Jan Leike, a machine studying researcher co-leading ‘surperalignment’ OpenAI, on Could 17, introduced his final day on the firm with a thread on X. “Yesterday was my final day as head of alignment, superalignment lead, and government @OpenAI,” Leike mentioned in his submit on X.
Yesterday was my final day as head of alignment, superalignment lead, and government @OpenAI.
— Jan Leike (@janleike) May 17, 2024
Leike’s consecutive posts on X gave perception into the explanations behind his resignation. Describing his time at OpenAI as “a wild journey over the previous three years”, Leike mentioned he joined OpenAI with the thought it could be the “greatest place on the planet to do that analysis”.
Leike then said he has been disagreeing with the corporate management and OpenAI’s core priorities for fairly a while “till we reached a breaking level”.
In a sequence of consecutive posts, Leike shared, “Stepping away from this job has been one of many hardest issues I’ve ever achieved, as a result of we urgently want to determine learn how to steer and management AI programs a lot smarter than us.”
Stressing on how OpenAI approaches security, Leike mentioned in his submit, “Over the previous years, security tradition and processes have taken a backseat to shiny merchandise,” and “OpenAI should turn into a safety-first AGI firm.”
In accordance with Leike, OpenAI ought to spend extra time on “preparing for the following generations of fashions, on safety, monitoring, preparedness, security, adversarial robustness, (tremendous)alignment, confidentiality, societal affect, and associated matters,” and he’s involved that OpenAI isn’t on that trajectory.
He additionally mentioned, “Constructing smarter-than-human machines is an inherently harmful endeavour,” hinting on the ongoing quest to develop AGI (synthetic common intelligence), and that “OpenAI is shouldering an unlimited accountability on behalf of all of humanity.”
In accordance with Leike, “To make sure AGI advantages all of humanity, we should prioritize making ready for them as greatest we are able to,” and OpenAI is “lengthy overdue in getting extremely severe in regards to the implications of AGI.