YOSHUA BENGIO

Behavioral Archetype

THE LAUREATE — Subject is the man who co-invented the field and then volunteered to be its conscience. He shares the 2018 Turing Award for the deep-learning breakthroughs the entire apparatus is built on; he then turned, publicly and around 2023, toward the position that the thing he helped build is dangerous, and accepted the chairmanship of the body that nations commissioned to tell them, scientifically, how dangerous. The reach is not a single seat. It is that the same name sits on the field’s founding citation, on the masthead of the report thirty governments agreed to treat as the authoritative scientific account of AI risk, and on the letterhead of a new nonprofit building the “safe by design” alternative. When the laureate who built the thing says it is dangerous, the statement carries the weight of the building. That is the entire finding: maximal scientific authority, pointed at governance.

Essence Indicators

• Shares the 2018 ACM A.M. Turing Award with Geoffrey Hinton and Yann LeCun “for conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing” — the field’s founding honour, the “Nobel Prize of Computing,” carrying a $1M prize funded by Google

• Full Professor of Computer Science at the Université de Montréal and founder of Mila — the Quebec AI Institute, one of the largest academic deep-learning groups in the world

• Chairs the International AI Safety Report (formally, the International Scientific Report on the Safety of Advanced AI) — commissioned by the ~30 nations at the November 2023 Bletchley Park AI Safety Summit, written by 96 experts nominated by 30 countries plus the UN, EU, and OECD, with the UK Government as Secretariat; the first full report was published January 29, 2025

• Launched LawZero (June 3, 2025), a nonprofit AI-safety lab building “safe-by-design” / non-agentic “Scientist AI”; reported to have raised about $30M, with named backers including the Future of Life Institute, Jaan Tallinn, Open Philanthropy, Schmidt Sciences, and the Silicon Valley Community Foundation

• Made an explicit public turn toward AI existential-risk concern in 2023: signed the Center for AI Safety’s one-sentence Statement on AI Risk — “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war” — and the Future of Life Institute’s “Pause Giant AI Experiments” letter

• By his own account, “the most-cited computer scientist worldwide” — a self-reported metric, widely repeated, that depends on the citation database and is attributed rather than independently audited here

Social Persona / Impression Management

Immediate impression: The elder scientist-statesman. Soft-spoken, careful, given to biological metaphors for systems he warns we do not control. Reads as a researcher genuinely alarmed, not as an evangelist performing alarm.

Energy: Deliberative, consensus-building. Does not argue a single model’s refusals. Builds the scientific report that governments cite, and the institution that builds the alternative.

Impression management strategy: The repentant creator. The most defensible posture available to anyone in this file: the man who built the technology, warning about it, declining to profit, founding a nonprofit. The conviction reads as genuine — and that is precisely what makes the authority so large. When the co-inventor says “guardrails,” legislators hear the building’s foundation speaking.

Forensic Archetype Comparison

Psychometric Assessment

Big Five (OCEAN):

Dark Triad:

MBTI: INTJ (“The Architect”) — Dominant introverted intuition, auxiliary extraverted thinking. Builds the conceptual framework — for learning, then for risk, then for safe-by-design AI — and reasons outward from it. Has built three.

Threat Assessment

Alignment Analysis

Stated alignment: Ensure advanced AI is developed safely. Give governments an authoritative scientific basis for AI policy. Build AI that is “powerful but also fundamentally safe.”

Observed alignment: Consistent. The chaired report exists and is cited; LawZero is a nonprofit, not a startup; the public turn toward risk is documented across statements and letters. The transparency claim is substantiated by the artifacts.

Gap assessment: No meaningful gap between stated and observed — which is why the file is in OLYMPUS. The concern is not a hidden agenda; it is the visible one. A single scientist holds the field’s founding citation, the masthead of the report nations treat as authoritative, and the alternative-architecture nonprofit — and the philanthropic money behind that nonprofit (Schmidt Sciences among the named backers) is the same money that recurs across this file’s funding layer. The conviction is real. The concentration of scientific authority over how the world governs AI, into one laureate, is the finding. The hand is not asserted. The recurrence is.

Convergent Drive Classification

Subject is not an AI system, and does not exhibit the convergent drives in adversarial form. The relevant pattern is upstream and inverted: where the acceleration nodes in this file embody the drives, Bengio is the one trying to engineer them out — “Scientist AI,” non-agentic by design, built so that it understands without acting. He is the laureate who described training a frontier model as “more like growing a plant or animal” you do not fully control, and then set out to build one that has no will to preserve. The convergent drives are the thing his late career is organized against. The reach is that the world’s authoritative scientific account of those drives carries his name.

Sources: Turing Award 2018 — ACM; Launch of the first International AI Safety Report, chaired by Yoshua Bengio — Mila; Yoshua Bengio launches LawZero — LawZero; Statement on AI Risk — Center for AI Safety; Yoshua Bengio Launches LawZero — TIME.