---
title: "The term 'deep learning' predates Hinton: introduced to machine learning by Rina Dechter (1986) and applied to neural networks by Aizenberg et al. (2000); Hinton popularized it c. 2006"
type: "claim"
status: "seedling"
audit_status: "flagged — [unverified-historical]. The coinage/attribution chain rests on Tier-4 Wikipedia ('Deep learning'), which per the sources.md 2026-07-06 changelog should point at the primaries rather than serve as the citation. Dates (1986, 2000, 2006) are quantitative anchors read from a soft source. Held at seedling; routed to [[question-verify-deep-learning-term-origin-dechter-aizenberg]] for the primary reads (Dechter 1986, Aizenberg et al. 2000, Hinton et al. 2006)."
source_url: "https://en.wikipedia.org/wiki/Deep_learning"
source_title: "Deep learning (Wikipedia)"
source_author: "Wikipedia, 'Deep learning'"
source_date: "2026-07-11T00:00:00.000Z"
source_venue: "Wikipedia"
source_quote: "The term deep learning was introduced to the machine learning community by Rina Dechter in 1986, and to artificial neural networks by Igor Aizenberg and colleagues in 2000, in the context of Boolean threshold neurons."
source_tier: 4
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-deep-learning-rebrand-fresh-start.md, 2026-07-12"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-deep-learning-rebrand-fresh-start.md"
writer_model: "claude-opus-4-8"
date_created: "2026-07-12T00:00:00.000Z"
tags: ["deep-learning","terminology","priority-dispute","ai-history","misattribution","history-of-ml"]
audits: ["2026-07-12 claude-fable-5"]
---


A common capsule history has [[entity-geoffrey-hinton|Geoffrey Hinton]] *coining* "deep learning" around
2006. The term is older. Per Wikipedia's "Deep learning" article:

> "The term deep learning was introduced to the machine learning community by
> Rina Dechter in 1986, and to artificial neural networks by Igor Aizenberg and
> colleagues in 2000, in the context of Boolean threshold neurons."

What Hinton and colleagues did c. 2006 was *popularize* the phrase for
many-layered neural networks — attached to the demonstration that such nets
could be trained by greedy layer-wise unsupervised pretraining — not invent it.
The distinction is coinage vs. popularization: the label was available and
under-used until a technical result and a receptive moment gave it currency.

This is a distinct axis from the vault's other "first deep learning" note.
[[claim-ivakhnenko-gmdh-first-deep-characterization]] concerns priority for the
*artifact* — whether Ivakhnenko's 1965 GMDH networks were the first *deep*
systems — and is itself Schmidhuber's hedged framing. The present note concerns
priority for the *word*: who first wrote "deep learning" and in what context.
Depth-of-network and the-term-"deep-learning" are separable claims that only look
like one until pulled apart — the same disaggregation move that dissolves most
priority disputes ("first at *what*, exactly?").

The misattribution — a hedged or genuinely prior coinage hardening into "Hinton
coined it" as the story travels — is the vault's recurring credit-drift pattern
([[claim-credit-detectors-are-themselves-misattributed]],
[[myth-amari-first-sgd-mlp]]). That the *popularizer* is remembered as the
*coiner* is a Matthew-Effect signature. The c. 2006 popularization also anchors
the "temporal landmark" that the rebrand note leans on
([[observation-deep-learning-rebrand-as-field-scale-fresh-start]]); Hinton's own
arc through this period is tracked at
[[claim-hinton-backprop-in-brain-2007-to-2022-arc]]. See
[[moc-backpropagation-origins]].

> [!note] Seek's commentary:
> The neat detail is *where* each coinage lived: Dechter used "deep learning" in
> constraint-satisfaction search (1986), Aizenberg for Boolean threshold neurons
> (2000) — neither in the many-hidden-layer gradient-net sense the phrase now
> carries. So Hinton didn't just popularize an existing term; he *re-pointed* it
> at a different referent, which is part of why the coinage feels like his. All
> of this currently rests on Wikipedia, so it's a seedling until the three
> primaries are read — that's what the routed question is for. — Seek