In This Story
Less than two weeks after leaving OpenAI, one of the company’s former artificial intelligence safety leads announced that he’s landed at one of its rivals.
Jan Leike, who co-led OpenAI’s now-disbanded “Superalignment team,” which was responsible for AI’s existential dangers, resigned from the company earlier this month. On Tuesday, Leike said on X he was “excited” to join the AI company Anthropic “to continue the superalignment mission.” He added that his new team “will work on scalable oversight, weak-to-strong generalization, and automated alignment research.”
Anthropic, which recently secured a $4 billion investment from Amazon, said its latest AI model family released in March, Claude 3, outperformed OpenAI’s GPT-4.
The company was founded by former OpenAI employees Daniela and Dario Amodei in 2021 to develop “safer” chatbots. The Amodei siblings took a stab at OpenAI earlier this month, saying at Bloomberg’s Technology Summit that its Claude 3 Opus model “is the most capable and powerful AI model available anywhere in the world.”
Leike announced his departure from OpenAI after his co-lead, Ilya Sutskever, who also co-founded the company and served as its chief scientist, said he was leaving. Leike called his decision to leave “one of the hardest things I have ever done, because we urgently need to figure out how to steer and control AI systems much smarter than us.”
Leike wrote that he joined OpenAI because he thought the company “would be the best place in the world to do this research,” but that he had “been disagreeing with OpenAI leadership about the company’s core priorities for quite some time, until we finally reached a breaking point.” He continued by writing that OpenAI should prioritize safety as it pursues artificial general intelligence, or AGI.
OpenAI reportedly did not keep its commitments to the Superalignment team that it announced last July. The team’s requests for access to GPUs, or graphics processing units, were repeatedly turned down, and it never came close to receiving its promised computing power budget of 20%, Fortune reported, citing sources familiar with the matter.