Revelation from OpenAI’s Superalignment Employee on Resignation Due to Lack of AGI Mission Focus by Company

In July last year, OpenAI announced the formation of a new research team called Superalignment to prepare for handling superintelligent AI that surpasses human intelligence, ensuring humans remain knowledgeable and in control to prevent any harmful behavior towards themselves.

Headed by Ilya Sutskever, a co-founder of OpenAI and the head of the scientific team, Superalignment, along with Jan Leike, agreed to allocate 20% of the company’s processing power to this team.

Last week, OpenAI unveiled its new AI model, GPT-4o, along with expanding access to various features for users for free, reflecting the decreasing cost of AI processing. However, shortly after, Ilya Sutskever announced his resignation from the company, followed by Jan Leike and several other employees from the Superalignment team. This event was certainly not ordinary.

Wired reported on the current situation within OpenAI, stating that the Superalignment department had been disbanded along with the resignation of key employees, with the team’s responsibilities transferred to other research divisions within OpenAI.

Jan Leike explained the reason for his resignation in X, stating his disagreement with the direction of OpenAI’s management and the organization’s urgent mission many times, reaching a breaking point where he could no longer tolerate.

He believed that OpenAI’s main mission should be to prepare for safety concerns, societal impact, uncontrollable intelligence, and other related issues when developing new AI models, which are difficult tasks. He expressed concerns that OpenAI was not moving in that direction any longer, likening the challenges his team faced to navigating a ship through a stormy sea, with securing agreed-upon processing power becoming increasingly challenging.

Leike pointed out that OpenAI was founded with the mission of overseeing responsible and beneficial AI for humanity. However, in recent years, the focus on creating safe AI has been overshadowed, with the company prioritizing introducing commercially appealing products to the market.

Wired confirmed that several employees from the Superalignment team had also resigned, with some citing reasons similar to Jan Leike’s, indicating a lack of confidence that OpenAI was still following its core mission of creating AGI that functions universally.

TLDR: OpenAI announced the Superalignment team to handle superintelligent AI, with key members resigning due to disagreements on the company’s focus on safety and its core mission.