Artificial Life intro
----------------------- by Teresa Dillon (03/12/03)------------
Artificial Life: terminology
· Artificial life (aLife, A-life) is a diverse and rich field of research.
It attempts to examine how life is more than the sum of its parts and draws
on evolution, chaos, complexity and systems theory
· aLife is not artificial intelligence (AI) they are two separate fields
with different preoccupations
· Opinions on the application of aLife have developed within many subject
areas, depending on the interests of those involved and their backgrounds
· As a science aLife focuses on the principles of self-organisation and
the emergent properties and processes through which organisms develop
· Common across all research within aLife, is the notion of 'teasing
out the fundamental principles of life by building detailed working models'
(Brooks, 1990a)
· According to Brooks one of the most ambitious goals of aLife research
is the construction of living systems out of non-living parts
· Grand (2000) notes aLife (as a technology) is about putting the 'soul
back into lifeless machines', the 're-vivificationn (i.e. rebirth) of technology'
through the 'art of creating and embedding living things
into machines, either in software or hardware'. This 're-birth' has been happening
through various computational conceptualisations of life from experimentations
with robots to computer simulations
· Ironically the term aLife itself, although still widely used has become
somewhat redundant as the field develops and becomes more specialised - researchers
in the field tend to align themselves with other
disciplines, such as cybernetics and the specialisms that are developing within
this and related fields
Artificial Intelligence: terminology
· Traditional AI has its roots within psychology, it assumes that the
mind is a virtual machine created by the brain and therefore can be studied
and presumably replicated at that level of description
· AI is not interested in replication, evolution or any of the other
aspects of life
· Traditional AI (GOFAI - Good Old-Fashioned AI) takes a top-down approach
to human thinking and behaviour. It attempts to make computers mimic intelligent
behaviour, without considering the structures that give
rise to such behaviour
· During the 80's and 90's, a New AI (AI Nouveau) started to emerge,
which is much more bottom-up and concerned with embodiment, situatedness, and
biological plausibility
ALife vs. AI:
· In comparison to traditional AI, New AI shares its conceptual framework
with aLife
· However although there is a historical connection between AI and aLife
(see following section), aLife has stronger links with other disciplines such
as cybernetics, self-organisation theory, biology, arts and chaos theory etc
· However the boundaries between New AI and aLife are not clear-cut and
for many New AI researchers, evolution and simple adaptation are regarded as
intelligence
· Additional New AI still fails to account for the complexities of
intelligence such as memory, learning, prediction and theory of mind
· It is important to remember (in mine view, others could argue against
this) both aLife and AI are just ways of interpreting the world and both fail
to account for the more complex and experiential nature of human life
A brief history of aLife
· Historically Alan Turing (1912-1954) and Von Neumann (1903-1957) influenced
both aLife and AI
The low down on Mr. Turning
· During World War 2, Turning worked in the UK on breaking the U-boat
Enigma
· He developed what is known as the Turning machine (1937), which theoretically
defined a computing system as an infinite tape, which can perform various computations
· Turing also explored what he called Unorganised Machines, which are
similar to what we now call neural networks
· Turning also conducted seminal work on embryology and has been called
by some the 'grand - daddy' of aLife
· In addition he developed the concept known as the Turning Waves, a
mathematical model of how a smooth, symmetrical egg cell might spontaneously
develop the asymmetry needed to enable it to develop into a
decidedly asymmetrical embryo
· Despite this the notion of Turning Waves was neglected for years, although
it now assists researchers to explain various kinds of morphogenic events such
as how animals get their skin patterns
The low down on Herr Von Neumann
· Von Neumann focused on computational neurosciences, in particular cellular
automata and self- reproducing systems - both have had a big impact on aLife
(but later on rather than when he was a live himself)
· He also defined the universal replicator (1946) as a computational
system capable of reproducing any system - he realised that such a system must
deal with both the instruction and the data and that errors in
copying self-description could lead to evolution, which could be examined through
computation.
· However as Herr Von Neumann died too early, he didn't publish much
of his work but his friend and colleague Arthur W. Burks did (after his death,
Burks, 1968). Later on Chris Langton (see next section) picked up the Von Neumann
cellular automata baton, claiming that cellular automata was capable to produce
self-reproducing patterns (loops to be precise) (e.g. Langton, 1984)...and then
· As noted despite the seminal work conducted by both Turing and Von
Neumann, within many disciplines AI rather than aLife became the main focus
of interest
· But in 1987 Chris Langton hosted the first aLife conference at Los
Alamos National Laboratories, Santa Fe Institute, New Mexico - it was here that
the term aLife was officially coined
· The conference brought together a small but growing body of international
researchers that believed, that life and its emergent properties could be modelled
or simulated by computers
· During the conference Langton gave a keynote paper, which is now generally
regarded as the doctrine on aLife
· Langton (1988) defined aLife as 'a property of the organisation of
matter rather than a property of matter itself'
· As noted the term aLife itself, although still widely used is becoming
somewhat redundant as the field develops and becomes more specialised
Notes:
Cellular automata (automata in 'computer speak', refers to all machines whose
output behaviour is not a direct consequence of the current input, but of some
past history of its inputs - thus the machine has some
kind of 'inner state') are dynamical systems whose behaviour is completely specified
in terms of a local relation. A cellular automaton can be thought of as a stylised
universe. e.g. A uniform grid represents space, with each cell containing a
few bits of data. The system's laws are local and uniform.
Self-reproducing systems = Self reproduction is an essential "property"
of life. Therefore in 'computer speak' if we want to understand what is life,
self-reproduction is important to understand i.e. how do systems continuously
produce themselves and their constitutive parts?
References:
A. M. Turing's ACE Report of 1946 and Other Papers (1986). In B. E. Carpenter
and R. W. Doran (Eds). MIT Press.
Brooks, R. (1989). A robot that walks: Emergent behaviours from a carefully
evolved network. In Brooks, R. Neural Computations, 1:2 Summer, 253-263
Brook, R. (1990a). The Behaviour Language: Users Guide. MIT, A.I, Lab Memo,
1227, April
Grand, S., Cliff, D., & Malhorta, A. (1997b). Creatures: artificial life,
autonomous software agents for home environment. Paper presentation at 1st International
Conference on Autonomous Agents, ACM Press, New
York, NY.
Grand, S. (2000). What is Cyberlife?
http://www.simons-rock.edu/rlovison/grand.html
Langton, C. (1988). Toward Artificial Life, Whole Earth Review, 58(74)
Von Neumann, J. (1946). The Principles of Large-Scale Computing Machines, reprinted
in Ann. History Computer, Vol. 3, No. 3, pp. 263-273