_______________________________
|Artificial Life
|
| Dave Ackley
| Computer Science
| University of New Mexico
| ackley cs. unm. ed u
|
| Visiting UCSD Cognitive Science
| 2001-2002 ackley hci. ucsd. ed u
|
| http://ackleyshack.com/
|
__________________________________
|
| Something happened July 19, 2001
|
|(animation)
__________________
|
| Code Red (v2 w/randomized seed)
|
|- Infected >359,000 hosts in 14 hours
|
|- Peak infection rate: 2,000 hosts/min
|
|- Global network routing instabilities
|
__________________
|
| Is Code Red alive?
|
__________________
|Outline
|
|- What is life
|
|- Artificial life
| - science & engineering
|
|- What we can learn
|
__________________
|What is life?
|
|- Is that the right question?
|
|- What are we looking for?
| A definition?
|
___________________________
|What is (the definition of) life?
|
| Webster's Unabridged (1913):
|
| 1. The state of being which begins with
| generation, birth, or germination, and ends
| with death; also, the time during which this
| state continues; that state of an animal or
| plant in which all or any of its organs are
| capable of performing all or any of their
| functions; -- used of all animal and
| vegetable organisms. ...
|
__________________
|What is life?
|
|- Can we find necessary &
| sufficient conditions?
|
|- What characteristics do all
| living things share?
|
__________________
|Characteristics of life
|
|Life is that which ...?
|- Grows
|- Consumes to survive
|- Reproduces itself
|- Heals itself
|
__________________
|What is life?
|
|- No firm "scientific definition"
|
|- Is "being life" an all-or-none
| proposition?
|
|- Big umbrella:
| "Dynamically maintained patterns"
|
___________________________
|What is (the definition of) life?
|Life is:
|
|"a self-sustaining chemical system capable
| of undergoing Darwinian evolution."
| (Gerald Joyce)
|
| "a self-organized non-equilibrium system
| such that its processes are governed by
| a program that is stored symbolically and
| it can reproduce itself, including the
| program." (Lee Smolin)
|
| "nature's way of preserving meat."
|
__________________
|Artificial life ("alife", "AL")
|
|- Biology: Compared to what?
|
|- `Life as it is' vs `Life as it could be'
|
|- The weak claim for AL: Computational
| systems can be models of living systems
|
|- The strong claim for AL: Computational
| systems can be living systems
|
__________________
|Artificial life
|
| - Strong claim advocates:
| = Talk about "systems"
| = Focus on building
| = Seek things that work
|
| - Weak claim advocates:
| = Talk about "models"
| = Focus on understanding
| = Seek things that predict
|
| - The "World In A Box" methodology
|
__________________
|Artificial life
|
|- The empirical study of
| evolutionary systems via
| computational modelling
|
|- Relationships:
| Biology, Cognitive science,
| Artificial intelligence, Economics
|
|(Also: Cultural evo;evo language;
| ethology; epistemology...)
|
__________________
|Artificial life: How to play
|
|- Focus on discovering consequences
| of evolutionary processes
|
|- Evolution needs:
| (1) Heritable traits
| (2) Source of variation
| (3) Source of differential survival
| "fitness"
__________________
|Artificial life: How to play
|
|Typical alife responses -
|(1) Heritable traits:
| - specified directly by the creator
|(2) Source of variation:
| - specified directly by the creator
|(3) Source of differential survival
| - specified indirectly by the creator
| via the "world" the creatures inhabit
|
__________________
|Artificial life: How to play
|
|- Points to note:
| = Full of outrageous simplifications
| (and yet)
| = Myriads of "parameters", with
| few motivations
|
____________________________________
|Artificial Life: Examples
|
|- Learning from evolution
| (e.g., Ackley & Littman, 1992)
|
|- Evolution of communication
| (e.g., Ackley & Littman, 1994)
|
__________________
|Learning from evolution
|
|- Learning and evolution:
| Mechanisms of adaptation
|
|- Learning algorithms depend on
| supervision/reinforcement/feedback
|
__________________
|Learning from evolution
|
|- Less feedback, generally harder to learn
|
|- Least feedback: Death
|
|- How to learn from death?
|
__________________
|Learning from evolution: A model
|
|- Need both behavior ("what to do")
| and evaluation ("how am I doing")
|
|- "Genes" specify evaluation (fixed
| for life) and initial behavior
| (modifiable by learning)
|
|- If evaluation score increases,
| reward previous action, and vv
|
__________________
|Learning from evolution: A model
|
|- World is 100x100 grid
|- Learning and evolving agents
|- Finite state handcoded predators
|- Growing grass supplies food
|- Growing trees supplies shelter
|
__________________
|Learning from evolution: A model
|
|Most typical outcome:
|- Agent populations quickly extinct
|
|Result:
|- L+E populations survive more often
| compared to E-only, L-only, Neither
|
__________________
|Learning from evolution: A model
|
|Other observations:
|- Predator-prey oscillations
|- Long-term adaptation
|
__________________
|Learning from evolution
|
|Other observations:
|- Baldwin effect
|- Shielding
|
|- Multilevel models
|- Scaling in time
|
____________________________________
|Artificial Life: Examples
|
|- Learning from evolution
| (e.g., Ackley & Littman, 1992)
|
|- Evolution of communication
| (e.g., Ackley & Littman, 1994)
|
__________________
|Evolution of communication
|
|- Also, e.g., Batali (1994, 1998), Hutchins &
| Hazelhurst (1995), Werner & Todd (1997), ...
|
|- Beyond the isolated individual
|
|- Evolution of cooperation
|
|- Would an agent ever `speak truth' if
| it only benefited the listener?
|
|- Assigning meaning to arbitrary signals
|
__________________
|Evolution of communication
|
|- Spatial levels: Individual, local, global
|
|- Reproduction & communication is local
|
|- `Migration' mixes nearby local populations
|
__________________
|Evolution of communication
|
|- Communication based on arbitrary signals
| can evolve and stabilize even when it
| provides no benefit for the individual speaker.
|
|- Key is partial alignment of communication
| range and reproductive range
|
|- A variety of `parasites' may arise
|
__________________
|Outline
|
|- What is life
|
|- Artificial life
| - science & engineering
|
|- What we can learn
|
__________________
|What we can learn
|
|- May need less than we think
|
|- Unexpected connections between
| phenomena, between local and
| global views, and short/small
| and long/large
|
|- "Physical intuitions"
| for evolutionary systems
|