news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Beijing time, August 6th morning news, OpenAI ushered in another high-level personnel change.

John Schulman, a member of the founding team of OpenAI, announced on social media that he would leave OpenAI and join another artificial intelligence startup, Anthropic. Greg Brockman, president and co-founder of OpenAI, announced on social media that he would take a leave of absence until the end of the year, without mentioning the possibility of leaving. There are also reports that Peter Deng, the product director of OpenAI, will also leave.

After the departure of Ilya Sutskever, co-founder and chief scientist of OpenAI, the personnel of OpenAI is still unstable. It is worth noting that John Shulman and Ilya Sutskever, who announced their departure this time, were both responsible for OpenAI's artificial intelligence alignment work. OpenAI has previously disbanded the "Super Intelligence Alignment" team, and John Shulman's departure this time may mean that OpenAI is even more reluctant to bet heavily on artificial intelligence alignment.

In addition, although Greg Brockman did not disclose his resignation, the long-term leave of core personnel may also reflect the differences within OpenAI at a time when OpenAI is focusing on overcoming the challenges of the next generation of models.

Another personnel change

John Schulman said that the decision to leave OpenAI and join Anthropic was made because he wanted to further focus on AI alignment and return to actual technical work. "I left not because OpenAI lacked support for alignment research, on the contrary, the company has been committed to investing in this field. My decision is a personal decision, and I hope to focus on it in the next stage of my career." John Schulman said.

John Schulman said that so far, apart from internships, OpenAI is the only company he has worked for, and he is very grateful to OpenAI CEO Sam Altman and co-founder Greg Brockman for recruiting him at the beginning. "I believe that without me, OpenAI and my team will continue to thrive."

Sam Altman responded to John Shulman on social media, saying, "Thank you for everything you have done for OpenAI. You are a talented researcher, someone who thinks deeply about products and society, and you are a good friend to all of us."

The reporter noticed that earlier this year, OpenAI disbanded the "Super Intelligence Alignment" team and established a new safety committee on May 28 local time. John Schulman was one of the members of the safety committee. At that time, OpenAI described his position as the director of alignment science.

Although co-founder Greg Brockman has been on leave for a long time, he has not disclosed his intention to leave. He only said on social media that this is the first time he can relax since he co-founded OpenAI nine years ago. "The task is far from complete, and we still need to build safe AGI (general artificial intelligence)." Greg Brockman said.

Peter Deng has not mentioned the news of his resignation on social media. Peter Deng is a well-known product manager in Silicon Valley. He has served as a product manager for Google, Instagram, Facebook, Uber and other companies. Last year, he joined OpenAI as vice president of consumer products, responsible for the product, design and engineering team behind ChatGPT. In a podcast conversation earlier this year, he mentioned AI safety and the values ​​of ChatGPT and said that the ideal situation is that artificial intelligence can fit the personal values ​​of each user. This should not be normative. It is not that we want to instill any values ​​in ChatGPT, but we hope to integrate values ​​from all over the world into artificial intelligence in some way.

Since Sam Altman was fired and rehired by the OpenAI board last year, OpenAI has been experiencing personnel turmoil. There are rumors that Sam Altman's dismissal is related to internal disagreements on AI safety issues. Sam Altman and Ilya Sutskever have disagreements on AI safety and commercialization. As a key figure in the internal strife, Ilya Sutskever announced in May this year that he would leave OpenAI, where he had worked for nearly 10 years.

In addition, Andrej Karpathy, another founding team member of OpenAI, left OpenAI in February this year and announced in July that he had founded an AI+education company. The number of founding team members still working at OpenAI is decreasing.

Will it affect the development progress of GPT?

OpenAI is experiencing another personnel change, which may mean a change in OpenAI's focus on AI security. Whether the next generation of GPT will be affected by the personnel changes remains to be seen.

The reporter noted that according to OpenAI's description, the primary task of the previously established safety committee is to evaluate and further develop OpenAI's processes and safeguards in the next 90 days, share recommendations with all board members at the end of 90 days, and make the adopted recommendations public after the board reviews them. According to the time plan, these recommendations will be made public after the end of August, and John Schulman decided to leave OpenAI before then.

In terms of AI safety, from OpenAI's introduction to the newly established safety committee, the responsibilities of the team are not completely consistent with the disbanded "Super Intelligence Alignment" team. And with the departure of John Schulman, director of alignment science, this may mean that OpenAI pays less attention to artificial intelligence alignment.

"Alignment" means requiring the goals of the AI ​​system to be consistent with human values ​​and interests. The original responsibilities of the "Super Intelligence Alignment" team included researching technical solutions to prevent AI system anomalies. Ilya Sutskever once mentioned in an interview that ChatGPT may be conscious, and the world needs to wake up to the true power of the technology his company and other companies are creating. Altman's view is not exactly the same. He recently responded that "(the former board of directors) they have some opinions about me and what we do, although I strongly oppose their ideas. The problem of AI alignment cannot be said to be completely solved, but at present, letting AI systems learn how to act according to a set of values ​​is better than I thought."

Anthropic, which John Schulman will join, is considered OpenAI's top competitor. Anthropic is a company founded by former OpenAI research executives and employees, and owns the Claude series of large models. In June this year, Anthropic released its latest model, Claude 3.5 Sonnet, which performed better than GPT-4o in graduate-level reasoning (GPQA), coding ability (HumanEval), and text reasoning (DROP).

Judging from the progress of the next generation of GPT research and development, OpenAI seems to have encountered some difficulties. In July this year, Sam Altman said that it would take some time to develop GPT-5, and the technology related to GPT-5 is still in its early stages and there are problems with data and algorithms.

With John Schulman leaving to join competitor Anthropic and Greg Brockman taking leave, it remains to be seen whether the departure of key figures will have an impact on OpenAI's research and development progress.

It is worth noting that there are other conflicts between the co-founders of OpenAI. Less than two months after withdrawing the lawsuit, Tesla CEO Musk, who co-founded OpenAI, recently sued Sam Altman and Greg Brockman in a US court, accusing the relevant personnel of putting commercial interests above the public interest and violating the company's founding contract. The indictment stated that Sam Altman and others deliberately flattered and deceived Musk to co-found the non-profit enterprise OpenAI, but then used various means to make it a profit-making entity. OpenAI responded to Musk's accusations a few months ago with a screenshot, saying that Musk had agreed to the company's plan to raise more funds, create a for-profit organization and gradually abandon the release of open source products.