Upload
ioan-muntean
View
4.351
Download
0
Tags:
Embed Size (px)
DESCRIPTION
An informal talk about genetic algorithms, numerical simulations and scientific discovery. March 2010, HPS, University of Leeds
Citation preview
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Numerical Simulations and Scientific Discoverypaper available at: http://papers.imuntean.net
Ioan Muntean
Department of Philosophyand
History and Philosophy of ScienceUniversity of Leeds
March 9, 2010
1
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Outline1 Philosophy of Numerical Simulations?
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
2 Genetic Algorithms in Numerical Simulations (GNS)Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
3 Philosophy of GNSWhat philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
4 FinisRisky conclusionsWeaker conclusions
5 References
2
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)
Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).
Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).
History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Definitions
Standard: “the use of a computer to build a model involvingequations that we cannot solve analytically”. (P. Humphreys,E. Winsberg)Non-Standard: NS track the dynamical evolution of realsystems (St. Hartmann).Broad: NS is a computational model that includes theequations of a model, assumptions, corrections,interpretations, justifications, and representations (Humphreys2004, 110).History prone NS: more fine-grained, historical distinctionsare needed (E. F. Keller).
I take them as “working definitions”
I agree that we need to be more philosophically nuanced
3
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:
the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:
the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:
the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:
the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:
the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:
the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:the microscope (I. Hacking),
the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What are they? What is their status?
Are numerical simulations similar to models? M. Morrison(2009): they have the same epistemic status)
Are NS mere experiments? E. Winsberg, W. Parker: they arenot
Are numerical experiments mere applications or spinoffs ofscientific theories? We’ll discuss this.
How do NS contribute to the progress of science? Not yet,not significantly.
Witness the philosophical importance of some scientific tools:the microscope (I. Hacking),the thermometer (H. Chang).
Why not a philosophy of NS?
4
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Mongrels and halfway houses
E. Winsberg: NS are “mongrels”between experiments and theories andhave features of both theories and ofexperiments, without being theories orexperiments.
S. Ulam (the father of the MonteCarlo method, late 1940s): NS are a“halfway house” between eleganttheory and experimental hardware(quoted in Keller 2003, 205)
5
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Mongrels and halfway houses
E. Winsberg: NS are “mongrels”between experiments and theories andhave features of both theories and ofexperiments, without being theories orexperiments.S. Ulam (the father of the MonteCarlo method, late 1940s): NS are a“halfway house” between eleganttheory and experimental hardware(quoted in Keller 2003, 205)
5
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Mongrels and halfway houses
E. Winsberg: NS are “mongrels”between experiments and theories andhave features of both theories and ofexperiments, without being theories orexperiments.S. Ulam (the father of the MonteCarlo method, late 1940s): NS are a“halfway house” between eleganttheory and experimental hardware(quoted in Keller 2003, 205)
5
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Mongrels and halfway houses
E. Winsberg: NS are “mongrels”between experiments and theories andhave features of both theories and ofexperiments, without being theories orexperiments.S. Ulam (the father of the MonteCarlo method, late 1940s): NS are a“halfway house” between eleganttheory and experimental hardware(quoted in Keller 2003, 205)
5
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The enthusiasts
It is thus reasonable to conclude that we are at thethreshold of an era of new scientific methodology. Inview of further technical developments in the near future,computer experts suggest that we are at present only atthe very beginning of this new era. [...] computersimulation offers a new tool for science: theoreticalmodel experiments of a scope and richness far exceedinganything available before. (Rohrlich 1990, 512,516)
Galison: NS constitute a new epistemology, as a new methodof extracting information from physical measurements, as wellas a new metaphysics that presupposed discrete entitiesinteracting through stochastic processes (Galison 1996, 120).
6
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The enthusiasts
It is thus reasonable to conclude that we are at thethreshold of an era of new scientific methodology. Inview of further technical developments in the near future,computer experts suggest that we are at present only atthe very beginning of this new era. [...] computersimulation offers a new tool for science: theoreticalmodel experiments of a scope and richness far exceedinganything available before. (Rohrlich 1990, 512,516)
Galison: NS constitute a new epistemology, as a new methodof extracting information from physical measurements, as wellas a new metaphysics that presupposed discrete entitiesinteracting through stochastic processes (Galison 1996, 120).
6
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Some skeptical stances
1 “Old stew in a new pot”: NS are not special for philosophy ofscience (Frigg and Reiss 2009; Stockler 2000).
2 “Wait-and-see”: We do not know what are the long-termconsequences of the NS
3 “Rage against the machine”: philosophical argumentspertaining to show that: “computers are dummy” “computerscannot create” etc.
7
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Some skeptical stances
1 “Old stew in a new pot”: NS are not special for philosophy ofscience (Frigg and Reiss 2009; Stockler 2000).
2 “Wait-and-see”: We do not know what are the long-termconsequences of the NS
3 “Rage against the machine”: philosophical argumentspertaining to show that: “computers are dummy” “computerscannot create” etc.
7
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Some skeptical stances
1 “Old stew in a new pot”: NS are not special for philosophy ofscience (Frigg and Reiss 2009; Stockler 2000).
2 “Wait-and-see”: We do not know what are the long-termconsequences of the NS
3 “Rage against the machine”: philosophical argumentspertaining to show that: “computers are dummy” “computerscannot create” etc.
7
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Some skeptical stances
1 “Old stew in a new pot”: NS are not special for philosophy ofscience (Frigg and Reiss 2009; Stockler 2000).
2 “Wait-and-see”: We do not know what are the long-termconsequences of the NS
3 “Rage against the machine”: philosophical argumentspertaining to show that: “computers are dummy” “computerscannot create” etc.
7
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Short answers to skeptics
My answer to 1: what is a novel philosophical problem? Werisk to get to “nothing new under the sun”
My answer to 2: philosophy of science is the history of futurescience (so to speak). Wait what? The next Ice Age? NS arehere, alive and kicking.
I focus on 3
It’s more subtle!
It has a respectable philosophical pedigree (Descartes, Leibniz,Kant)
It is mathematically and scientifically challenging (see Godel,Turing, J.R. Lucas)
8
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Short answers to skeptics
My answer to 1: what is a novel philosophical problem? Werisk to get to “nothing new under the sun”
My answer to 2: philosophy of science is the history of futurescience (so to speak). Wait what? The next Ice Age? NS arehere, alive and kicking.
I focus on 3
It’s more subtle!
It has a respectable philosophical pedigree (Descartes, Leibniz,Kant)
It is mathematically and scientifically challenging (see Godel,Turing, J.R. Lucas)
8
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Short answers to skeptics
My answer to 1: what is a novel philosophical problem? Werisk to get to “nothing new under the sun”
My answer to 2: philosophy of science is the history of futurescience (so to speak). Wait what? The next Ice Age? NS arehere, alive and kicking.
I focus on 3
It’s more subtle!
It has a respectable philosophical pedigree (Descartes, Leibniz,Kant)
It is mathematically and scientifically challenging (see Godel,Turing, J.R. Lucas)
8
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Short answers to skeptics
My answer to 1: what is a novel philosophical problem? Werisk to get to “nothing new under the sun”
My answer to 2: philosophy of science is the history of futurescience (so to speak). Wait what? The next Ice Age? NS arehere, alive and kicking.
I focus on 3
It’s more subtle!
It has a respectable philosophical pedigree (Descartes, Leibniz,Kant)
It is mathematically and scientifically challenging (see Godel,Turing, J.R. Lucas)
8
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Short answers to skeptics
My answer to 1: what is a novel philosophical problem? Werisk to get to “nothing new under the sun”
My answer to 2: philosophy of science is the history of futurescience (so to speak). Wait what? The next Ice Age? NS arehere, alive and kicking.
I focus on 3
It’s more subtle!
It has a respectable philosophical pedigree (Descartes, Leibniz,Kant)
It is mathematically and scientifically challenging (see Godel,Turing, J.R. Lucas)
8
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Short answers to skeptics
My answer to 1: what is a novel philosophical problem? Werisk to get to “nothing new under the sun”
My answer to 2: philosophy of science is the history of futurescience (so to speak). Wait what? The next Ice Age? NS arehere, alive and kicking.
I focus on 3
It’s more subtle!
It has a respectable philosophical pedigree (Descartes, Leibniz,Kant)
It is mathematically and scientifically challenging (see Godel,Turing, J.R. Lucas)
8
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 1: “Look! This is something new”?
For instance, if, rather than spilling much ink onconvincing ourselves that simulations are unlikeeverything else, we recognize that the epistemologicalproblems presented to us by simulations have much incommon with the ones that arise in connection withmodels, we can take the insights we gain in both fieldstogether and try to make progress in constructing thesought-after new epistemology. (Frigg and Reiss 2009,611).
With this, I agree
I will try to do something similar in last section
9
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 1: “Look! This is something new”?
For instance, if, rather than spilling much ink onconvincing ourselves that simulations are unlikeeverything else, we recognize that the epistemologicalproblems presented to us by simulations have much incommon with the ones that arise in connection withmodels, we can take the insights we gain in both fieldstogether and try to make progress in constructing thesought-after new epistemology. (Frigg and Reiss 2009,611).
With this, I agree
I will try to do something similar in last section
9
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 1: “Look! This is something new”?
For instance, if, rather than spilling much ink onconvincing ourselves that simulations are unlikeeverything else, we recognize that the epistemologicalproblems presented to us by simulations have much incommon with the ones that arise in connection withmodels, we can take the insights we gain in both fieldstogether and try to make progress in constructing thesought-after new epistemology. (Frigg and Reiss 2009,611).
With this, I agree
I will try to do something similar in last section
9
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 2: The Greek Chorus attitude
Wait and see who’s winning the battle. Do not hedge your betstoo early in the game. Philosophers are like Greek chorus, theycome at the end to explain the victory.
10
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 2: The Greek Chorus attitude
Wait and see who’s winning the battle. Do not hedge your betstoo early in the game. Philosophers are like Greek chorus, theycome at the end to explain the victory.
10
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 3: NS as “glorified slide rules”
claim C-1
A computer algorithm is no better than the assumptions which itwas built on
The Analytical Engine has no pretensions to originateanything. It can do whatever we know how to order it toperform. A letter of A. Lovelace quoted in (Hatree, 1949)
An argument
Computers do not thinkScience is a creative process that involves reason and skillsComputers do not contribute to progress (or discovery) in science
Computers are erring in anything, like slide rules do.
I call this argument and those of its ilk: the “glorified sliderules” argument
11
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Stance 3: NS as “glorified slide rules”
claim C-1
A computer algorithm is no better than the assumptions which itwas built on
The Analytical Engine has no pretensions to originateanything. It can do whatever we know how to order it toperform. A letter of A. Lovelace quoted in (Hatree, 1949)
An argument
Computers do not thinkScience is a creative process that involves reason and skillsComputers do not contribute to progress (or discovery) in science
Computers are erring in anything, like slide rules do.I call this argument and those of its ilk: the “glorified sliderules” argument
11
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Consequences of the glorified slide rules argument
Demoting NS again
Whenever the analytic solution is discovered, a real experimentis possible or new data is available, NS can be tossed away.
Nothing that NS have achieved could not have been done byan “army of well-trained scientists working with slide rules”.
12
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
Consequences of the glorified slide rules argument
Demoting NS again
Whenever the analytic solution is discovered, a real experimentis possible or new data is available, NS can be tossed away.
Nothing that NS have achieved could not have been done byan “army of well-trained scientists working with slide rules”.
12
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:
NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).
NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;
NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,
there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,
computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.
Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
The inherently limited nature of NS
NS are subordinate in their nature because they do providenovel scientificknowledge only when other, more rigorous ways ofrepresenting the real world fail:NS are unreal because unlike experiments and models, theydo not latch directly onto reality
NS lack materiality; Materiality, maybe the most relevant, isdiscussed in (Parker 2009).NS bear no causal connection to the world;NS are very brute idealizations (Parker, M. Morgan2005-2009) etc. argue for or against some of these.
NS are fundamentally flawed.
computers cannot simulate the continuum quantities ofphysics,there are inherent errors of digitization and,computer arithmetic is fundamentally limited by Godel’sincompleteness theorem.Hence: mathematics can show us what “machines cannot inprinciple do”.
13
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
science and NS
NS are not able to falsify or confirm scientific theories;
NS do not explain
NS do not augment scientific knowledge.
NS are limited predicting tools, at best, and only when apre-existing theoretical model permits it.
Science is about explanation/understanding/unification etc.
No philosophy of slide rules
In philosophy of science, no country for old slide rules
14
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
A rejoinder to the “glorified slide rules” arguments
How do we argue against the “glorified slide rules” arguments?
1 Deny C-1: computers do help us understanding and explainbecause they show us how to decompose systems, separatelevels and see the organization of mechanisms. (Simon 1969)
2 Show that historically it is inaccurate (Keller, 2003): CellularAutomata (CA), Neural Networks (NN) and GeneticAlgorithms (GA) are counterexamples to C2. (CA illustratethe third stage in Keller).
3 Quantum computing may be the “best” candidate forsurpassing C-1.
Here I combine 1 and 2, but insist on the paradigm shift a laKeller.
15
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
A rejoinder to the “glorified slide rules” arguments
How do we argue against the “glorified slide rules” arguments?
1 Deny C-1: computers do help us understanding and explainbecause they show us how to decompose systems, separatelevels and see the organization of mechanisms. (Simon 1969)
2 Show that historically it is inaccurate (Keller, 2003): CellularAutomata (CA), Neural Networks (NN) and GeneticAlgorithms (GA) are counterexamples to C2. (CA illustratethe third stage in Keller).
3 Quantum computing may be the “best” candidate forsurpassing C-1.
Here I combine 1 and 2, but insist on the paradigm shift a laKeller.
15
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
A rejoinder to the “glorified slide rules” arguments
How do we argue against the “glorified slide rules” arguments?
1 Deny C-1: computers do help us understanding and explainbecause they show us how to decompose systems, separatelevels and see the organization of mechanisms. (Simon 1969)
2 Show that historically it is inaccurate (Keller, 2003): CellularAutomata (CA), Neural Networks (NN) and GeneticAlgorithms (GA) are counterexamples to C2. (CA illustratethe third stage in Keller).
3 Quantum computing may be the “best” candidate forsurpassing C-1.
Here I combine 1 and 2, but insist on the paradigm shift a laKeller.
15
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
A rejoinder to the “glorified slide rules” arguments
How do we argue against the “glorified slide rules” arguments?
1 Deny C-1: computers do help us understanding and explainbecause they show us how to decompose systems, separatelevels and see the organization of mechanisms. (Simon 1969)
2 Show that historically it is inaccurate (Keller, 2003): CellularAutomata (CA), Neural Networks (NN) and GeneticAlgorithms (GA) are counterexamples to C2. (CA illustratethe third stage in Keller).
3 Quantum computing may be the “best” candidate forsurpassing C-1.
Here I combine 1 and 2, but insist on the paradigm shift a laKeller.
15
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
A rejoinder to the “glorified slide rules” arguments
How do we argue against the “glorified slide rules” arguments?
1 Deny C-1: computers do help us understanding and explainbecause they show us how to decompose systems, separatelevels and see the organization of mechanisms. (Simon 1969)
2 Show that historically it is inaccurate (Keller, 2003): CellularAutomata (CA), Neural Networks (NN) and GeneticAlgorithms (GA) are counterexamples to C2. (CA illustratethe third stage in Keller).
3 Quantum computing may be the “best” candidate forsurpassing C-1.
Here I combine 1 and 2, but insist on the paradigm shift a laKeller.
15
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,
find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,
discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
What do i argue for?
Pace Frigg&Reiss, there is philosophical novelty in NS.
Not all NS are “dumb slide rules”; Some are more interestingthan others.
More attention to the historical developments of NS (a laKeller and Galison)
Some NS are able to:
build models,find invariants,discover non-trivial conserved quantities etc.
The NS under scrutiny here are: Genetic Algorithms
16
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Outline1 Philosophy of Numerical Simulations?
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
2 Genetic Algorithms in Numerical Simulations (GNS)Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
3 Philosophy of GNSWhat philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
4 FinisRisky conclusionsWeaker conclusions
5 References
17
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochasticity and NS
Biomimetics
Q-1: How can computers be made to do what needs to be done,without being told exactly how to do it?
The “glorified slide rules” argument uses the Turing machineparadigm.
Are all machine Turing machines?
Build machines inspired by learning, discovery, game playing,solving real-life problems, etc.
Hence adopt biomimetics
1 Go stochastic in building algorithms!2 Go Darwinian in programming computers
18
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochasticity and NS
Biomimetics
Q-1: How can computers be made to do what needs to be done,without being told exactly how to do it?
The “glorified slide rules” argument uses the Turing machineparadigm.
Are all machine Turing machines?
Build machines inspired by learning, discovery, game playing,solving real-life problems, etc.
Hence adopt biomimetics
1 Go stochastic in building algorithms!2 Go Darwinian in programming computers
18
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochasticity and NS
Biomimetics
Q-1: How can computers be made to do what needs to be done,without being told exactly how to do it?
The “glorified slide rules” argument uses the Turing machineparadigm.
Are all machine Turing machines?
Build machines inspired by learning, discovery, game playing,solving real-life problems, etc.
Hence adopt biomimetics
1 Go stochastic in building algorithms!2 Go Darwinian in programming computers
18
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochasticity and NS
Biomimetics
Q-1: How can computers be made to do what needs to be done,without being told exactly how to do it?
The “glorified slide rules” argument uses the Turing machineparadigm.
Are all machine Turing machines?
Build machines inspired by learning, discovery, game playing,solving real-life problems, etc.
Hence adopt biomimetics
1 Go stochastic in building algorithms!2 Go Darwinian in programming computers
18
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochasticity and NS
Biomimetics
Q-1: How can computers be made to do what needs to be done,without being told exactly how to do it?
The “glorified slide rules” argument uses the Turing machineparadigm.
Are all machine Turing machines?
Build machines inspired by learning, discovery, game playing,solving real-life problems, etc.
Hence adopt biomimetics
1 Go stochastic in building algorithms!
2 Go Darwinian in programming computers
18
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochasticity and NS
Biomimetics
Q-1: How can computers be made to do what needs to be done,without being told exactly how to do it?
The “glorified slide rules” argument uses the Turing machineparadigm.
Are all machine Turing machines?
Build machines inspired by learning, discovery, game playing,solving real-life problems, etc.
Hence adopt biomimetics
1 Go stochastic in building algorithms!2 Go Darwinian in programming computers
18
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Do philosophers talk about non-Turing machines?
CA, GA and NN can go beyond what a Turing machine isable to do. So did the Monte Carlo method (first NS).
Are these solutions philosophically attractive?
The literature on NS ignores GA
There are interesting discussions on CA and NN. (Keller,2003) (Barberousse, Franceschelli, and Imbert 2007) and,more dogmatically, (Wolfram 2002).
The literature on NN is well-known to philosophers: (Pauland Patricia Churchland)
I focus here on GA and GP
19
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Do philosophers talk about non-Turing machines?
CA, GA and NN can go beyond what a Turing machine isable to do. So did the Monte Carlo method (first NS).
Are these solutions philosophically attractive?
The literature on NS ignores GA
There are interesting discussions on CA and NN. (Keller,2003) (Barberousse, Franceschelli, and Imbert 2007) and,more dogmatically, (Wolfram 2002).
The literature on NN is well-known to philosophers: (Pauland Patricia Churchland)
I focus here on GA and GP
19
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Do philosophers talk about non-Turing machines?
CA, GA and NN can go beyond what a Turing machine isable to do. So did the Monte Carlo method (first NS).
Are these solutions philosophically attractive?
The literature on NS ignores GA
There are interesting discussions on CA and NN. (Keller,2003) (Barberousse, Franceschelli, and Imbert 2007) and,more dogmatically, (Wolfram 2002).
The literature on NN is well-known to philosophers: (Pauland Patricia Churchland)
I focus here on GA and GP
19
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Do philosophers talk about non-Turing machines?
CA, GA and NN can go beyond what a Turing machine isable to do. So did the Monte Carlo method (first NS).
Are these solutions philosophically attractive?
The literature on NS ignores GA
There are interesting discussions on CA and NN. (Keller,2003) (Barberousse, Franceschelli, and Imbert 2007) and,more dogmatically, (Wolfram 2002).
The literature on NN is well-known to philosophers: (Pauland Patricia Churchland)
I focus here on GA and GP
19
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Do philosophers talk about non-Turing machines?
CA, GA and NN can go beyond what a Turing machine isable to do. So did the Monte Carlo method (first NS).
Are these solutions philosophically attractive?
The literature on NS ignores GA
There are interesting discussions on CA and NN. (Keller,2003) (Barberousse, Franceschelli, and Imbert 2007) and,more dogmatically, (Wolfram 2002).
The literature on NN is well-known to philosophers: (Pauland Patricia Churchland)
I focus here on GA and GP
19
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Do philosophers talk about non-Turing machines?
CA, GA and NN can go beyond what a Turing machine isable to do. So did the Monte Carlo method (first NS).
Are these solutions philosophically attractive?
The literature on NS ignores GA
There are interesting discussions on CA and NN. (Keller,2003) (Barberousse, Franceschelli, and Imbert 2007) and,more dogmatically, (Wolfram 2002).
The literature on NN is well-known to philosophers: (Pauland Patricia Churchland)
I focus here on GA and GP
19
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,
the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation and
natural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Evolution of algorithms
Speculated by Turing in 1948.
Based on genetic or evolutionary search by which a“combination of genes is looked for, the criterion being thesurvival value”.
Turing in Mind (1950): “the child-machine needs to be taughtand surveyed. Then another child-machine tried andcompared to the first etc.”
the child machine = hereditary material,the changes within it = genetic mutation andnatural selection = “judgment of the experimenter”
In a unpublished paper, Turing realized that such a geneticsearch implied randomness (Turing 1996).
20
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
The 1970s and 1980s
Alien (1979)Aliens (1986)Alien (1992)Alien Resurrection (1997)ABBA’s “Take a chance on me”Koza, Holland et al.: Birth of the Genetic Programming andGenetic Algorithms: 1986 to 1995
21
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.
Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproduced
Offspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.
The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.
GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolution
Optimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteria
Unlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
J. Holland’s genetic programming
Starts from a given number of initial programs randomlydistributed in a given space of solutions.Based on relative results, the best competitors are chosen andreproducedOffspring have some (randomly chosen) features of thepredecessors.The best competitor wins and constitutes the solutions of theproblem.GA are implementations of the biological evolutionOptimization: searching for the best solution is based on somepre-established criteriaUnlike in Turing, selection occurs at the level of population,not at the level of individual algorithms.
Adaptation in Natural and Artificial Systems (1975)
22
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Stochastic algorithms
Output is manifestly stochastic.
23
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Developments in GA
Crossover takes two individuals (parents) and produces twonew individuals, the offspring,
by swapping parts of the parents (the simplest crossoveroperator exchanges substrings of the parents)
Crossover moves the search towards those regions of thesearch space in which results are most likely to befound.(Tomassini 1995)
Mutation is a background redistribution of strings to preventpremature convergence to local optima.
Termination condition: when the sought-for level of optimalityis reached or when all the solutions converge to one candidate
24
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Developments in GA
Crossover takes two individuals (parents) and produces twonew individuals, the offspring,
by swapping parts of the parents (the simplest crossoveroperator exchanges substrings of the parents)
Crossover moves the search towards those regions of thesearch space in which results are most likely to befound.(Tomassini 1995)
Mutation is a background redistribution of strings to preventpremature convergence to local optima.
Termination condition: when the sought-for level of optimalityis reached or when all the solutions converge to one candidate
24
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Developments in GA
Crossover takes two individuals (parents) and produces twonew individuals, the offspring,
by swapping parts of the parents (the simplest crossoveroperator exchanges substrings of the parents)
Crossover moves the search towards those regions of thesearch space in which results are most likely to befound.(Tomassini 1995)
Mutation is a background redistribution of strings to preventpremature convergence to local optima.
Termination condition: when the sought-for level of optimalityis reached or when all the solutions converge to one candidate
24
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Developments in GA
Crossover takes two individuals (parents) and produces twonew individuals, the offspring,
by swapping parts of the parents (the simplest crossoveroperator exchanges substrings of the parents)
Crossover moves the search towards those regions of thesearch space in which results are most likely to befound.(Tomassini 1995)
Mutation is a background redistribution of strings to preventpremature convergence to local optima.
Termination condition: when the sought-for level of optimalityis reached or when all the solutions converge to one candidate
24
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Developments in GA
Crossover takes two individuals (parents) and produces twonew individuals, the offspring,
by swapping parts of the parents (the simplest crossoveroperator exchanges substrings of the parents)
Crossover moves the search towards those regions of thesearch space in which results are most likely to befound.(Tomassini 1995)
Mutation is a background redistribution of strings to preventpremature convergence to local optima.
Termination condition: when the sought-for level of optimalityis reached or when all the solutions converge to one candidate
24
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Results of GA/GP
GA
are able to reinvent patented inventions: the negativefeedback (Black 1927)
are able to find algorithms in quantum computing–toocomplicated to be discovered by humans.
find patterns in the EEG before the epileptic seizure (butcan’t predict it!)
find more interesting patterns before earthquakes (but too lateand in general without impact)
25
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Results of GA/GP
GA
are able to reinvent patented inventions: the negativefeedback (Black 1927)
are able to find algorithms in quantum computing–toocomplicated to be discovered by humans.
find patterns in the EEG before the epileptic seizure (butcan’t predict it!)
find more interesting patterns before earthquakes (but too lateand in general without impact)
25
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Results of GA/GP
GA
are able to reinvent patented inventions: the negativefeedback (Black 1927)
are able to find algorithms in quantum computing–toocomplicated to be discovered by humans.
find patterns in the EEG before the epileptic seizure (butcan’t predict it!)
find more interesting patterns before earthquakes (but too lateand in general without impact)
25
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Results of GA/GP
GA
are able to reinvent patented inventions: the negativefeedback (Black 1927)
are able to find algorithms in quantum computing–toocomplicated to be discovered by humans.
find patterns in the EEG before the epileptic seizure (butcan’t predict it!)
find more interesting patterns before earthquakes (but too lateand in general without impact)
25
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
A simple result
The algorithm takes pairs of numbers between −1 and +1that fit perfectly the quadratic polynomial x2 + x + 1 and aset of elementary or “desired” operators F = {+;−;×; / }used as primitive of the symbolic regression.
The initial population randomly constructed had theindividuals: x + 1; x2 + 1; 2; x and were fitted with the databased on a classical metric: the area covered by the curverepresented by the competitor and the given set of data.
After the second generation and after applying the crossoveroperator, the result was as expected x2 + x + 1.
26
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
A simple result
The algorithm takes pairs of numbers between −1 and +1that fit perfectly the quadratic polynomial x2 + x + 1 and aset of elementary or “desired” operators F = {+;−;×; / }used as primitive of the symbolic regression.
The initial population randomly constructed had theindividuals: x + 1; x2 + 1; 2; x and were fitted with the databased on a classical metric: the area covered by the curverepresented by the competitor and the given set of data.
After the second generation and after applying the crossoveroperator, the result was as expected x2 + x + 1.
26
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
A simple result
The algorithm takes pairs of numbers between −1 and +1that fit perfectly the quadratic polynomial x2 + x + 1 and aset of elementary or “desired” operators F = {+;−;×; / }used as primitive of the symbolic regression.
The initial population randomly constructed had theindividuals: x + 1; x2 + 1; 2; x and were fitted with the databased on a classical metric: the area covered by the curverepresented by the competitor and the given set of data.
After the second generation and after applying the crossoveroperator, the result was as expected x2 + x + 1.
26
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
A simple result
The algorithm takes pairs of numbers between −1 and +1that fit perfectly the quadratic polynomial x2 + x + 1 and aset of elementary or “desired” operators F = {+;−;×; / }used as primitive of the symbolic regression.
The initial population randomly constructed had theindividuals: x + 1; x2 + 1; 2; x and were fitted with the databased on a classical metric: the area covered by the curverepresented by the competitor and the given set of data.
After the second generation and after applying the crossoveroperator, the result was as expected x2 + x + 1.
26
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
GA in a snapshot
produce an initial population of individuals
while termination condition not met do
evaluate the fitness of all individuals
select fitter individuals for reproduction
produce new individuals
generate a new population by inserting some new good
individuals and by discarding some old/bad
individuals
mutate some individuals
endwhile
27
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
GA in a snapshot
produce an initial population of individuals
while termination condition not met do
evaluate the fitness of all individuals
select fitter individuals for reproduction
produce new individuals
generate a new population by inserting some new good
individuals and by discarding some old/bad
individuals
mutate some individuals
endwhile
27
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Problems of GA
GA are sensitive to the initial population. For a different set ofinitial population, the generation can take several iterations, butfor simple enough expressions, the GA converge
28
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON, the software
BACON is a software best on induction (1980s by H. Simon,P. Langley, J. Shrager, etc)
Other software: OCCAM, GALILEO, HUYGENS.
BACON rediscovered Kepler’s laws, Prout’s hypothesis aboutatomic structure, etc.
Bacon had a bad reception among philosophers, despiteSimon’s arguments (Simon, 1992).
BACON is marred with the problem of induction, does notexplain, does not help us with understanding.
It’s based on a totally weak analogy.
It threatens rationality of science
Does not explain and help us with understanding science.
29
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Planck’s discoveryFrom:
To:U(ν) = U(ω)dωdν = 8π
c3hν3
e~ω/kT−1
30
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON against the roulette (skip)
GA score better than BACON because they avoid localoptima.
Too complicated to discuss here.
31
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
BACON against the roulette (skip)
GA score better than BACON because they avoid localoptima.
Too complicated to discuss here.
31
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
ADAM, the robot scientist (skip)
R.D. King coined this term in 2004.
ADAM is a new software based on inductive reasoning. Itidentifies genes encoding orphan enzymes in Saccharomycescerevisiae for which the encoding gene(s) are not known
A robot scientist automatically originates hypothesesto explain observations, devises experiments to testthese hypotheses, physically runs the experiments byusing laboratory robotics, interprets the results, andthen repeats the cycle.
Limitations of Adam: the scientific knowledge “discovered” byAdam is implicit in the formulation of the problem and istherefore not novel.
We accept that the knowledge automatically generatedby Adam is of a modest kind. However, this knowledge isnot trivial, and in the case of the genes encoding 2A2OA,it sheds light on, and perhaps solves, a 50-year-oldpuzzle. (King et al. 2009, 88)
32
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
ADAM, the robot scientist (skip)
R.D. King coined this term in 2004.ADAM is a new software based on inductive reasoning. Itidentifies genes encoding orphan enzymes in Saccharomycescerevisiae for which the encoding gene(s) are not known
A robot scientist automatically originates hypothesesto explain observations, devises experiments to testthese hypotheses, physically runs the experiments byusing laboratory robotics, interprets the results, andthen repeats the cycle.
Limitations of Adam: the scientific knowledge “discovered” byAdam is implicit in the formulation of the problem and istherefore not novel.
We accept that the knowledge automatically generatedby Adam is of a modest kind. However, this knowledge isnot trivial, and in the case of the genes encoding 2A2OA,it sheds light on, and perhaps solves, a 50-year-oldpuzzle. (King et al. 2009, 88)
32
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
ADAM, the robot scientist (skip)
R.D. King coined this term in 2004.ADAM is a new software based on inductive reasoning. Itidentifies genes encoding orphan enzymes in Saccharomycescerevisiae for which the encoding gene(s) are not known
A robot scientist automatically originates hypothesesto explain observations, devises experiments to testthese hypotheses, physically runs the experiments byusing laboratory robotics, interprets the results, andthen repeats the cycle.
Limitations of Adam: the scientific knowledge “discovered” byAdam is implicit in the formulation of the problem and istherefore not novel.
We accept that the knowledge automatically generatedby Adam is of a modest kind. However, this knowledge isnot trivial, and in the case of the genes encoding 2A2OA,it sheds light on, and perhaps solves, a 50-year-oldpuzzle. (King et al. 2009, 88)
32
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
ADAM, the robot scientist (skip)
R.D. King coined this term in 2004.ADAM is a new software based on inductive reasoning. Itidentifies genes encoding orphan enzymes in Saccharomycescerevisiae for which the encoding gene(s) are not known
A robot scientist automatically originates hypothesesto explain observations, devises experiments to testthese hypotheses, physically runs the experiments byusing laboratory robotics, interprets the results, andthen repeats the cycle.
Limitations of Adam: the scientific knowledge “discovered” byAdam is implicit in the formulation of the problem and istherefore not novel.
We accept that the knowledge automatically generatedby Adam is of a modest kind. However, this knowledge isnot trivial, and in the case of the genes encoding 2A2OA,it sheds light on, and perhaps solves, a 50-year-oldpuzzle. (King et al. 2009, 88)
32
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
ADAM, the robot scientist (skip)
R.D. King coined this term in 2004.ADAM is a new software based on inductive reasoning. Itidentifies genes encoding orphan enzymes in Saccharomycescerevisiae for which the encoding gene(s) are not known
A robot scientist automatically originates hypothesesto explain observations, devises experiments to testthese hypotheses, physically runs the experiments byusing laboratory robotics, interprets the results, andthen repeats the cycle.
Limitations of Adam: the scientific knowledge “discovered” byAdam is implicit in the formulation of the problem and istherefore not novel.
We accept that the knowledge automatically generatedby Adam is of a modest kind. However, this knowledge isnot trivial, and in the case of the genes encoding 2A2OA,it sheds light on, and perhaps solves, a 50-year-oldpuzzle. (King et al. 2009, 88)
32
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
What’s GA Got To Do With NS?
There are genetic numerical algorithms ( GNS)
Best example I know of is (Schmidt&Lipson 2009)
GNS do not output data, but symbolic representations, i.e.laws and mathematical expressions from data based on GA.
It may have a direct impact on science in the future.(remember I started by rejecting the the wait-and-see attitude)
I also want to defuse blind optimism
33
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
What’s GA Got To Do With NS?
There are genetic numerical algorithms ( GNS)
Best example I know of is (Schmidt&Lipson 2009)
GNS do not output data, but symbolic representations, i.e.laws and mathematical expressions from data based on GA.
It may have a direct impact on science in the future.(remember I started by rejecting the the wait-and-see attitude)
I also want to defuse blind optimism
33
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
What’s GA Got To Do With NS?
There are genetic numerical algorithms ( GNS)
Best example I know of is (Schmidt&Lipson 2009)
GNS do not output data, but symbolic representations, i.e.laws and mathematical expressions from data based on GA.
It may have a direct impact on science in the future.(remember I started by rejecting the the wait-and-see attitude)
I also want to defuse blind optimism
33
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
What’s GA Got To Do With NS?
There are genetic numerical algorithms ( GNS)
Best example I know of is (Schmidt&Lipson 2009)
GNS do not output data, but symbolic representations, i.e.laws and mathematical expressions from data based on GA.
It may have a direct impact on science in the future.(remember I started by rejecting the the wait-and-see attitude)
I also want to defuse blind optimism
33
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
What’s GA Got To Do With NS?
There are genetic numerical algorithms ( GNS)
Best example I know of is (Schmidt&Lipson 2009)
GNS do not output data, but symbolic representations, i.e.laws and mathematical expressions from data based on GA.
It may have a direct impact on science in the future.(remember I started by rejecting the the wait-and-see attitude)
I also want to defuse blind optimism
33
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
What’s GA Got To Do With NS?
There are genetic numerical algorithms ( GNS)
Best example I know of is (Schmidt&Lipson 2009)
GNS do not output data, but symbolic representations, i.e.laws and mathematical expressions from data based on GA.
It may have a direct impact on science in the future.(remember I started by rejecting the the wait-and-see attitude)
I also want to defuse blind optimism
33
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
GNS and meaningfulness
Here things are getting really exciting:
These GNS can discover : Hamiltonians, Lagrangians, laws ofconservation, symmetries, and other invariants.
Based on meaningfulness and interestingness.
34
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
GNS and meaningfulness
Here things are getting really exciting:
These GNS can discover : Hamiltonians, Lagrangians, laws ofconservation, symmetries, and other invariants.
Based on meaningfulness and interestingness.
34
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
GNS and meaningfulness
Here things are getting really exciting:
These GNS can discover : Hamiltonians, Lagrangians, laws ofconservation, symmetries, and other invariants.
Based on meaningfulness and interestingness.
34
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
GNS and meaningfulness
Here things are getting really exciting:
These GNS can discover : Hamiltonians, Lagrangians, laws ofconservation, symmetries, and other invariants.
Based on meaningfulness and interestingness.
34
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
A simple illustration of a “best expression”
35
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Schmidt and Lipson about their result
We have demonstrated the discovery of physical laws,from scratch, directly from experimentally captured datawith the use of a computational search [...] detectnonlinear energy conservation laws, Newtonian forcelaws, geometric invariants, and system manifolds invarious synthetic and physically implemented systemswithout prior knowledge about physics, kinematics, orgeometry. The concise analytical expressions that wefound are amenable to human interpretation and help toreveal the physics underlying the observed phenomenon.[...] Might this process diminish the role of futurescientists? Quite the contrary: Scientists may useprocesses such as this to help focus on interestingphenomena more rapidly and to interpret their meaning.
(Schmidt and Lipson 2009, 85)
36
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Can BACON and ADAM compete with the roulette? (skip)
difficult to compare directly.
But GNS score better than BACON because they avoid localoptima.
Too complicated to discuss here.
37
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Can BACON and ADAM compete with the roulette? (skip)
difficult to compare directly.
But GNS score better than BACON because they avoid localoptima.
Too complicated to discuss here.
37
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
Can BACON and ADAM compete with the roulette? (skip)
difficult to compare directly.
But GNS score better than BACON because they avoid localoptima.
Too complicated to discuss here.
37
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Outline1 Philosophy of Numerical Simulations?
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
2 Genetic Algorithms in Numerical Simulations (GNS)Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
3 Philosophy of GNSWhat philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
4 FinisRisky conclusionsWeaker conclusions
5 References
38
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!
In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.
the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!
There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”
If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experiments
In the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Are GNS philosophically attractive?
In the framework of the debate about the “philosophy of NS”they are!In GNS we simulate:
the process of scientific discovery.the relation between experiments, theories and models.
With GNS, we do not rely on theories, we discover them!
It is prima facie a bottom-up view!There is literally a guesswork in GNS, but:
It optimizes the result based on meaning and “interestingness”If in science we transform data into phenomena (Woodwardand Bogen), then GNS qualifies as a possible candidatetogether with experimentsIn the context of discovery, science is, for better or worse,guesswork
39
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Creativity, Luck and Chance in GNS
In the context of scientific discovery, yes GNS arephilosophically interesting
GNS make room to chance in scientific discovery.
Novelty and chance may be stronger related.
As a biological metaphor, GNS are up to their neck instochasticity.
How is chance (as part of a search procedure) related tocreativity?
F. Crick “chance is the only source of true novelty” (Crick1981, 58).
Maybe scientific discovery is in fact closer to playing a gameand hedging one’s bets.
40
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
How inventive are we?
We are extremely inventive beings! Can you read this?
To xllxstxatx, I cxn rxplxce xvexy txirx lextex of x sextexce xitx anx, anx yox stxll xan xanxge xo rxad xt wixh sxme xifxicxltx.
Then be happy, because no Turing machine can!
Searching patterns and comparing them is a difficult business.
41
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
How inventive are we?
We are extremely inventive beings! Can you read this?
To xllxstxatx, I cxn rxplxce xvexy txirx lextex of x sextexce xitx anx, anx yox stxll xan xanxge xo rxad xt wixh sxme xifxicxltx.
Then be happy, because no Turing machine can!
Searching patterns and comparing them is a difficult business.
41
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
How inventive are we?
We are extremely inventive beings! Can you read this?
To xllxstxatx, I cxn rxplxce xvexy txirx lextex of x sextexce xitx anx, anx yox stxll xan xanxge xo rxad xt wixh sxme xifxicxltx.
Then be happy, because no Turing machine can!
Searching patterns and comparing them is a difficult business.
41
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Stochasticism (a philosophical view)
Maybe the nature is stochastic/chancy (we don’t know, butwe have reasons to think so)
Our representation of the world should mimic it
Then we’d better go stochastic in modeling the world
If the world is not stochastic, no big fuss: as humans we takechances: we are built to be stochastic (gamble, guess)
We can go intuitionistic in mathematics (as opposed toPlatonism, formalism,etc.)
42
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Stochasticism (a philosophical view)
Maybe the nature is stochastic/chancy (we don’t know, butwe have reasons to think so)
Our representation of the world should mimic it
Then we’d better go stochastic in modeling the world
If the world is not stochastic, no big fuss: as humans we takechances: we are built to be stochastic (gamble, guess)
We can go intuitionistic in mathematics (as opposed toPlatonism, formalism,etc.)
42
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Stochasticism (a philosophical view)
Maybe the nature is stochastic/chancy (we don’t know, butwe have reasons to think so)
Our representation of the world should mimic it
Then we’d better go stochastic in modeling the world
If the world is not stochastic, no big fuss: as humans we takechances: we are built to be stochastic (gamble, guess)
We can go intuitionistic in mathematics (as opposed toPlatonism, formalism,etc.)
42
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Stochasticism (a philosophical view)
Maybe the nature is stochastic/chancy (we don’t know, butwe have reasons to think so)
Our representation of the world should mimic it
Then we’d better go stochastic in modeling the world
If the world is not stochastic, no big fuss: as humans we takechances: we are built to be stochastic (gamble, guess)
We can go intuitionistic in mathematics (as opposed toPlatonism, formalism,etc.)
42
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Stochasticism (a philosophical view)
Maybe the nature is stochastic/chancy (we don’t know, butwe have reasons to think so)
Our representation of the world should mimic it
Then we’d better go stochastic in modeling the world
If the world is not stochastic, no big fuss: as humans we takechances: we are built to be stochastic (gamble, guess)
We can go intuitionistic in mathematics (as opposed toPlatonism, formalism,etc.)
42
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Stochasticism (a philosophical view)
Maybe the nature is stochastic/chancy (we don’t know, butwe have reasons to think so)
Our representation of the world should mimic it
Then we’d better go stochastic in modeling the world
If the world is not stochastic, no big fuss: as humans we takechances: we are built to be stochastic (gamble, guess)
We can go intuitionistic in mathematics (as opposed toPlatonism, formalism,etc.)
42
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Argument from stochasticity
world/science The world is:
deterministic non-deterministic
Non-stochastic Long Live Laplace! Ooops! We are off!!
Stochastic It is OK1 The best combination
1It’s ok, it’s just a representation. Some deterministic systems are betterrepresented by statistical models
43
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS as better “epistemic enhancers” (Humphreys)
P. Humphreys thinks that NS in general are epistemicenhancers.
They extend ourselves
If so, GNS pushes the limits farther than ordinary NS.Pushes the limits of what is “scientifically discoverable”
44
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS as better “epistemic enhancers” (Humphreys)
P. Humphreys thinks that NS in general are epistemicenhancers.
They extend ourselves
If so, GNS pushes the limits farther than ordinary NS.Pushes the limits of what is “scientifically discoverable”
44
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS as better “epistemic enhancers” (Humphreys)
P. Humphreys thinks that NS in general are epistemicenhancers.
They extend ourselves
If so, GNS pushes the limits farther than ordinary NS.Pushes the limits of what is “scientifically discoverable”
44
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS as better “epistemic enhancers” (Humphreys)
P. Humphreys thinks that NS in general are epistemicenhancers.
They extend ourselves
If so, GNS pushes the limits farther than ordinary NS.
Pushes the limits of what is “scientifically discoverable”
44
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS as better “epistemic enhancers” (Humphreys)
P. Humphreys thinks that NS in general are epistemicenhancers.
They extend ourselves
If so, GNS pushes the limits farther than ordinary NS.Pushes the limits of what is “scientifically discoverable”
44
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Mathematics as a constraint?
Think of the “unreasonable effectiveness of mathematics”argument (Steiner, Colyvain, etc)
Parenthetically, I have doubts about this argument.
In some cases mathematics is not the driving force but theconstraint.
GNS illustrates the mathematics as “constraint”: it generatessymbolic expression from data based on mathematicalconstraints
First you need to decide what limits to put to yourmathematics, and then GNS delivers an optimal result
45
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Mathematics as a constraint?
Think of the “unreasonable effectiveness of mathematics”argument (Steiner, Colyvain, etc)
Parenthetically, I have doubts about this argument.
In some cases mathematics is not the driving force but theconstraint.
GNS illustrates the mathematics as “constraint”: it generatessymbolic expression from data based on mathematicalconstraints
First you need to decide what limits to put to yourmathematics, and then GNS delivers an optimal result
45
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Mathematics as a constraint?
Think of the “unreasonable effectiveness of mathematics”argument (Steiner, Colyvain, etc)
Parenthetically, I have doubts about this argument.
In some cases mathematics is not the driving force but theconstraint.
GNS illustrates the mathematics as “constraint”: it generatessymbolic expression from data based on mathematicalconstraints
First you need to decide what limits to put to yourmathematics, and then GNS delivers an optimal result
45
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Mathematics as a constraint?
Think of the “unreasonable effectiveness of mathematics”argument (Steiner, Colyvain, etc)
Parenthetically, I have doubts about this argument.
In some cases mathematics is not the driving force but theconstraint.
GNS illustrates the mathematics as “constraint”: it generatessymbolic expression from data based on mathematicalconstraints
First you need to decide what limits to put to yourmathematics, and then GNS delivers an optimal result
45
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Mathematics as a constraint?
Think of the “unreasonable effectiveness of mathematics”argument (Steiner, Colyvain, etc)
Parenthetically, I have doubts about this argument.
In some cases mathematics is not the driving force but theconstraint.
GNS illustrates the mathematics as “constraint”: it generatessymbolic expression from data based on mathematicalconstraints
First you need to decide what limits to put to yourmathematics, and then GNS delivers an optimal result
45
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Mathematics as a constraint?
Think of the “unreasonable effectiveness of mathematics”argument (Steiner, Colyvain, etc)
Parenthetically, I have doubts about this argument.
In some cases mathematics is not the driving force but theconstraint.
GNS illustrates the mathematics as “constraint”: it generatessymbolic expression from data based on mathematicalconstraints
First you need to decide what limits to put to yourmathematics, and then GNS delivers an optimal result
45
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worlds
C’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worlds
C’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Conceivability
Any experiment explores possible worldsC’mon, that’s trite
Science is not only about the actual, but about the possible.
The typical way to explore possibility: change initialconditions of an experiment but keep the laws.
Is science indeed about the possible? Maybe more aboutcounterfactuals.
That’s a class of nomical worlds based on physical possibilities
Science teaches us that the world is richer that we canconceive with the “naked mind” (Van Fraassen)
My argument is that NS and GNS, enhance the mind bypushing the limits of conceivability.
46
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
What is really new?
Strictly speaking, data are new.
The way we discover it.
47
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
What is really new?
Strictly speaking, data are new.
The way we discover it.
47
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
What is really new?
Strictly speaking, data are new.
The way we discover it.
47
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)
My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancy
It’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)
It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)
You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
GNS and possibilia
The “observable”
The “physical”
The “logical”
The “conceivable”
Other less discussed possibilia
The “computable” (a la Church-Turing). But is physicscomputable? (some systems in condensed matter physics arenot NP-computable)My proposal: the “evolvable” (a la GNS)
The only one that’s chancyIt’s natural (maybe too literally)It’s always open to re-runs (not replicable)You can always hope for a better solution
48
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;General Relativity has too many choices for the global inertialframe;Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;General Relativity has too many choices for the global inertialframe;Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;
General Relativity has too many choices for the global inertialframe;Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;General Relativity has too many choices for the global inertialframe;
Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;General Relativity has too many choices for the global inertialframe;Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;
Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;General Relativity has too many choices for the global inertialframe;Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.
There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The surplus mathematical structure
Physics is replete with surplus mathematical structure:
Newtonian spacetime has too many rest frames among theinertial frames;General Relativity has too many choices for the global inertialframe;Quantum Field Theory has unitarily non-equivalent butobservationally equivalent representations available;Gauge theories provide too many solutions to the fieldequations which are all acceptable from an empirical point ofview, etc.There are 10500 String Theories (models?) etc.
49
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The GNS and symmetry, invariance and objectivity
Are invariance and symmetry related to objectivity? Redheadand Debs think they are not
Objectivity is conventional.
50
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The GNS and symmetry, invariance and objectivity
Are invariance and symmetry related to objectivity? Redheadand Debs think they are not
Objectivity is conventional.
50
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The GNS and symmetry, invariance and objectivity
Are invariance and symmetry related to objectivity? Redheadand Debs think they are not
Objectivity is conventional.
50
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
No more partial differential equations?
Laws of evolutions are based on partial differential equations:the most powerful tools in the history of science
51
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
No more partial differential equations?
Laws of evolutions are based on partial differential equations:the most powerful tools in the history of science
51
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnature
MRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicproperties
Laws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.
For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”
Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:
If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse data
for a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,
for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”
Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
The Humeans and GNS
GNS mesh well with the Mill-Ramsey-Lewis view of laws ofnatureMRL=laws of nature supervene on the collection of intrinsicpropertiesLaws of nature are the best balance between strength,simplicity, expressiveness, etc.For GNS: new data, new laws. But some laws have a betterstability to new data. Those are “laws of nature”Statistical and stochastic MRL:If a large population of algorithms
for diverse datafor a large range of contraints,for a large number of runs of GNS
converge to the same expression, then dub it “a law of nature”Laws of nature as stable maximal sets (Lange, Woodward, S.Mitchell)
52
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Advice for NS coming from GNS
Take biomimetics seriously.
Take seriously the context of discovery.
Naturalize philosophy of science.
53
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Advice for NS coming from GNS
Take biomimetics seriously.
Take seriously the context of discovery.
Naturalize philosophy of science.
53
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Advice for NS coming from GNS
Take biomimetics seriously.
Take seriously the context of discovery.
Naturalize philosophy of science.
53
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Advice for NS coming from GNS
Take biomimetics seriously.
Take seriously the context of discovery.
Naturalize philosophy of science.
53
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
An argument for distributive discovery
P. Humphreys suggests that the future of science belongs tothe collaboration humans between computers (Humphreys2004, 2009)
We need a division of labor
This will be more evident in discovery than in justificationI do not think NS score better in justification
54
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
An argument for distributive discovery
P. Humphreys suggests that the future of science belongs tothe collaboration humans between computers (Humphreys2004, 2009)
We need a division of labor
This will be more evident in discovery than in justificationI do not think NS score better in justification
54
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
An argument for distributive discovery
P. Humphreys suggests that the future of science belongs tothe collaboration humans between computers (Humphreys2004, 2009)
We need a division of labor
This will be more evident in discovery than in justificationI do not think NS score better in justification
54
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
An argument for distributive discovery
P. Humphreys suggests that the future of science belongs tothe collaboration humans between computers (Humphreys2004, 2009)
We need a division of labor
This will be more evident in discovery than in justification
I do not think NS score better in justification
54
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
An argument for distributive discovery
P. Humphreys suggests that the future of science belongs tothe collaboration humans between computers (Humphreys2004, 2009)
We need a division of labor
This will be more evident in discovery than in justificationI do not think NS score better in justification
54
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Ways to retort: some objections
1 Wait-and-see (again!) GNS are not ripe to harvest.
2 GNS bring irrationality into science. Do we need it?
3 If the world is stochastic/chancy then we do not need torepresent it by stochastic tools.
4 You do not honor the discovery/justification distinction.
5 If nature is deterministic why should we go stochastic? (seebelow)
6 The “three armies” argument (see below)
7 This is anthropomorphism in sheep clothes!(see below)
55
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Back to Objection 5: Maybe Nature abhors chance!
Maybe at the bottom She abhors chance.
If so, all our models are pseudo-stochastic
If so, we can represent parts the world with stochastic models
M. Strevens: “Not only is Nature red in tooth and claw; she isirrepresibly stochastic in her violence” (Strevens, 2009, 459)
models in biology have a probabilistic element.
56
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Back to Objection 5: Maybe Nature abhors chance!
Maybe at the bottom She abhors chance.
If so, all our models are pseudo-stochastic
If so, we can represent parts the world with stochastic models
M. Strevens: “Not only is Nature red in tooth and claw; she isirrepresibly stochastic in her violence” (Strevens, 2009, 459)
models in biology have a probabilistic element.
56
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Back to Objection 5: Maybe Nature abhors chance!
Maybe at the bottom She abhors chance.
If so, all our models are pseudo-stochastic
If so, we can represent parts the world with stochastic models
M. Strevens: “Not only is Nature red in tooth and claw; she isirrepresibly stochastic in her violence” (Strevens, 2009, 459)
models in biology have a probabilistic element.
56
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Back to Objection 5: Maybe Nature abhors chance!
Maybe at the bottom She abhors chance.
If so, all our models are pseudo-stochastic
If so, we can represent parts the world with stochastic models
M. Strevens: “Not only is Nature red in tooth and claw; she isirrepresibly stochastic in her violence” (Strevens, 2009, 459)
models in biology have a probabilistic element.
56
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Back to Objection 5: Maybe Nature abhors chance!
Maybe at the bottom She abhors chance.
If so, all our models are pseudo-stochastic
If so, we can represent parts the world with stochastic models
M. Strevens: “Not only is Nature red in tooth and claw; she isirrepresibly stochastic in her violence” (Strevens, 2009, 459)
models in biology have a probabilistic element.
56
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
Back to Objection 5: Maybe Nature abhors chance!
Maybe at the bottom She abhors chance.
If so, all our models are pseudo-stochastic
If so, we can represent parts the world with stochastic models
M. Strevens: “Not only is Nature red in tooth and claw; she isirrepresibly stochastic in her violence” (Strevens, 2009, 459)
models in biology have a probabilistic element.
56
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 6 The three armies argument against GNS
Whatever you told us, Nothing that NS have achieved couldnot have been done by:
an army of well-trained scientists working with slide rules
an army of well-trained evolutionary biologists
an army of well-trained poker dealers and gamblers
Computers, NS are still dumb tools
57
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 6 The three armies argument against GNS
Whatever you told us, Nothing that NS have achieved couldnot have been done by:
an army of well-trained scientists working with slide rules
an army of well-trained evolutionary biologists
an army of well-trained poker dealers and gamblers
Computers, NS are still dumb tools
57
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 6 The three armies argument against GNS
Whatever you told us, Nothing that NS have achieved couldnot have been done by:
an army of well-trained scientists working with slide rules
an army of well-trained evolutionary biologists
an army of well-trained poker dealers and gamblers
Computers, NS are still dumb tools
57
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 6 The three armies argument against GNS
Whatever you told us, Nothing that NS have achieved couldnot have been done by:
an army of well-trained scientists working with slide rules
an army of well-trained evolutionary biologists
an army of well-trained poker dealers and gamblers
Computers, NS are still dumb tools
57
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 6 The three armies argument against GNS
Whatever you told us, Nothing that NS have achieved couldnot have been done by:
an army of well-trained scientists working with slide rules
an army of well-trained evolutionary biologists
an army of well-trained poker dealers and gamblers
Computers, NS are still dumb tools
57
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 6 The three armies argument against GNS
Whatever you told us, Nothing that NS have achieved couldnot have been done by:
an army of well-trained scientists working with slide rules
an army of well-trained evolutionary biologists
an army of well-trained poker dealers and gamblers
Computers, NS are still dumb tools
57
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 7: Wait: this is anthropomorhism!
is biomimetics nothing more than anthropomorphizing?
It seems the nature abhors many things
58
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 7: Wait: this is anthropomorhism!
is biomimetics nothing more than anthropomorphizing?
It seems the nature abhors many things
58
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
What philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
O 7: Wait: this is anthropomorhism!
is biomimetics nothing more than anthropomorphizing?
It seems the nature abhors many things
58
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
Outline1 Philosophy of Numerical Simulations?
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
2 Genetic Algorithms in Numerical Simulations (GNS)Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
3 Philosophy of GNSWhat philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
4 FinisRisky conclusionsWeaker conclusions
5 References
59
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
What emerges from my analysis?
Surprisingly, a less central role for mathematics in scientificdiscovery
More or less a bottom up image of science
A new type of possibilia, the numerical possibility
60
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
What emerges from my analysis?
Surprisingly, a less central role for mathematics in scientificdiscovery
More or less a bottom up image of science
A new type of possibilia, the numerical possibility
60
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
What emerges from my analysis?
Surprisingly, a less central role for mathematics in scientificdiscovery
More or less a bottom up image of science
A new type of possibilia, the numerical possibility
60
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
What emerges from my analysis?
Surprisingly, a less central role for mathematics in scientificdiscovery
More or less a bottom up image of science
A new type of possibilia, the numerical possibility
60
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
Moving from Turing to stochastic algorithms makes NS moreattractive
Relevance and meaning can be generated by machines
For better, for worse, GNS are not mere glorified slide rules
61
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
Moving from Turing to stochastic algorithms makes NS moreattractive
Relevance and meaning can be generated by machines
For better, for worse, GNS are not mere glorified slide rules
61
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
Moving from Turing to stochastic algorithms makes NS moreattractive
Relevance and meaning can be generated by machines
For better, for worse, GNS are not mere glorified slide rules
61
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Risky conclusionsWeaker conclusions
Moving from Turing to stochastic algorithms makes NS moreattractive
Relevance and meaning can be generated by machines
For better, for worse, GNS are not mere glorified slide rules
61
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
Outline1 Philosophy of Numerical Simulations?
What are Numerical Simulations (NS)?Philosophical questionsThree stancesThe “glorified slide rule argument”My position
2 Genetic Algorithms in Numerical Simulations (GNS)Beyond TuringSurvival and chance in computer scienceInductive programming (skip)Genetic numerical algorithms (GNS)
3 Philosophy of GNSWhat philosophy for GNS?Arguments for GNSMetaphysics of GNSGNS and mathematicsGNS and invarianceGNS and laws of natureObjections
4 FinisRisky conclusionsWeaker conclusions
5 References
62
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References I
Affenzeller, M., 2009. Genetic Algorithms and GeneticProgramming: Modern Concepts and Practical Applications.No. v. 6 in Numerical insights. CRC Press, Boca Raton, Fla.
Barberousse, A., Franceschelli, S., Imbert, C., 2007. Cellularautomata, modeling, and computation.http://philsci-archive.pitt.edu/archive/00003579/.URLhttp://philsci-archive.pitt.edu/archive/00003579/
Callender, C., Cohen, J., 2010. Special sciences, conspiracyand the better best system account of lawhood.
63
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References II
Chaitin, G., Feb. 1995. Randomness in arithmetic and thedecline and fall of reductionism in pure mathematics. Chaos,Solitons & Fractals 5 (2), 143–159.
Cohen, J., Callender, C., 2009. A better best system accountof lawhood. Philosophical Studies 145 (1), 1–34.
Crick, F., 1981. Life Itself: Its Origin and Nature, 1ST Edition.Simon and Schuster.
Davies, P. C. W., Brown, J. R., 1988. Superstrings: a theoryof everything? Cambridge University Press, Cambridge U.K.;New York, book, Edited.
64
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References III
Debs, T., 2007. Objectivity, invariance, and convention :symmetry in physical science. Harvard University Press,Cambridge Mass.
Floridi, L., 2008. Philosophy of computing and information : 5questions. Automatic Press, [S.I.].
Frigg, R., Reiss, J., 2009. The philosophy of simulation: Hotnew issues or same old stew? Journal for Epistemology169 (3), 593–613.
65
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References IV
Galison, P. L., 1996. Computer simulations and the tradingzone. In: Galison, P., Stump, D. J. (Eds.), The Disunity ofscience : boundaries, contexts, and power. Stanford UniversityPress, Stanford Calif.
Hacking, I., 1983. Representing and intervening: introductorytopics in the philosophy of natural science. CambridgeUniversity Press, Cambridge, Cambridgeshire ; New York,book, Whole.
Hartmann, S., 2008. Modeling in Philosophy of Science. OntosVerlag, Heusenstamm bei Frankfurt, book, Whole.
66
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References V
Humphreys, P., 2004. Extending ourselves: Computationalscience, empiricism, and scientific method. 2004.
Humphreys, P., 2009. The philosophical novelty of computersimulation methods. Synthese 169 (3), 615–626.
Keller, E., 2003. Models, simulation, and ’Computerexperiments’. In: Radder, H. (Ed.), The Philosophy ofScientific Experimentation. University of Pittsburgh Press, pp.198–215.
67
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References VI
Koza, J. R., Keane, M., Streeter, M., Mydlowec, W., Yu, J.,Lanza, G. (Eds.), 2003. Genetic Programming IV: RoutineHuman-Competitive Machine Intelligence. Kluwer AcademicPublishers, Norwell, Mass.
Maddy, P., 2007. Second philosophy : a naturalistic method.Oxford University Press, Oxford ;;New York.
Morrison, M., 2009. Models, measurement and computersimulation: the changing face of experimentation.Philosophical Studies 143 (1), 33–57, journal Article.
68
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References VII
Parker, W., 2009. Does matter really matter? computersimulations, experiments, and materiality. Synthese 169 (3),483–496.
Simon, H., 1969. The sciences of the artificial,. M.I.T. Press,Cambridge Mass.
Simon, H. A., 1992. Scientific discovery as problem solving.International Studies in the Philosophy of Science 6 (1), 3.
Simon, H. A., Langley, P. W., Bradshaw, G. L., 1981.Scientific discovery as problem solving. Synthese 47 (1), 1–27.
Simonton, D. K., 2004. Creativity in Science: Chance, Logic,Genius, and Zeitgeist. Cambridge University Press, Cambridge.
69
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References VIII
Steiner, M. A., 1998. The applicability of mathematics as aphilosophical problem. Harvard University Press, Cambridge,Mass., book, Whole.
Thagard, P., 1988. Computational philosophy of science. MITPress, Cambridge, Mass., book, Whole.
Tomassini, M., 1995. A survey of genetic algorithms. AnnualReviews of Computational Physics 3, 87–118.
Turing, A., 1950. Computing machine and intelligence. Mind:A Quarterly Review of Philosophy 59 (236), 433–460.
Turing, A., 1996. Intelligent machinery, a heretical theory.Philosophia Mathematica 4 (3), 256–260.
70
Philosophy of Numerical Simulations?Genetic Algorithms in Numerical Simulations (GNS)
Philosophy of GNSFinis
References
References IX
Winsberg, E., 1999. Sanctioning models: The epistemology ofsimulation. Science in context. 12 (2), 275.
Winsberg, E., 2001. Simulations, models, and theories:Complex physical systems and their representations.Philosophy of Science 68 (3), S442–S454.
Wolfram, S., 2002. A new kind of science. Wolfram Media,Champaign IL.
71