More Inside Information Revealed After OpenAI's "Superalignment" Team Disbandment, Concerning Computing Power and AGI Safety

PANews reported on May 20 that, according to Financial World, on May 14 Eastern Time, OpenAI's Chief Scientist Ilya Sutskever announced his resignation. On the same day, Jan Leike, one of the heads of OpenAI's Superalignment team, also announced his departure. This Friday, OpenAI confirmed that the "Superintelligence Alignment Team" co-led by Sutskever and Leike has been disbanded. In the early hours of May 18, Jan Leike posted 13 tweets on social platform X, revealing the real reasons for his resignation and more insider information. In summary, the main issues were insufficient computing resources and a lack of emphasis on safety at OpenAI. Leike stated that more resources and effort should be dedicated to preparing for the next generation of AI models, but the current development path cannot achieve this goal smoothly. His team faced significant challenges over the past few months, sometimes struggling to obtain enough computing resources. Leike also emphasized that creating machines that surpass human intelligence is fraught with risks, and OpenAI is taking on this responsibility, but safety culture and processes have been marginalized in the pursuit of product development. OpenAI must transform into an AGI company that prioritizes safety. In response to Leike's revelations, Altman urgently responded on May 18: "I am very grateful for Jan Leike's contributions to OpenAI's AI superalignment research and safety culture, and I am very sorry to see him leave the company. He pointed out that we still have a lot of work to do, and we agree with this and are committed to advancing this work. In the coming days, I will write a more detailed article to discuss this issue." Tesla CEO Elon Musk commented on the disbandment of the "Superalignment" team, saying: "This shows that safety is not OpenAI's top priority."

Additionally, on May 18, the American tech blog TechCrunch reported that OpenAI abandoned safety research in favor of launching new products like GPT-4o, ultimately leading to the resignation of the two heads of the "Superalignment" team. It is still unclear when or if the tech industry will achieve the necessary breakthroughs to create AI capable of performing any task that humans can. However, the disbandment of the "Superalignment" team seems to confirm one thing: OpenAI's leadership—especially Altman—has chosen to prioritize products over safety measures. Over the past year, OpenAI has filled its chatbot store with spam and violated platform service terms by scraping data from YouTube... Safety seems to be a secondary concern at the company—an increasing number of safety researchers have reached this conclusion and chosen to seek opportunities elsewhere.

On May 19, OpenAI CEO Sam Altman responded on social platform X to previous rumors about "gag" resignation clauses: the company has never reclaimed anyone's vested equity, even if someone did not sign a resignation agreement (or disagreed with a non-disparagement agreement), it would not take such actions.