“It may be that today's large neural networks are slightly conscious.”
“Sequence-to-sequence is all you need.”
“If you have a very powerful AI that is not aligned with humanity, bad things can happen.”
“The goal of the superalignment team is to solve the alignment problem for superintelligent AI within four years.”
“I am proud of and grateful for everything we've done together and I'm confident OpenAI will build safe and beneficial AGI.”
Ilya Sutskever is widely regarded as one of the most brilliant technical minds in modern AI. A student of Geoffrey Hinton, he co-authored the AlexNet paper that ignited the deep learning revolution and went on to serve as Chief Scientist at OpenAI, where he was instrumental in the development of GPT-series models.
Sutskever’s public statements are rare and carefully considered, making each one disproportionately influential. His 2022 tweet suggesting large neural networks may be “slightly conscious” sparked enormous debate and revealed a willingness to consider profound implications that many researchers avoid.
His role in the November 2023 OpenAI board crisis—where he initially voted to fire Sam Altman before expressing regret—became one of the most consequential moments in AI industry history. The episode appeared to reflect genuine internal conflict about the pace and direction of AI development.
Sutskever’s subsequent departure from OpenAI and founding of Safe Superintelligence Inc. in June 2024 signaled a complete pivot from capability research to alignment. SSI’s stated mission—solving superintelligence alignment before building superintelligent systems—represents perhaps the purest expression of the “safety first” philosophy.
His relative silence since founding SSI has only amplified interest in his thinking, with the AI community closely watching for any signals about SSI’s research direction and Sutskever’s evolving views on the timeline to superintelligence.