SIMULATIONS, REALIZATIONS,AND THEORIES OF LIFE

H. H. PATTEE (1989)

By Hyojung SeoDept. of Psychology

CONTENTS

• How Universal is a Computer?• Realizations, Simulations, and Theories• The Limitations of Theory-free Simulations• Simulations do not become Realizations• The Symbol-Matter Problem• Weak or Strong Artificial Evolution?• Three Levels of Emergent Behavior• Can we Artificially Evolve New Measurements?• Conclusions

How would we distinguish computer simulationsfrom realizations of life?

How does this relate to theories of life?: how can the living be distinguished from

the non-living?

• Hope of Strong AL: Langton (1987)‘We would like to build models that are so lifelike that they would cease to be models of life and become examples of life themselves’

• Needed Discussion

1. Simulations and Realizations belong different categories of modeling. Simulations: metaphorical models that symbolically

‘stand for’ something else.Realizations: literal, material models that

implement functions

2. The criteria for good simulations and realizations of a system depend on our theory of the system. 3. Our theory of living systems must include evolvability.

evolution: symbolic genotypes, material phenotypes, selective environments

Main Ideas

How Universal is a Computer?

• Artificial Life studies have closer roots in Artificial Intelligence and Computational modelling than in Biology itself.

• Platonic ideal & AI- rule-based formalists- law-based ecological realists- neural network, connectionist,

& parallel distributed processor

How Universal is a Computer? (Cont’d)

• The criticism of AI , which will be aimed at AL as well, is that it has for the most part neglected the fundamental biology of the nervous system.

• Who decides ‘What is fundamental about biology’? : A theory of life must be decisive.

• neuroscientists: a model of the brain that is empirically testablein biological systems.

Realizations, Simulations, and Theories

1) Computer-dependent realizations ofliving systems

2) Computer simulations of living-systems behavior

3) Theories of life that derive from simulations

4) Theories of life that are testable only by computer simulations

Need to distinguish between,

Realizations, Simulations, and Theories (Cont’d)

Realization

• Judged by how well it can function as an implementation of a design specification

• The problem of AL is, What is the operation or function of living?

• The classical theory of life requires, symbolic genotype,

material phenotype, environment

• Two possibilities for strong AL: (1) include robotics to realize the phenotype-environment interactions

(2) treat the symbolic domain of the computer as an artificial environment in which symbolic phenotype properties of

artificial life are realized

Realizations, Simulations, and Theories (Cont’d)

Realization (cont’d)

Simulation

Realizations, Simulations, and Theories (Cont’d)

• Judged by how well they generate similar morphologies or parallel behaviors.

of some specified aspects of the system

• The simulation, no matter how accurate, is not the same as the thing simulated. : has extra features of simulation medium

that are not found in the system

• The simulation depends on theory. : represents the essential functions of

a realization of the system

Theories

Realizations, Simulations, and Theories (Cont’d)

• Judged by abstract tests of universality,

conceptual coherence, simplicity and elegance

• Judged by concrete tests of how well they can predict specific values for the observables

of the system being modelled.

• Physics: formal theoretical structures- mathematical models of generality and formal simplicity that do not have perceptual or behavioral similarities with the world they model

• The study of dynamical systems by computeroften blurs the distinction between theoriesand simulations.

The Limitations of Theory-Free Simulations

• Computational Universality

Newell & Simon, Physical Symbol System Hypothesis : Computation can realize intelligent thoug

ht. ‘This form of symbolic behavior is all the

re is; it includes human symbolic behavio

r.’

In AI

• The power of computer simulation obscures the basic requirements of a scientific theory.

The Limitations of Theory-Free Simulations (Cont’d)

• Problem :It has not been verified by the delicate criteria

for theory; A working simulation is not really evidence for the theory.

both simulations and realizations must be evaluated in terms of

a theory of the brain, and the empiricalevidence for that theory.

Simulations

Theory Realization

Theory

Simulations Realization

In AL

The Limitations of Theory-Free Simulations (Cont’d)

• Working simulations are not by themselves evidence for or against theories of life.

• There are many alternative ways to successfully simulate any behavior.

• All the models must be evaluated by theory.

• We must allow only universal physical laws andthe theory of natural selection to restrict the evolution of artificial life.

Simulations do not become Realizations

• Categorical difference:Realization: literal, substantial replacementSimulation: metaphorical representation of

specific structure or behavior - symbolic forms, not material substance

• Measurement in Simulation:A mapping from observable aspects of the system to corresponding symbolic elementsof the simulation

problem of formalization of life

The Symbol-Matter Problem

• The molecular facts do not constitue a theory of how symbolic forms and material structures

must interact to evolve.

• Need for the theory of evolution Von Neumann(1966) : evolution of complexity - the essential characteristic of life - symbolic instructions and universal construction

necessary for heritable & open-ended evolution

: the problem of the relation of symbol and matter

The Symbol-Matter Problem (Cont’d)

• The Physical Symbol System Hypothesisis half-truth:

- We can construct symbol system from matter. - Matter can not be constructed from symbols. - Evolutionary Hypothesis:

under the symbolic control of genotypes, material phenotypes in an environmentcan realize endless varieties of structures and behaviors.

(1) we can simulate everything by universal symbol systems

(2) we can realize universal symbol systemswith material constructions

(3) we can realize endless types of structuresand behaviors by symbolic constraintson matter

The Symbol-Matter Problem (Cont’d)

we cannot realize material systems with symbols alone

Three symbol-matter possibilities:

Fundamental impossibilities:

Weak or Strong Artificial Evolution

• AL realization should include the genotype, phenotype, environment distinctions as well as the corresponding mutability, heritability and natural selection of neo-Darwinian theory.

• A realization of life should have the emergent or novel evolutionary behavior that goes beyond adaptation to an environment.

• Whether formal environment like a computer can realize evolutionary novelty or merely simulate?

• Computation / physical models

- Computational universality and physical universality refer to different domains. physical theories: observable world.

universality -- restrictions on theories computational universality: symbolic domain

universality -- no restriction on symbolic rewriting rule

Weak or Strong Artificial Evolution (Cont’d)

Three Levels of Emergent Behaviour

• Syntactic emergence: symmetry-breaking, chaotic dynamics

strong emergence / weak emergence

• Semantic emergencenon-dynamical symbol systems

• Measurement itself

Can we artificially evolve New Measurement?

• New measurements as one of the more fundamental test cases for emergent behavior

in artificial life models.

• Generalized measurement is a record stored in the organism of some type of classification of the environment and this classification must be realized by a measuring device.

Can we artificially evolve New Measurement?

• The primary processes of evolution is the construction of new measuring devices,and this type of substantive emergence isout of the domain of symbolic emergence,so we cannot expect to realize this naturaltype of evolutionary emergence with computers alone.

• Autonomous classification: reclassification of the results of measurement,no new measurement

:realization of semantic emergence

Conclusions

1. AL must evaluate its models by the strength of its theories of living systems, and not by technological mimicry alone.

2. There is nothing wrong with a good illusion as long as one does not claim it is reality.

3. AL should pay attention to the enormous knowledge base of biology.

4. The process of measurement is proposed as a test case for the distinction between simulation and realization