5 Revelations from Jan Leike's Resignation: OpenAI's Internal Struggles

Freshly Resigned Superalignment Head Criticizes OpenAI: Chasing Flashy Products, Neglecting AGI Safety

From left to right: Andrej Karpathy, Jan Leike and Ilya Sutskever

OpenAI Loses Two More Leaders—What Does That Mean for AI Safety?

Ilya Sutskever and Jan Leike left the company on May 15th, hours after the GPT-4o launch. Andrej Karpathy departed in February. All are focused on responsible AI development.

Jan Leike Speaks Out After Resignation

Jan Leike, the former Head of Superalignment at OpenAI, spoke publicly after resigning.

To understand why Altman was fired last November and why Ilya and Jan left recently, Jan's statement is key.

Reason for Resignation

Jan Leike had long disagreed with OpenAI's leadership on the company's core priorities. This disagreement reached a tipping point, making his resignation one of the hardest decisions of his life.

OpenAI Priorities

Leike believes OpenAI must figure out how to control AI systems smarter than humans. He sees it as a heavy responsibility for humanity.

We need to work on new models. We need to make them safe. We need to watch them. We need to stop bad things. We need to make them strong. We need to align them. We need to keep things secret. We need to think about the impact on people. We need to work on related topics.

A Hopeful Message

Though he resigned, Leike hopes his colleagues at OpenAI will transform the company culture. He urges them to take their work seriously and understand the implications of general AI. The world relies on them, and so does he.

My Opinion

I believe Ilya and Jan resigned for the same reason: lack of resources. Their task was to research AGI risks.

With ChatGPT's success, much of the company's resources shifted to supporting this product and its growing user base. This shift diverted focus from principles and values aimed at safer and more responsible development.

Maybe Sam Altman wants to make money quickly. After all, without money, there is nothing. Or perhaps GPT-5 is truly nearing AGI. Time will tell.

Hi! My name is Meng. Thank you for reading and engaging with this piece. If you’re new here, make sure to follow me. (Tip: I’ll engage with, and support, your work. 🤫thank me later).

Enjoy