Previous Next Up Comments
SEHR, volume 4, issue 2: Constructions of the Mind
Updated 4 June 1995

conference review

artificial intelligence, artificial life, and alternatives to computationalism and objectivism


Stefan Helmreich

In late 1993, a primarily European group of computer scientists, biologists, roboticists, and philosophers congregated in San Sebastian, Spain to discuss the lessons that Artificial Intelligence (AI) might learn from the nascent discipline of Artificial Life.[1] Against the claims of traditional AI that cognition is an abstract phenomenon that can be studied and simulated independent of its biological contexts, Artificial Life researchers hold that cognition is inseparable from life processes and that the path to AI might be found by modeling -- on computers and in robots -- the evolutionary and vital dynamics that support mental life. This essay is my report and reflection on the San Sebastian workshop, which I attended as part of an ethnographic project I am conducting among Artificial Life scientists.

Artificial Life was christened in 1987 in New Mexico by Chris Langton, a researcher at Los Alamos National Laboratories and at the Santa Fe Institute for the Sciences of Complexity. Artificial Life, in its U.S. incarnation, takes as its charter the idea that life "is a property of the organization of matter, rather than a property of matter itself,"[2] maintaining, as a consequence, that life may be faithfully modeled -- or more ambitiously, created -- on computers. In the last few years, a European school of Artificial Life has emerged which, while deriving inspiration from its counterpart across the Atlantic, has its own set of histories and questions. Participants at the Spanish workshop were eager to point out: 1) that "Artificial Life" was not born ex nihilo in New Mexico, but has descended from tangled international lineages of cybernetics, AI, self-organization theory, and theoretical biology; 2) that the sociologically contingent birth of Artificial Life at Santa Fe has focused too much attention on computational conceptualizations of life (the experimental apparatus of choice at the Institute is the computer rather than the wet lab or the robot): and 3) that Artificial Life should not be about generating total explanations or replications of life processes, but about using modeling/simulation techniques to test intuitions about aspects of life.

For many Artificial Life researchers -- particularly lab biologists and roboticists -- computational understandings of life are as problematic as computational conceptions of intelligence. San Sebastian conferees noted that notions of life as an abstract, formal process sound curiously close to AI's pictures of cognition. Standard AI and standard Artificial Life deny the embodied character of cognition and life, and both mistake symbols systematically interpretable as a phenomenon for the phenomenon itself. In Artificial Life, this allows some workers to convince themselves that self-reproducing computer programs are alive.

If there was one theme that characterized the San Sebastian conference, it was that life cannot be understood as purely computational and cannot be appreciated independent of its embodiment. One of the organizers of the conference, biologist Francisco Varela, contended that living systems produce the identities they have through histories of "structural coupling" to an environment that they themselves help call forth. Understanding a living thing requires that the reference point of the system be taken into account, not that the system be objectified. Theoretical biologist George Kampis argued that just because both life and computation can be described as processes does not mean that life can be reduced to computation. To ask if life can be fully defined by a particular formalism is like asking if a map is the territory. In his talk, "Situated Embodied Autonomous Agents: A Natural Artificial Intelligence," roboticist Tim Smithers spoke of building robots that acted effectively in an unpredictable real world and did so without using internal representations of that world. Philosopher Evan Thompson noted that, in living systems, form and matter can't be easily separated. He declared that "'Strong' Artificial Life is a bizarre mathematical Platonism in which only form matters."

Participants were concerned not just with examining alternatives to computationalism in AI and Artificial Life, but with fastening on why computational approaches had been so prevalent. They were keenly aware that the naming of Artificial Life on the analogy to Artificial Intelligence had encouraged formalistic descriptions of life. As practitioners operating on the periphery of a U.S. centered discipline, they also realized that the institutional power of the Santa Fe Institute was decisive in setting the field's methodological course. A few European researchers told me that they thought that some American scientists could afford to remain ignorant of the many historical precedents for Artificial Life -- and the many non-computational visions of life available in self organization theory, theoretical biology, and cybernetics -- because of their privileged position in the global economy of science.

I would like now to reflect on politics, post-modernity, and Artificial Life. If traditional AI has been accused (for example, by Joseph Weizenbaum, author of one of the first AI programs, ELIZA) of supporting an objectivist, instrumentalist epistemology which has dangerous political consequences (e.g. the computerization of nuclear weapons networks using game-theoretic logic), we might ask whether approaches to AI and Artificial Life that emphasize embodiment, situatedness, and historical contingency could articulate to different political and epistemological commitments.

Some ideas proposed at the San Sebastian workshop can be read as having resonances with certain post-modern/progressive (not always identical) thought -- a possibility that some scientists seem happy with. They are pleased that Artificial Life might question foundational stories about life, especially as simulation practices push people to recognize the socially constructed character of the very realities they are trying to simulate. Embodied robots, for example, are built with dispersed control systems. This questions the notion that action must proceed from a unified, directed, all-knowing agency. Varela's contention that organisms are located, with identities that emerge from histories of interactions with similar and dissimilar others, could advance views of evolution that stress contingent connections rather than narratives of progress. Discussions of embodiment and situatedness in non mainstream Artificial Life might call to mind the work of feminist science critic Donna Haraway, who insists that an epistemology that is tethered to socially located lives (rather than abstract, "transcendent" ones) is the only kind that can make sense and that can be politically enabling.[3] We might discern a sympathetic philosophy in Varela, Thompson, and Rosch's book on cognitive science, The Embodied Mind, where they write: "Freedom does not mean escape from the world; it means transformation of our entire way of being, our mode of embodiment, within the lived world itself."[4]

Both mainstream and alternative constructions of Artificial Life promise to blur the borders between the natural and the artificial, a project that may help us see "nature" as a precipitate of our social and cultural imaginations and activities. This is a view Donna Haraway advocates: "If the world exists for us as 'nature,' this designates a kind of relationship, an achievement among many actors, not all of them human, not all of them organic, not all of them technological."[5] Such a picture of nature can promote a more responsible engagement with the world, for no part of it is ever separate from us. Chris Langton has proposed that Artificial Life is not just about life-as-we-know-it, but about life-as-it-could-be. There are provocative possibilities here, especially if we consider that Artificial Life technology could be used to refashion our modes of embodiment in radical or subversive ways.

It would be simpleminded, however, to see either dominant or alternative Artificial Life discourse as connecting to progressive political agendas just because there are some words and theoretical constructions that resonate with the formulations of recent left academic writing. Describing life as situated and embodied, dismantling the distinctions between nature and artifice, and claiming interest in possible biologies are not necessarily liberating practices in themselves. Artificial Life ideas, if appropriated by the wrong institutions -- the military, the world "development" industry, socially irresponsible corporations -- could certainly be used in regressive ways. Artificial Life, and the new AI it supports, could help build autonomous, "self-reproducing" robot tanks; could be used for multinational agribusiness surveillance of "development" projects; and could rationalize oppressive manufacturing regimes along "adaptive" and "evolutionary" lines. I would also say that computer models of biology will be powerfully inflected by researchers' conceptions of gender, sexuality, race, and economy, and that these will tend to reflect perspectives grounded in positions of relative privilege, just as "objective" rationalist views of intelligence regnant in AI linked up with the rationalities of people -- like defense intellectuals -- who were empowered to make decisions about the lives of people who were very differently positioned in global structures of social inequality.

How new theories in AI and Artificial Life connect with the political and practical world in which they are necessarily embedded remains to be seen. But tracking these articulations will require attention to issues of power and language, and demand that diaproceed with an awareness that science is situated in culture.

Previous Next Up Comments

Notes

1. "Artificial Life: A Bridge Towards a New Artificial Intelligence," a workshop sponsored by the Department of Logic and Philosophy of Science, University of the Basque Country, Donostia/San Sebastian, Spain, was held on December 10-11, 1993.

2. Chris Langton, "Toward Artificial Life," Whole Earth Review, 58 (1988) 74.

3. Donna Haraway, "Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective," Simians, Cyborgs, and Women (New York: Routledge, 1991) 183-201.

4. Francisco Varela, Evan Thompson and Eleanor Risch, The Embodied Mind: Cognitive Science and Human Experience (Cambridge: MIT Press, 1991) 234.

5. Donna Haraway, "The Promises of Monsters: A Regenerative Politics for Inappropriate/d Others," Cultural Studies, ed. Lawrence Grossberg, Gary Nelson and Paula Treichler (New York: Routledge, 1991) 297.