Dreyfus, Hubert

Dreyfus, Hubert

(1929–  ) American

Philosopher, Cognitive Psychologist

As the possibilities for computers going beyond “number crunching” to sophisticated information processing became clear starting in the 1950s, the quest to achieve artificial intel-ligence (AI) was eagerly embraced by a number of innovative researchers. For example, Allen Newell, Herbert Simon, and Cliff Shaw at the RAND Corporation, attempted to write programs that could “understand” and intelligently manipu-late symbols rather than just literal numbers or characters. Similarly, MIT’s Marvin Minsky (see Minsky, Marvin) was attempting to build a robot that could not only perceive its environment, but in some sense understand and manipulate it. (See artificial intelligence and robotics.)

Into this milieu came Hubert Dreyfus, who had earned his Ph.D. in philosophy at Harvard. Dreyfus had special-ized in the philosophy of perception (how meaning can be derived from a person’s environment) and phenomenology (the understanding of processes). When Dreyfus began to teach a survey course on these areas of philosophy, some of his students asked him what he thought of the artificial intelligence researchers who were taking an experimental and engineering approach to the same topics the philoso-phers had discussed abstractly.

Philosophy had attempted to explain the process of per-ception and understanding (see also cognitive science). One tradition, the rationalism represented by such think-ers as Descartes, Kant, and Husserl took the approach of formalism and attempted to elucidate rules governing the process. They argued that in effect the human mind was a machine (albeit a wonderfully complex and versatile one). The opposing tradition, represented by the phenomenolo-gists Wittgenstein, Heidegger, and Merleau-Ponty, took a holistic approach in which physical states, emotions, and experience were inextricably intertwined in creating the world that people perceive and relate to.

If computers, which at that time had only the most rudimentary “senses” and no emotions could perceive and understand in the way humans did, then the rules-based approach of the rationalist philosophers would be vindi-cated. But when Dreyfus had examined the AI efforts, he wrote a paper titled “Alchemy and Artificial Intelligence.” His comparison of AI to alchemy was provocative in that it suggested that like the alchemists, the modern AI research-ers had met with only limited success in manipulating their materials (such as by teaching computers to perform such intellectual tasks as playing checkers and even prov-ing mathematical theorems). However, Dreyfus concluded that the kind of flexible, intuitive, and ultimately robust intelligence that characterizes the human mind couldn’t be matched by any programmed system. Each time AI research-ers demonstrated the performance of some complex task, Dreyfus examined the performance and concluded that it lacked the essential characteristics of human intelligence. Dreyfus expanded his paper into the book What Computers Can’t Do. Meanwhile, critics complained that Dreyfus was moving the goal posts after each play, on the assumption that “if a computer did it, it must not be true intelligence.”

Two decades later, Dreyfus reaffirmed his conclusions in What Computers Still Can’t Do, while acknowledging that the AI field had become considerably more sophisticated in creat-ing systems of emergent behavior (such as neural networks).

Currently a professor in the Graduate School of Phi-losophy at the University of California, Berkeley, Dreyfus continues his work in pure philosophy (including a com-mentary on phenomenologist philosopher Martin Hei-degger’s Being and Time) while still keeping an eye on the computer world in his latest publication, On the Internet.