Ilya Sutskever

Ilya Sutskever (born 1986) is among the most consequential researchers in deep learning. Born in the Soviet Union and raised in Israel and Canada, he co-authored AlexNet with Alex Krizhevsky and Geoffrey Hinton in 2012 — the result that ignited the deep-learning revolution — then pioneered sequence-to-sequence learning at Google Brain. In 2015 he co-founded OpenAI as chief scientist, where his conviction that scaling neural networks would keep working underwrote the GPT series.

From the boardroom crisis to a straight shot

Sutskever’s belief in AGI was never casual — colleagues describe him leading chants of “feel the AGI” — and neither were his worries about it. He co-led OpenAI’s superalignment team, and in November 2023 joined the board vote that briefly fired Sam Altman, a decision he publicly said he deeply regretted within days. He left OpenAI in May 2024, and a month later announced Safe Superintelligence Inc. with Daniel Gross and Daniel Levy: no products, no revenue roadmap, one goal — safe superintelligence, pursued in “a straight shot,” insulated from commercial pressure. Investors valued the productless lab at a reported $32 billion by 2025.

At NeurIPS 2024 he told researchers that pre-training as we know it will end because data — the “fossil fuel” of AI — is finite, and that truly agentic, reasoning systems will be qualitatively different and less predictable. His position confounds simple labels: he is certain superintelligence is coming and certain it is dangerous, and his answer to both is to build it right.