Leike: OpenAI's loss, Anthropic's gain?

Jan Leike, who previously co-led OpenAI’s “Superalignement” AI safety team–which focused on long-term AI safety risks–abrubtly left OpenAI over concerns that they were prioritizing the release of “shiny new products” over AI safety.

He’s now joined Anthropic, creator of Claude, a Large Language Model (LLM) that rivals OpenAI’s GPT.

What will Leike do at Anthropic?  

It’s thought that Leike will be able to continue the work he started at OpenAI, which was to make sure AI systems are aligned with human values. In a statement he released on X, he expressed enthusiasm for being able to “work on scalable oversight, weak-to-strong generalization, and automated alignment research,” which will include techniques to control the behavior of large-scale AI.

Why has Anthropic hired Leike?

Anthropic was co-founded by former OpenAI employees who left, as Leike did, over safety concerns.

Take Anthropic’s CEO, Dario Amodei–ex-VP of Research at OpenAI–for instance. He left after a public disagreement with OpenAI about its strategic direction, following its $1B deal with Microsoft: He claimed that its mission had changed from focusing on developing safer, more ethical AI, to achieving commercial success.

Leike and Anthropic seem to be aligned on the importance of AI safety which can only be a good thing.