Ilya Sutskever leaves OpenAI after nearly a decade

Ilya Sutskever, OpenAI cofounder and chief scientist, announced Tuesday that he is leaving the company after nearly a decade. It is a significant departure for the company that popularized ChatGPT: Sutskever has been one of the most influential scientific figures in the development of modern language models and played a central role in OpenAI’s leadership crisis last November.

In a message posted on X, Sutskever described OpenAI’s trajectory as extraordinary and expressed confidence that the company will build safe and beneficial artificial general intelligence — AI capable of handling a wide range of intellectual tasks — under the leadership of Sam Altman and Greg Brockman. He also said he is working on a personal project that he will share more details about later.

OpenAI said that Jakub Pachocki, previously its head of research, will take over as chief scientist.

One of the architects of today’s OpenAI

Sutskever founded OpenAI in 2015 alongside Altman, Brockman, Elon Musk and other researchers and entrepreneurs. He had previously worked at Google, where he contributed to foundational research in deep learning, the technique that allows machines to find patterns in vast amounts of data.

His name is linked to some of the breakthroughs that laid the groundwork for today’s systems. He coauthored influential papers on neural networks for computer vision and on neural machine translation. At OpenAI, he led a substantial portion of the research that produced the GPT families, the models that power ChatGPT.

The importance of that track record goes beyond the titles. In just a few years, OpenAI has become a company with mass-market products and enormous influence over the commercial direction of generative AI. The departure of its top scientific leader shifts the balance between research, product development and safety inside a company facing growing competitive pressure.

The shadow of the November crisis

Sutskever’s departure comes six months after the most turbulent episode in OpenAI’s history. The board removed Sam Altman on November 17, 2023, a decision Sutskever took part in as a board member. The news triggered an internal revolt: more than 700 employees threatened to resign if Altman did not return.

Days later, Sutskever publicly said he regretted taking part in the decision to remove Altman and pledged to do everything he could to reunite the company. Altman returned as chief executive on November 22, with a revamped board that no longer included Sutskever.

That crisis exposed a tension that remains essential to understanding OpenAI. The organization was founded with a mission focused on developing advanced AI that would benefit all of humanity, but it now also competes in a market of enormous economic value against Google, Anthropic, Meta and other technology companies. The pace of product launches and the need to manage risks do not always move in lockstep.

Sutskever was particularly associated with the latter. In 2023, he co-led the Superalignment team, created to research how to supervise and control AI systems far more capable than today’s models. The goal was to solve, within four years, safety problems that could emerge with future models. His departure leaves open the question of how much weight that line of research will retain within OpenAI.

A scientific handover at a decisive moment

Jakub Pachocki takes over as chief scientist after leading OpenAI’s research. He has worked on the development of GPT-4 and other company systems, making the appointment a bet on technical continuity rather than an external hire.

Sutskever’s departure does not, by itself, signal a change in strategy or reveal the nature of his next project. But it does remove one of OpenAI’s founders — and a voice with unique authority in the debate over how to build increasingly powerful AI systems — from the company’s front line.

For OpenAI, the immediate challenge will be to show that it can maintain its research capabilities while commercializing its products and addressing the safety demands that come with its most advanced models. For the industry, Sutskever’s next initiative is worth watching: few people have had as much influence on the technology that has brought generative AI to the center of the economy and public debate.