OpenAI has disbanded its team focused on the long-term risks of artificial intelligence just one year after the team’s announcement. Some team members are being reassigned to other teams within the company, following the departures of team leaders Ilya Sutskever and Jan Leike. Leike expressed concerns about OpenAI’s safety culture and processes taking a backseat to the development of new products, prompting his decision to leave. He emphasized the need for more focus on security, monitoring, preparedness, safety, and societal impact, suggesting that the company needs to prioritize becoming a “safety-first AGI company.”
The dissolution of OpenAI’s Superalignment team, which was announced last year and focused on steering and controlling AI systems smarter than humans, was first reported by Wired. The news of the team’s disbandment came shortly after the departures of Sutskever and Leike, who shared further details about their reasons for leaving the company. Leike stated that he disagreed with OpenAI’s leadership about the company’s core priorities, with a belief that more attention should be given to safety and societal impact. He expressed concerns about the difficulty of obtaining computing resources necessary for crucial research and highlighted the importance of building machines safely, as the endeavor of creating smarter-than-human AI is inherently dangerous.
The recent departures at OpenAI follow a leadership crisis involving co-founder and former CEO Sam Altman, who was ousted by the board in November. Altman’s removal led to resignations or threats of resignations, with employees signing an open letter and investors, including Microsoft, expressing concern. Altman later returned to the company, while several board members were replaced, including Sutskever. Altman praised Sutskever as one of the greatest minds in the field of AI and expressed sadness over his departure. Research director Jakub Pachocki was named as Sutskever’s replacement as chief scientist.
News of Sutskever’s and Leike’s departures, along with the dissolution of the superalignment team, comes as OpenAI launched a new version of its AI model, ChatGPT, with improved capabilities and an updated user interface. The new GPT4 model, GPT-4o, offers enhanced text, video, and audio capabilities, with plans to enable users to video chat with ChatGPT in the future. Technology chief Mira Murati described the update as a significant step forward in improving the ease of use for ChatGPT users. The company aims to expand the use of its chatbot by providing a faster and more capable model to all users, including those on the free platform.
