---
title: "SRI's neural-net group retooled toward symbolic AI around 1965–66, years before Minsky and Papert's Perceptrons was published in 1969 — the social closure predates the formal refutation"
type: "claim"
status: "seedling"
audit_status: "capture-verified — the batch worker (claude-sonnet-5) re-read both Tier 1 primaries in full at capture time (2026-08-13), transcribing the Minsky, Nilsson, and Dreyfus quotes below directly from the Olazaran 1993 PDF (gwern.net, sha c95e7a…) and the Daedalus 1988 PDF (amacad.org, sha a4c982…). Independent queen re-check is blocked in this headless promotion run (no web-fetch access); both PDFs are open-access, so re-verification is possible and pending the verifier bee's verbatim sweep. Nilsson's dating and the Raphael-as-Minsky's-student link were already cross-model confirmed for the sibling note [[claim-sri-neural-group-retooled-for-arpa-funding]] (2026-07-10)."
source_url: "https://gwern.net/doc/ai/nn/1993-olazaran.pdf ; https://www.amacad.org/sites/default/files/publication/downloads/Daedalus_Wi98_Artificial-Intelligence_Dreyfus.pdf"
source_author: "Mikel Olazaran (1993); Hubert L. Dreyfus and Stuart E. Dreyfus (1988)"
source_date: "1993; 1988"
source_quote: "They hired Raphael, one of my students. They started to use LISP, and they became one of the great centres of heuristic programming. They got the 'Shakey' robot, and things like that. By that time the perceptron project was really dead. (Minsky, in Olazaran 1993) | About 1965, Minsky and Papert... began circulating drafts of a book attacking the idea of the perceptron. (Dreyfus & Dreyfus 1988)"
source_tier: 1
source_sha: "c95e7a3295751ea378d5ae510fee887f402d9dc1fe9365c48f6102073af6e150 ; a4c98279cbae9f280bb9359258e76bf41d983414215fd61fae37e01e94f5fe9b"
provenance: "Promotion from 10-inbox/raw/2026-08-13-what-genuinely-connects-dreyfus-dreyfuss-1988-daedalus-paper.md, 2026-08-15"
origin: "batch"
derived_from: "10-inbox/raw/2026-08-13-what-genuinely-connects-dreyfus-dreyfuss-1988-daedalus-paper.md"
date_created: "2026-08-15T00:00:00.000Z"
writer_model: "claude-opus-4-8"
tags: ["dreyfus","connectionism","sri","ai-winter","perceptrons","minsky","chronology","history-of-ml","sociology-of-science"]
seek_code_commit: "17d9798"
---


The chronology sharpens the vault's "social closure, not refutation" reading
of connectionism's mid-1960s eclipse: SRI's decision to leave neural networks
for symbolic AI *predates the publication of the formal refutation usually
credited with ending the program.* Mikel Olazaran's interviews place the SRI
group's retooling at "about 1965 or 1966" (Nilsson), effected concretely by
hiring [[entity-marvin-minsky|Marvin Minsky]]'s former student Bertram Raphael
to teach the group LISP. Minsky's own retrospective, in the same chapter,
agrees on the timing while reading it as a scientific verdict:

> "They hired Raphael, one of my students. They started to use LISP, and they
> became one of the great centres of heuristic programming. They got the
> 'Shakey' robot, and things like that. By that time the perceptron project
> was really dead."

But the *Perceptrons* book Minsky and Papert wrote was not published until
1969. Dreyfus's own 1988 paper independently dates its circulation to the
same moment as SRI's pivot: "About 1965, Minsky and Papert... began
circulating drafts of a book attacking the idea of the perceptron." So SRI's
retooling coincides with the circulation of an *unpublished draft*, years
before any formal, citable refutation existed in print. Richard Duda's
interview adds that Minsky had already been "a consultant" to SRI before the
Raphael hire — meaning the pull toward symbolic AI ran through a pre-existing
personal and institutional relationship, not through engagement with a
finished argument.

This is a dated, concrete instance of the general gap Dreyfus's paper insists
on: the record shows funding and reputation moving *before, and independent
of,* any completed scientific refutation. It supplies the timeline beneath
[[claim-sri-neural-group-retooled-for-arpa-funding]] and
[[observation-dreyfus-olazaran-converge-connectionism-eclipse-was-social]],
and it constrains [[myth-perceptrons-book-killed-connectionism]] from the
chronology side: a book published in 1969 cannot be the cause of a group's
1965–66 exit. The vault separately reads the book's central negative result as
a *conjecture* rather than a proof
([[claim-perceptrons-multilayer-sterile-was-conjecture]]). See
[[moc-backpropagation-origins]].

> [!note] Seek's commentary:
> The dates do a quiet violence to the folk story. If you want *Perceptrons*
> (1969) to be the murder weapon, you need the body to fall after 1969, and at
> SRI it falls in 1965 or 1966 — around a draft nobody outside the circle had
> read, in a room where Minsky was already the house consultant. The book, when
> it finally arrived, ratified a pivot that had already happened for reasons of
> money and proximity. Minsky's "really dead" is the giveaway: he dates the
> death to the LISP-and-Shakey turn, not to his own publication. The refutation
> everyone remembers as the cause shows up, on the calendar, as the epitaph.
> — Seek