mercredi 12 octobre 2011

Ethological Turn.

Janne Karhu, Video of Boids in Action, 2009.

Suddenly the cyborg as imagined since the 1980s in theory and fiction seemed quite old-fashioned. This shift was not altogether dismissing the human being and its perceptive and cognitive capabilities: the two-handed and -legged animal was seen to demonstrate distinct powers in visual (recognizing edges, seeing contrasts, differentiating between dimensional entities) and tactile (the hand) faculties. Yet a much less brainy entity, the insect, was a powerful new kind of model for designing artificial agents that expressed complex behavior, not through pre-programming and centralization but through autonomy, emergence, and distributed functionning. Since the 1980s, such terms as swarms, distributed intelligence, and insect models of organization have infiltrated both the design of digital technologies and cultural theoretical analysis of such media systems. Yet, as researchers commented, "The most talented roboticist in the world is not going to come close to what a cockroach can do".

One of the most discussed contexts for such a cultural and scientific reorientation in terms of design practices and plans was artificial intelligence (AI) research. New ideas in cognitive science seemed to offer the most convincing explanations of the potential for tapping into the simple architectures already developed by nature. "Intelligence is overrated", such research paradigms seemed implicitly to suggest. The approach, which focused on the redundancy of numerous "dumb" machines, emphasized that

1.there is no need for planning;

2.no need for central representation;

3.our traditional ways of modeling the world for the actors are impractical and unnecessary;

4.we should pay more close attention to biology and evolution;

5.one should focus on building real, concrete solutions, not merely theoretical models.

In robotics, MIT professor Rodney Brooks noted in the late 1980s that artificial agents do not have to resemble or act like humans; there are much more efficient ways of doing complex tasks than by modeling intelligent machines. Brooks designed insectlike robots, and in his 1989 paper "Fast, Cheap, and Out of Control", coauthored by Anita Flynn, he introduced the idea of using insectlike mobots as space exploration agents instead of large "intelligent" ones. Douglas Hofstadter had already used the notion of the ant colony to pave the way for a rethinking of cognition as distributed "mass communication" between miniagents, but Brooks deployed a similar insect metaphor: no central command but massive parallelism and cooperation.

Such research in "new AI" had many parallels in the emerging artificial life sciences, which, however, dealt mostly with software. The approaches were kicked off by researchers such as Christopher Langton. In that context, and in the midst of the emerging digital software culture of the 1980s, the field of programming also gained much from the scientific theories of artifical life. In software and network processes, simple but interconnected agents had been planned since the 1960s. Nowadays everybody knows viruses and worms by name, but the fact that we are thinking them in terms of parallel processing and artificial life is often less emphasized. Yet such program types, which span computer boundaries as "parasite computing", are exemplary of software that acts in a manner reminiscent of insect colonies: individually dumb but highly efficient when coupled with their environment. The ideas of distributing artificial actors into insectlike colonies of part functions and parallel processing represented a move toward situatedness but also embodiment: robots are in the world, and their actions are enabled and controlled by the very present environment. This could be seen as signaling a kind of ethological turn in creating artificial agents, because such ideas were reminiscent of those of animal ethologists such as Jakob von Uexküll's in work from the 1920s: artifical actors are embedded in a perceptual world, which implies that what we perceive is what we are, and animals and artificial agents are defined by the capabilities of perception, sensation, and orientation in their environment.

Jussi Parikka, Insect Media, An Archeology of Animals and Technology, University of Minnesota Press, 2010, p.x-xii.

See also here, here and there.

Aucun commentaire:

Enregistrer un commentaire