Yann LeCun

persona · active · confidence: medium · reviewed 2026-09-07 · tier: quarterly

Who he is

Affiliation: Co-founder & Chairman, AMI Labs (Advanced Machine Intelligence; Paris HQ, offices NY/Montreal/Singapore) — left Meta (departure announced 2025-11 after 12 years as Chief AI Scientist) and raised a $1.03B seed at $3.5B pre-money (2026-03; Cathay Innovation, Greycroft, Hiro Capital, HV Capital, Bezos Expeditions); CEO is Alexandre LeBrun, CSO Saining Xie, CRIO Pascale Fung. Still Silver Professor, NYU. Turing Award (2018, with Bengio & Hinton); pioneer of convolutional neural networks (1989+).
One-line position: Current LLMs are an off-ramp from human-level AI — "basically a dead end" — the path is non-autoregressive world-model architectures (JEPA), and x-risk concerns are overstated by people who don't understand how these systems work. He has now bet a billion dollars of other people's money on it.

Discipline & technical bet

Deep technical researcher turned founder. The wager, now institutional: V-JEPA-lineage world models trained on video and spatial data, not text — with early prototypes targeted within ~a year of the raise (~early 2027) and commercial systems years out ("fundamental research," per CEO LeBrun). First disclosed partner: Nabla (digital health), premised on LLM hallucinations being intolerable in medicine. The exit itself was a bet: he refused to report into Alexandr Wang's TBD Lab after Meta's 2025-06 Scale deal — "You certainly don't tell a researcher like me what to do."

Key claims (Says)

Notable predictions — with falsifiable checks

Revealed behavior (Does)

Feels

Vindication-seeking and liberated: a decade of being right early (ConvNets, SSL) fuels certainty that he's right early again; visible contempt for being managed and for the "LLM-pilled." The dead-end framing is identity now, not just analysis.

Hears

Academic vision/robotics researchers, the JEPA lineage (Xie, Fung, Rabbat followed him out), European tech-sovereignty circles (Paris HQ; Euronews frames him as "French AI godfather"), and his own X audience — notably no longer inside any frontier lab's information flow.

Sees

Where autoregressive scaling disappoints inside a hyperscaler — he watched Meta's Llama program from the inside through 2025 — and what video/embodied data can teach that text can't. Lost: visibility into frontier-lab internals going forward.

Incentive map

Inverted since v1. The old discount (Meta's strategic interest in open-source; AGI-skepticism reframing Meta's lag) is obsolete. New rule-25 structure: he is a founder who raised $1.03B on the thesis that LLMs are a dead end and world models are the path — every public statement is now also fundraising narrative for AMI. Cannot easily say: "LLM agents turned out more capable than I expected," "JEPA scaling is disappointing," or anything that validates the paradigm his investors paid him to displace. The healthcare partnership also gives him a commercial stake in maximizing LLM-hallucination salience.

Theories aligned with

What he's reacting against

Where he overlaps / splits (with Rich)

Track record

Empirical vs normative

Weak spots / open questions

Rich's take

Delta log

2026-09-07 — v1→v2 migration + validation (weekly batch)

Sources

migrated v1→v2 2026-09-07 · backup: _archives/yann-lecun.html.bak-20260907 · converts-from: personas/yann-lecun.md · AI & Society domain