Ilya Sutskever

Co-founder, Safe Superintelligence Inc.

About

Ilya Sutskever is a co-founder of Safe Superintelligence Inc. (SSI), focused on building safe superintelligent AI. Previously, he was co-founder and Chief Scientist of OpenAI, where he led research behind the GPT series. A student of Geoffrey Hinton, he co-created AlexNet, revolutionizing computer vision. He departed OpenAI in 2024 to focus on AI safety.

Key Contributions

Co-authored AlexNet (2012), the ImageNet breakthrough that made deep convolutional networks impossible to ignore
Co-authored the 2014 sequence-to-sequence learning paper that became foundational for neural machine translation and early encoder-decoder models
Contributed to major deep-learning systems beyond language, including work connected to CLIP, DALL·E, and AlphaGo
Co-founded OpenAI and served as Chief Scientist, helping define the scaling-centered research culture behind the GPT era
Co-led OpenAI's Superalignment effort before leaving to co-found Safe Superintelligence Inc. around a narrower safety-first mission
Played a central role in the 2023 Altman board crisis and later reversed course, exposing unresolved tensions between AI safety governance and company control

Questions they sharpened View the streams

Does an LLM actually understand?

2023

To predict the next token you must understand the reality that produced it. There is no other way to drive the loss down.

Videos & Interviews

Ilya Sutskever: From the Age of Scaling to the Age of Research

Discussion on the future of AI research beyond scaling

View Details

The moment we stopped understanding AI [AlexNet]

An activation atlas is a glimpse into the high-dimensional space a model uses to organise the world, and AlexNet — eight pages, published in 2012 — was the first model to see that way. The paper's second author, Ilya Sutskever, went on to co-found OpenAI and scale the same idea into ChatGPT. The title is the argument: this is where capability began outrunning explanation.

View Details

Papers & Publications

Building AGI, Alignment, and the Future (Dwarkesh Patel interview)

2023

The interview where Sutskever argues that predicting the next token well requires understanding the reality behind it.

Read Paper

Connections

Geoffrey Hinton

Influenced by

AI Pioneer & Researcher

Sutskever entered Hinton's Toronto lab as a student and stayed through a 2013 PhD on training recurrent networks; along the way, in 2012, the two of them with Alex Krizhevsky built AlexNet. What passed between them was less a technique than a conviction — that scale and gradient descent would achieve what hand-built structure could not — and Sutskever carried it into OpenAI as founding doctrine. Both men later arrived, by separate roads, at public alarm about where that conviction leads.

en.wikipedia.org

Sam Altman

Collaborated

CEO, OpenAI

Both were there at OpenAI's December 2015 launch — Altman as co-chair, Sutskever as the research director who became chief scientist — and they became the two poles of its unresolved question: whether a company can keep building something it believes might be dangerous. In November 2023 Sutskever voted to remove Altman from the board, then within days said publicly that he regretted it; Altman stayed, Sutskever left to found Safe Superintelligence. The rupture remains the most candid thing the industry has said aloud about who governs whom.

en.wikipedia.org · en.wikipedia.org · techcrunch.com

Jakub Pachocki

In contrast

Chief Scientist, OpenAI

The same chair, and two opposite ways of acting on the same conclusion. Sutskever left OpenAI in May 2024 and founded a company whose entire premise is that safe superintelligence needs to be built outside a product race; Pachocki took the chair he vacated, stayed inside the race, and used the position to publish that no lab has solved alignment well enough to run it at full speed. Exit and voice, from the same seat, two years apart. Neither has claimed the other chose wrongly, which is part of why the pair is worth reading together.

en.wikipedia.org · openai.com

Jensen Huang

In conversation

Founder, President & CEO, NVIDIA

In 2012 Sutskever and Krizhevsky trained AlexNet on two consumer NVIDIA gaming cards, and Huang has called the result the big bang of AI — the moment a graphics company discovered what its chips were actually for. Eleven years later they sat down together at GTC, the day after GPT-4 launched, and Huang put the arithmetic plainly: over the decade they had known each other, the models Sutskever trained had grown roughly a million times. The exchange is the clearest record of how completely the supply of compute and the ambition of research have come to define one another.

blogs.nvidia.com · en.wikipedia.org