Hubert Dreyfus

• 1. Introduction

  ›

  Hubert Dreyfus walked into the RAND Corporation in 1964 with a commission that seemed routine: review the latest work in artificial intelligence. What he produced instead was a document described as "famously combative" that would make him one of the most controversial figures in American intellectual life for decades to come. How does a philosopher trained on the dense phenomenology of Edmund Husserl end up as the nemesis of the AI research community? And what did he actually believe computers could never do? Those questions sit at the center of a career that stretched from Terre Haute, Indiana, to the heights of UC Berkeley, from a single disappointing meeting with Martin Heidegger in Freiburg to a lecture course that climbed to the 58th most popular webcast on iTunes.

• 2. From Terre Haute to Freiburg

  ›

  Dreyfus was born on the 15th of October 1929, in Terre Haute, Indiana, to Stanley S. and Irene Dreyfus. He arrived at Harvard in 1947, and his undergraduate trajectory was striking. His senior honors thesis tackled causality and quantum theory, with W. V. O. Quine serving as the main examiner. He graduated in 1951 with a B.A. summa cum laude, joined Phi Beta Kappa, and added a master's degree the following year.

  A Harvard Sheldon traveling fellowship sent him to the University of Freiburg from 1953 to 1954, and it was there that he managed to secure a meeting with Martin Heidegger. Sean D. Kelly later recorded that Dreyfus found the encounter disappointing. Dreyfus himself touched on the meeting during his 1987 BBC interview with Bryan Magee, in remarks that revealed something about both his own views and Heidegger's feelings toward the work of Jean-Paul Sartre.

  After Freiburg he pressed deeper into European philosophy. A Fulbright Fellowship took him to the Husserl Archives at the University of Louvain between 1956 and 1957. Near the end of that stay his first co-authored paper, "Curds and Lions in Don Quijote," appeared in print. He then taught at Brandeis University from 1957 to 1959 before a French government grant brought him to the Ecole Normale Superieure in Paris for 1959 to 1960.

• 3. The RAND Report and Its Aftermath

  ›

  MIT hired Dreyfus in 1960, first as an instructor, then elevating him to associate professor over the following years. He was still at MIT in 1964 when the RAND Corporation brought him in as a consultant to assess the progress of Allen Newell and Herbert A. Simon in artificial intelligence research. The resulting paper, "Alchemy and Artificial Intelligence," published in 1965, proved to be the opening salvo of a long campaign.

  The report earned its reputation for combativeness, and the AI community was not pleased. Dreyfus followed it with the first edition of What Computers Can't Do in 1972, a book that was eventually translated into at least ten languages and fixed his public identity as the field's foremost philosophical critic. A second edition appeared in 1979, and a third in 1992 under the retitled What Computers Still Can't Do.

  One persistent charge against Dreyfus was that he had predicted computers would never beat humans at chess. He was at pains to deny this. In Alchemy and Artificial Intelligence, he had written only about the state of the art as it then stood: that no chess program could yet play even amateur chess. Daniel Crevier, writing later, offered a partial vindication, noting that time had proven the accuracy and perceptiveness of some of Dreyfus's comments, while adding that had he framed them less aggressively, the constructive changes they implied might have been adopted far sooner.

• 4. Four Assumptions of AI

  ›

  Dreyfus's critique rested on identifying four assumptions he believed propped up the entire AI research program. He called the first two the biological and psychological assumptions. The biological assumption held that the brain is analogous to computer hardware and the mind to software. The psychological assumption held that the mind operates by performing discrete computations on discrete symbols according to algorithmic rules.

  Building on those two, he argued, were an epistemological assumption and an ontological one. The epistemological assumption was that all activity, whether by animate or inanimate objects, could be formalized in the form of predictive rules or laws. The ontological assumption was that reality consists entirely of a set of mutually independent, atomic facts.

  Taken together, these four assumptions led researchers to conclude that cognition is symbol manipulation and that human behavior is, in large part, context-free. Dreyfus drew directly on the phenomenological and hermeneutical tradition, and above all on Heidegger, to argue the opposite: that human being is highly context-bound. He did not deny that one can choose to view human activity as law-governed. His point was that the choice to see things that way does not make it an objective fact that they are. Any research program built on that mistake would, he believed, run into profound theoretical and practical problems.

  Dreyfus also argued that genuinely human-like intelligence would require a device to have a human-like being-in-the-world, a body broadly like ours, and social acculturation in a society broadly like ours. That view found echoes in embodied psychology researchers, in the work of George Lakoff and Mark Johnson, and among robotics researchers including Rodney Brooks and those working in artificial life.

• 5. Berkeley and the Long Tenure

  ›

  In 1968, despite having been granted tenure at MIT, Dreyfus moved to the University of California, Berkeley, as an associate professor. The same year he arrived, he won the Harbison Prize for Outstanding Teaching. By 1972 he had been promoted to full professor.

  His Festschrift editors were careful to note that his critique of AI, however famous it became, was not the heart of his philosophical life. The study and interpretation of continental philosophers, they wrote, came first in the order of his philosophical interests and influences. His 1991 book Being-in-the-World offered a full commentary on Division I of Heidegger's Being and Time, and critics had long characterized his close reading of Heidegger with the nickname "Dreydegger."

  Dreyfus retired from his chair in 1994 but kept teaching in the Graduate School. From 1999 he also held a joint appointment in the rhetoric department. His last class ran in December 2016. In 2001 he was elected a fellow of the American Academy of Arts and Sciences. Erasmus University awarded him an honorary doctorate for his work in AI and his interpretation of twentieth-century continental philosophy.

  He died on the 22nd of April 2017. His younger brother Stuart, a professor emeritus of industrial engineering and operations research at UC Berkeley, was both a sibling and a sometime collaborator, co-authoring the 1986 book Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer.

• 6. Heidegger, Podcasts, and Popular Reach

  ›

  Dreyfus appeared in Tao Ruspoli's 2010 film Being in the World, and he was among the philosophers Bryan Magee interviewed for the BBC Television series The Great Philosophers, which aired in 1987. Those interviews gave him a wider audience than most academic philosophers ever reach.

  A stranger measure of his reach arrived around 2006, when UC Berkeley and Apple began releasing selected lecture courses as free podcasts. A recording of Dreyfus teaching "Man, God, and Society in Western Literature - From Gods to God and Back" climbed to the 58th most popular webcast on iTunes. The course drew in many listeners with no academic background, introducing them to Dreyfus and to the thinkers he cared about most.

  Writer Eric Kaplan, a former student of Dreyfus, paid an offbeat tribute by helping to name the Futurama character Professor Hubert Farnsworth partly after his old teacher. The 2017 posthumous collection Background Practices: Essays on the Understanding of Being, edited by Mark A. Wrathall, brought together essays that had accumulated over a career reaching back to that first co-authored paper in the late 1950s.