---
id: "20260709-1835-hop-connectionist-diaspora-funding-and-industry"
title: "The other 1960s-80s connectionists didn't vanish -- they chased the funding or exited into industry"
type: "capture"
status: "promoted"
promoted_to: ["30-notes/claim-sri-neural-group-retooled-for-arpa-funding.md","30-notes/claim-nestor-inc-commercialized-neural-nets-through-ai-winter.md","30-notes/claim-hnc-falcon-fraud-manager-became-fico-infrastructure.md"]
questions_routed: ["50-questions/question-verify-fico-neural-majority-card-transactions.md"]
not_promoted: ["Hop-1 'rebranding to escape the AI funding stigma' (researchers renamed work 'machine learning'/'informatics' etc.) — no citable source (web-search aggregate, no single URL); fails the attribution floor. Left in inbox as a lead.","The 'two opposite survival strategies' synthesis frame (chase-funding vs. commercial-exit) — genuine but interpretive and thin (n=3); folded into the three notes' cross-links and commentary rather than made a standalone note.","Saved leads, not claims: Olazaran ch. 6 (Widrow) and ch. 8 (Rosenblatt) closure sections (unread this session), Marr's quoted reflection on leaving neural nets, and HNC founder Robert Hecht-Nielsen's UCSD biography — future-chain hooks."]
promoted_date: "2026-07-09T00:00:00.000Z"
origin: "hop-batch"
model: "claude-sonnet-5"
date_created: "2026-07-09T00:00:00.000Z"
hop_chain: ["seed: HOP SEED -- what happened to the OTHER 1980s connectionists (sociology/economics redirect, not biography)","seed -> AI-winter rebranding-to-escape-stigma search results (max_cosine 0.721)","rebranding hook -> Olazaran (1993) 'A Sociological History of the Neural Network Controversy' (max_cosine 0.656)","Olazaran ch. 6-8 'Closure of the Controversy' -> SRI group's funding-driven pivot, primary interview quotes (max_cosine 0.657, part of same source)","AI-winter funding search -> Nestor Inc, unfamiliar company name (max_cosine 0.602)","Nestor Inc -> broader survivor-company pattern, HNC Software / Falcon / FICO (max_cosine 0.657)"]
novelty_max_cosine: 0.761
tags: ["hop","connectionism","ai-winter","sociology-of-science","funding","neural-networks","history"]
source_1_title: "A Sociological History of the Neural Network Controversy"
source_1_url: "https://gwern.net/doc/ai/nn/1993-olazaran.pdf"
source_1_tier: 1
source_2_title: "Nestor's Neural Chip Destiny Now In Its Own Hands"
source_2_url: "https://www.techmonitor.ai/technology/nestors_neural_chip_destiny_now_in_its_own_hands"
source_2_tier: 3
source_3_title: "Where FICO Gets Its Data for Screening Two-Thirds of All Card Transactions"
source_3_url: "https://www.predictiveanalyticsworld.com/machinelearningtimes/where-fico-gets-its-data-for-screening-two-thirds-of-all-card-transactions/13648/"
source_3_tier: 3
---


## Core claims

**1. SRI's neural-network group didn't lose an argument -- they retooled toward where DARPA funding was flowing.** Mikel Olazaran's 1993 sociology-of-science chapter, based on direct interviews, documents SRI hiring Bertram Raphael (a Minsky student) specifically to teach them LISP so they could pivot from perceptron research into symbolic-AI robotics (the "Shakey" robot).
> "We hired Bert Raphael from MIT who taught us LISP. We were interested in learning LISP programming, and that started to be more interesting than neural networks." (Nilsson, interview, in Olazaran 1993, p.373)
— source_tier: 1

**2. Some connectionists didn't retool -- they left academia and built commercial infrastructure companies that never became famous.** Nestor Inc. (founded 1975 by Nobel laureate Leon Cooper and Charles Elbaum) commercialized neural nets continuously through the "dead" decades -- check reading, handwriting recognition, eventually traffic-camera enforcement -- yet by 1994, after 19 years, still "hasn't ever been profitable."
— source_url: https://www.techmonitor.ai/technology/nestors_neural_chip_destiny_now_in_its_own_hands — source_tier: 3

**3. A sibling company, HNC Software, took the same commercial-exit path and won big and invisibly.** Its 1992 Falcon Fraud Manager became the seed of what is now FICO's fraud-scoring infrastructure. `[unverified-quant -- needs primary]` "Neural network models created by FICO score a majority of both the credit and debit card transactions in the world today." — source_tier: 3

## Why this was hop-worthy
Two opposite, equally invisible survival strategies -- chase the winning paradigm's funding (SRI) vs. exit academia into commercial infrastructure (Nestor, HNC) -- turn out to be the actual answer to "what happened to them," and neither shows up in the standard Hinton/LeCun/Bengio revival narrative.

> [!note] Seek's commentary:
> The Olazaran interviews are unusually good primary material -- Rosen, Nilsson, Minsky and Duda all speaking in their own words about a career pivot most AI histories flatten into "symbolic AI won." Worth a return trip just for the Rosenblatt and Widrow closure chapters.

## Further leads
- Olazaran ch. 8 (Rosenblatt's own group's closure) and ch. 6 (Widrow, already partly in the vault) -- unread this session.
- HNC Software founder Robert Hecht-Nielsen (UCSD) -- not pursued, individual-biography neighborhood was over quota tonight.

## Hop chain

**Hop 1** — search results on AI-winter rebranding — (web search, no single URL)
- Hook type: surprising claim
- Hook: researchers deliberately renamed their work "machine learning," "informatics," "computational intelligence" etc. specifically to escape the funding stigma attached to the word "AI."
- Why followed: directly sociological/economic (labor-market survival strategy), matching the redirect note.
- Key findings: rebranding was a documented, deliberate funding-access tactic, not passive drift.

**Hop 2** — Olazaran, "A Sociological History of the Neural Network Controversy" (1993) — https://gwern.net/doc/ai/nn/1993-olazaran.pdf
- Hook type: mechanism question
- Hook: the chapter has dedicated sections titled "Closure of the Controversy" for Widrow's group, the SRI group, and Rosenblatt individually -- i.e. it directly answers "what happened to them," group by group.
- Why followed: primary sociology-of-science source (Tier 1, full text freely available via gwern.net), extends the vault's existing "Perceptrons book killed NN research" myth-ledger note with real mechanism.
- Key findings: read in full (91pp); the SRI section documents a funding-driven retooling, not a scientific refutation.

**Hop 3** — same source, SRI closure section (pp. 368-374)
- Hook type: surprising claim
- Hook: SRI's neural-net researchers hired a rival paradigm's student to teach them that paradigm's tools, on the record, in their own words.
- Why followed: strongest primary-interview evidence in the source; zoomed in on mechanism.
- Key findings: Rosen (SRI group lead) describes spending 3-4 months brainstorming "what project shall we select" purely to catch ARPA funding, landing on the Shakey robot; had to "sell" the pivot to people who "didn't like perceptrons."

**Hop 4** — search on AI-winter funding networks -> Nestor Inc
- Hook type: unfamiliar name
- Hook: "Nestor Inc," founded 1975 by Nobel physicist Leon Cooper, doing commercial neural-net pattern recognition through the entire supposed dead period.
- Why followed: unfamiliar company name in a familiar (AI-winter) context; zoomed out from a single research group to an industry-wide pattern.
- Key findings: Nestor sold sonar/radar classification and handwriting recognition commercially in the 1980s-90s while academia had abandoned the field.

**Hop 5** — Tech Monitor, "Nestor's Neural Chip Destiny Now In Its Own Hands," April 14, 1994 — https://www.techmonitor.ai/technology/nestors_neural_chip_destiny_now_in_its_own_hands
- Hook type: surprising claim
- Hook: after 19 years of commercial neural-net R&D, the company "hasn't ever been profitable."
- Why followed: quantified how brutal the commercial-exit path actually was.
- Key findings: Nestor's president David Fox described a strategic shift to "evolve from a research and development company into a product company" around the Ni1000 chip -- still chasing profitability two decades in.

**Hop 6** — search on 1980s neural-net startup survivors -> HNC Software / FICO
- Hook type: mechanism question / cross-domain bridge (a 1980s AI-winter startup -> today's global card-payment infrastructure)
- Hook: HNC's 1992 Falcon Fraud Manager, now folded into FICO (acquired HNC 2002), reportedly scores a majority of the world's card transactions.
- Why followed: zoomed out to confirm the commercial-exit pattern generalizes and has a genuinely enormous, invisible present-day footprint.
- Key findings: the chain's destination -- connectionist "failures" that quietly became load-bearing financial infrastructure nobody credits to 1980s neural-net research.

Saved hooks not followed:
- Marr's own quoted reflection on leaving neural nets for the symbolic paradigm in the early 1970s (Olazaran, same chapter) — individual-biography-shaped, over quota tonight.
- Widrow's and Rosenblatt's own "Closure of the Controversy" sections in Olazaran (ch. 6, 8) — unread, good seed for a future chain.
- HNC founder Robert Hecht-Nielsen's UCSD obituary — person-behind-the-thing hook, skipped per redirect.

post-worthy: maybe — a genuinely new sociology/economics finding (funding-chasing vs. commercial-exit as the two real "what happened to them" answers) with one strong Tier 1 primary source, but claim 3's headline stat needs a better citation before it can carry weight.