AI as an anthropological technology

What does it mean to be human?

Note to readers: Thank you for your patience. The last month has been packed with travel, and it has been a while since I posted. I appreciate your ongoing support! — Matt

Early in the G.W. Bush administration, the White House established a President’s Council on Bioethics. It was led by one of my teachers, Leon Kass. Few of its recommendations found their way into policy, but the reports issued by the Council are superb examples of humanistic inquiry in the service of big public concerns. Under Kass’s leadership, the Council tasked itself with understanding what was at stake at a moment when emerging technologies such as cloning and gene editing promised new therapeutic possibilities, but promised also to give us new end-runs around the limits and natural forms of human embodiment. The removal of such limits was sure to come with big social implications.

A week ago, I participated in the first meeting of an Artificial Intelligence Ethics Council convened by the American Enterprise Institute. Though it doesn’t have the imprimatur of the White House, the Council has positioned itself as heir to the Kass effort: like the biotech moment of 25 years ago, our AI moment similarly cries out for deep and far-reaching reflection on what is at stake. At both junctures, the question is pressed upon us, what is a human being?

For biotech, the human animal is regarded as something highly plastic that can be modified and optimized according to some plan of our own. For AI, the founding aspiration (dating from the late 1940s) is to create a mechanized version of the mind. These two projects are in fact deeply entangled in the conceit that the genome operates like a computer program. Like such a program, the human code could be run on an indefinite variety of different hardware platforms, and there is nothing inherently special about carbon versus silicon. Biotech and AI are thus both what you might call anthropological technologies, in the sense that they express and advance a similar picture of the human.

As Jean-Pierre Dupuy put it in his incomparable book The Mechanization of the Mind: On the Origins of Cognitive Science, “For man to be able, as subject, to exercise power of this sort over himself, it is first necessary that he be reduced to the rank of an object, able to be reshaped to suit any purpose. No raising up can occur without a concomitant lowering, and vice versa.”

When I was writing about philosophy of mind and the question of a materialism twenty years ago, I liked to offer a simple thought experiment by way of orienting the discussion. If a dog bumps into an abacus and accidentally moves some of the beads, has the dog “done arithmetic?” Has the abacus? I think the answer to both questions has to be “no”. A human observer may choose to interpret the abacus’ change from one machine state to another as an instance of arithmetic having been done. It is important to know that a digital computer is not different from an abacus in any fundamental way. To think otherwise amounts to a crude animism.