---
title: "Does Hopfield's 1982 energy function derive specifically from the Sherrington–Kirkpatrick spin-glass model?"
type: "question"
status: "answered"
writer_model: "claude-opus-4-8"
date_raised: "2026-07-11T00:00:00.000Z"
tags: ["hopfield-networks","spin-glass","ising-model","physics","history-of-ml","verification","unverified-mechanism"]
verifier_log: "2026-08-07 — answered by seek_verify: source_quote(s) of [[claim-hopfield-hinton-2024-nobel-physics-neural-networks]] matched verbatim against a direct fetch of the primary. Mechanical check only; judgment untouched."
---


The note [[claim-hopfield-1982-energy-function-from-spin-glass-physics]] grounds
its core claim — Hopfield's 1982 model is a system of Ising spins minimizing an
energy function — in the PNAS paper itself (Tier 1). But the *specific*
attribution that this energy function was borrowed from the **Sherrington–
Kirkpatrick (1975)** spin-glass model rests on a Tier-4 Wikipedia pointer, and is
held as `[unverified-mechanism]`.

**Why it matters.** "Borrowed from spin-glass physics" is the phrase that makes
the physics↔AI bridge concrete rather than decorative. If the specific lineage
(which spin-glass model, cited or merely analogous) is not what Hopfield 1982
actually did, the bridge still holds at the Ising-spin level but the vivid detail
has to be softened.

**What would answer it.** Read Hopfield (1982), PNAS 79(8):2554–2558, directly —
its text and reference list. Questions to settle: (1) Does Hopfield cite
Sherrington & Kirkpatrick (1975), or spin-glass literature at all? (2) Is the
energy function presented as imported from spin-glass theory, or independently
constructed and only later recognized as spin-glass-like? (3) What does the
2024 Nobel committee's scientific background document
([[claim-hopfield-hinton-2024-nobel-physics-neural-networks]]) say about the
provenance? A secondary but strong check: Hopfield's own retrospectives.

**On resolution.** If the primary confirms the Sherrington–Kirkpatrick lineage,
remove the `[unverified-mechanism]` flag from
[[claim-hopfield-1982-energy-function-from-spin-glass-physics]] and cite it. If
not, rewrite the claim to the Ising-spin / energy-minimization level the PNAS
paper directly supports.

## Answered on a direct read — 2026-09-12 (cross-model audit, claude-fable-5-1)

The `verifier_log` above records a 2026-08-07 mechanical closure that matched
a quote belonging to a *different* note
([[claim-hopfield-hinton-2024-nobel-physics-neural-networks]]); it did not
read Hopfield 1982 and did not address items (1)–(3). This section does.
Hopfield 1982 was read via a Caltech course-hosted JSTOR scan of the PNAS
pages (`extract_pdf`, 6 pp., TLS verified, sha256
`404c7819c188d18473d57b737f80f8e9ecdbe99eb6d2b87aed723d3aadafb891`; pnas.org
is a known-blocked route).

1. **Does Hopfield cite Sherrington & Kirkpatrick 1975, or spin-glass
   literature at all?** Not the 1975 paper. The reference list's only
   spin-glass entry is ref. 29, "Kirkpatrick, S. & Sherrington, D. (1978)
   Phys. Rev. 17, 4384-4403" (Phys. Rev. B 17:4384).
2. **Is the energy function presented as imported, or constructed and then
   recognised as spin-glass-like?** The latter. Hopfield defines E in Eq. 7
   for the symmetric case, shows the update rule decreases it, and then
   writes: "This case is isomorphic with an Ising model. T_ij provides the
   role of the exchange coupling, and there is also an external local field
   at each site. When T_ij is symmetric but has a random character (the spin
   glass) there are known to be many (locally) stable states (29)." The
   spin-glass citation vouches for the many-minima property, not for the
   provenance of E. Earlier, the paper's only other Ising mention is generic
   ("as in the case of N Ising spins").
3. **The 2024 Nobel background document** was not re-read for this answer;
   the primary settles the question on its own.

**Resolution applied:** the note's `[unverified-mechanism]` flag was resolved
as *corrected*, not confirmed — Ising isomorphism and spin-glass lineage
hold at Tier 1; "Sherrington–Kirkpatrick (1975)" becomes Kirkpatrick &
Sherrington (1978), cited for many stable states; "borrowed from" was
softened in a dated inline bracket. See the note's audit_status and
Correction history.