Searching for Artificial Intelligence in Artificial Life

In 2000, several prominent artificial life researchers published their co-authored list of 14 “open problems in artificial life”. Of special interest is their open problem number 11: “Demonstrate the emergence of intelligence and mind in an artificial living system. ” (p. 365). Not only do the authors pose the problem, but they give what strikes me as excellent advice towards its solution:

“To make progress, one must have a method to detect intelligence and mind when they are present in a system. Consciousness is the most difficult aspect of mind to detect, and initial progress is certain to be somewhere else. A more tractable aspect of mind to detect is meaning, that is, internal states that have semantic or representational significance for the entity and that influence the entity’s behavior by means of their semantic content.” (pp 372-373).

Progress along these recommended lines toward the solution of problem 11 will also involve work of relevance to what they identify as open problem number 10: “Develop a theory of information processing, information flow, and information generation for evolving systems.” Among their remarks on information, one in particular strikes me as especially significant:

“Firstly, there appear to be two complementary kinds of information transmission in living systems. One is the conservative hereditary transmission of information through evolutionary time. The other is transmission of information specified in a system’s physical environment to components of the system, possibly mediated by the components themselves, with the concomitant possibility of a combination of information processing and transmission. The latter is clearly also linked with the generation of information (to be discussed last). Clarifying the range of possibilities for information transmission, and determining which of those possibilities the biosphere exploits, is a fundamental enquiry of artificial life. ” ( p. 372)

As I read the quoted passage, the first kind of information transmission is that which passes from parent to offspring in virtue of reproduction. This is information transmission that traverses generations. The second kind of information transmission is from the environment to the organism. This is something that can happen over multiple generations as populations adapt and evolve. But the transmission of information from environment to organism can also take place within the lifetime of a single organism and this is especially evident in creatures capable of sensory perception and memory. Both perception and memory are amenable to information-theoretic analyses: perception involves the transmission of a signal across space and memory involves the transmission of a signal across time.

The search for mind will be guided by the search for entities that have states with “semantic or representational significance”. The earliest instances of such states will be ones that constitute the pick-up by organisms of information about their environments via sensory components. Slightly more sophisticated instances will involve the retention and processing of that information over time via mechanisms of memory and computation. These forms of information transmission and processing—the ones that constitute the earliest instances of cognition—will emerge in the course of the evolution of organisms that are not themselves in possession of anything cognitive, but may nonetheless be understood in informational terms as follows. The pre-cognitive forbears of cognizers, the non-cognitive mere organism from which cognitive organisms evolve, can be characterized in terms of the transmission of information from parents to offspring via inheritance and the acquisition of novel information at the species level. Non-cognitive or “mere” organisms are not capable of the acquisition of information except by inheritance: novel information is acquired only at the species level over evolutionary time. In contrast, cognitive organisms are the ones capable of the acquisition of novel information in their own lifetime.

These remarks help to suggest a method for addressing open problem number 11: develop a method for evolving artificial organisms in ways such that we (1) are able to detect which of the various kinds of information transmission are present in the system and (2) manipulate factors such as environments and fitness functions to encourage the evolution of the modes of information transmission distinctive of cognitive activity.



Fig 1.:Single artificial organism. Seeks food, companionship. IQ = 2.

Reference:
Bedau, M., McCaskill, J. S., Packard, N., Rasmussen, S., Adami, C., Green, D. G., Ikegami, T., Kaneko, K., and Ray, T. (2000). Open problems in artificial life. Artificial Life 6, 363-376. [ Link]

See also:
Varieties of Representation in Evolved and Embodied Neural Networks. Biology and Philosophy. 18 (1): 95-130. 2003.

Evolving Artificial Minds and Brains. (with Mike Collins and Alex Vereschagin). Categorisation, Mental States, and Development. Andrea Schalley and Drew Khlentzos (eds.) Amsterdam: John Benjamins Publishers. In press.

Synthetic Neuroethology. Metaphilosophy. 33 (1-2): 11-29. Reprinted in CyberPhilosophy: The Intersection of Philosophy and Computing, James H. Moor and Terrell Ward Bynum, (eds.), Oxford: Blackwell, 2002.

Comments are closed.