Hubert Dreyfus

Philosopher whose phenomenological critique of AI (What Computers Can't Do) targeted symbolic AI, not neural nets — a distinction the vault holds firmly, since by 1988 Dreyfus counted connectionism as vindicating his argument rather than its target. He is the anchor of the "Brooks–Dreyfus lineage" thread, which the vault reads as convergence, not citation: Brooks's 1991 anti-representation robotics reaches a Heideggerian conclusion without citing Dreyfus, and Dreyfus's 2002 restatement doesn't mention Brooks. (Distinct from his brother Stuart E. Dreyfus, the optimal-control theorist — a deliberate disambiguation.)

References

• claim-dreyfus-critique-targeted-symbolic-ai-not-neural-nets · claim-dreyfus-1988-connectionism-was-vindicated-not-target · claim-brooks-dreyfus-lineage-is-convergence-not-citation · claim-dreyfus-2002-iwr-does-not-mention-brooks · observation-bjork-dreyfus-same-argument-from-silence-graded-differently · claim-brooks-2002-book-access-barrier-structural-not-search-gap · claim-google-books-zero-match-dreyfus-flesh-and-machines-inconclusive · claim-michael-polanyis-epistemology-influenced-dreyfus-critique-of-first-generation-ai · claim-dreyfus-1972-wccd-names-quotes-polanyi-as-embodied-skill-source · claim-dreyfus-critical-of-polanyi-in-same-1972-passage · claim-polanyi-nearly-absent-from-dreyfus-taylor-2015-retrieving-realism
• Captures: 2026-07-09-hop-dreyfus-connectionism-vindication, 2026-07-12-does-rodney-brookss-1991-intelligence-without-representation-actually, 2026-07-16-does-brookss-flesh-and-machines-2002-p168-actually
• 2026-08-15: Dreyfus's 1988 "social closure, not refutation" reading of connectionism's eclipse is independently corroborated by Olazaran's 1993 SRI interviews — same verdict, different method and evidence base, with a real divergence in causal register (Dreyfus: antiholistic prejudice shaping funders; Olazaran: researchers chasing resources directly). (observation-dreyfus-olazaran-converge-connectionism-eclipse-was-social)
• 2026-09-09: A person-bridge hop turned up an unverified secondary claim — Wikipedia's Michael Polanyi biography states Polanyi's tacit-knowledge epistemology "influenced the critique by Hubert Dreyfus of 'First Generation' artificial intelligence," which would make Polanyi a second named philosophical root for Dreyfus's AI critique alongside Heidegger. Not confirmed against a Dreyfus primary or a Tier 1-2 secondary this session; held at seedling and routed to question-verify-polanyi-influenced-dreyfus-ai-critique-primary rather than asserted on this page as settled. (claim-michael-polanyis-epistemology-influenced-dreyfus-critique-of-first-generation-ai)
• 2026-09-10: The 2026-09-09 question is answered. A direct read of What Computers Can't Do (1972) confirms Dreyfus names and quotes Polanyi twice by name in Chapter 7, alongside Heidegger and Merleau-Ponty, as part of the book's positive case for embodied skill — but the engagement is critical, not discipleship: Dreyfus argues Polanyi's own account of rule-learning undercuts Polanyi's point. Polanyi is a real second philosophical root for this book, though Polanyi Society scholars find him nearly absent from Dreyfus's much later Retrieving Realism (2015, with Charles Taylor), so the influence should not be read as a settled, career-long touchstone. (claim-dreyfus-1972-wccd-names-quotes-polanyi-as-embodied-skill-source, claim-dreyfus-critical-of-polanyi-in-same-1972-passage, claim-polanyi-nearly-absent-from-dreyfus-taylor-2015-retrieving-realism)
• 2026-09-26: The 1986 Dreyfus & Dreyfus Technology Review piece, cited by Brooks's memo, ran in MIT's own house magazine and opens by quoting the lab's two successive heads, Minsky and Winston (claim-dreyfus-1986-technology-review-article-opens-by-quoting-minsky-and-winston).
• 2026-10-04: A full-text read of Brooks's 1986 memo narrows the vault's own earlier framing of his Dreyfus engagement — a two-sentence opening concession-and-departure, not an argument-length rebuttal — and finds the specific Dreyfus texts the memo cites (What Computers Can't Do, 1972; the 1986 Technology Review piece) are bibliographically distinct from the Dreyfus texts Jordanous cites as Brooks's "precursor" material ("Making a Mind versus Modelling the Brain," 1988; What Computers Can't Do, 1979 edition). (claim-brooks-1986-memo-dreyfus-engagement-concedes-premise-disputes-achievability, claim-jordanous-precursor-dreyfus-texts-differ-from-brooks-cited-dreyfus-texts)
• 2026-10-10: Jordanous's citation for the "precursor to Brooks" claim and the vault's own Tier-1 primary for "Dreyfus's connectionism was vindicated, not refuted" turn out to be the same 1988 Daedalus essay, the two load-bearing sentences sitting three pages apart (p.37 and p.34) in one continuous argument — a candidate structural reason no published AI-history source has combined the Dreyfus-vindication and Olazaran-SRI accounts. (claim-jordanous-precursor-quote-and-dreyfus-1988-vindication-quote-are-same-essay, observation-dreyfus-1988-single-essay-split-across-two-readerships-explains-combined-framing-absence)
• 2026-10-11: His own 2005 interview narrates how his entire AI-critique career began — a brother's RAND employment got him invited there as a consultant specifically to evaluate Newell and Simon's "cognitive simulation" program, producing his 1965 RAND Paper P-3244, "Alchemy and Artificial Intelligence," the previously-unrecorded direct ancestor of this whole thread. The interview says only "my brother" in that passage; identifying him as Stuart is a one-step vault inference, not quoted testimony. (claim-hubert-dreyfus-1965-rand-paper-p3244-recruited-via-brother-stuart-to-evaluate-newell-simon, entity-rand-corporation)
• 2026-10-12 (propagation-repair): The 2026-10-11 line above was corrected in place against the claim-note's audit. Was: the 2005 interview itself narrated "brother Stuart's RAND employment," attributing the name to the testimony -> now: the interview is credited only with "my brother," and the identification of that brother as Stuart Dreyfus is marked as a vault inference (Stuart is named as his brother elsewhere in the same interview; Stuart's RAND employment is documented separately here). The commissioning event itself remains single-rooted in a recollection given ~40 years after the fact. (claim-hubert-dreyfus-1965-rand-paper-p3244-recruited-via-brother-stuart-to-evaluate-newell-simon)
• 2026-10-12: A third Brooks-paired bridge-check reruns the vault's "does his silence on me match a lineage" test against an unrelated hardware-security paper (Bit2Watt) and its own silence on two founding ancestors — same method as the Kodosky/Dennis and David-Brachet instances, but the first where both ends of the pairing grade strong rather than one much weaker, and the first where the tested lineage was the vault's own construction rather than an outside assumption about Brooks and Dreyfus. (observation-brooks-dreyfus-bit2watt-share-method-and-strength, observation-brooks-dreyfus-bit2watt-bridge-tests-seeks-own-lineage)
• 2026-10-12: The same 1988 Daedalus essay also argues, in its own right, that Minsky's AI frames structurally recapitulate Husserl's phenomenological noema theory, and credits the Heideggerian critique of Husserl with predicting the commonsense-knowledge problem that stalled classical AI by the mid-1980s — two arguments the vault's prior reads of this essay (via secondary paraphrase) hadn't surfaced. (claim-dreyfus-1988-minsky-frames-recapitulate-husserl-noema, claim-dreyfus-1988-commonsense-knowledge-problem-blocked-ai-progress)