John Schulman's Departure For Anthropic Is Casting Another Shadow Over OpenAI's Future

Internal turmoil continues at OpenAI, as two key figures of ChatGPT choose to leave the company.

This week, John Schulman, a co-founder of AI giant OpenAI, announced on Twitter that he would be leaving OpenAI to join Anthropic, a main competitor of OpenAI.

Schulman, who joined at the inception of OpenAI nine years ago, was responsible for leading the later training work of ChatGPT and APIs.

He stated: This choice stems from my desire to deepen my focus on AI alignment and to start a new chapter of my career where I can return to hands-on technical work. I've decided to pursue this goal at Anthropic

However, he specifically emphasized that his departure was not due to a lack of support from OpenAI for alignment research, On the contrary, company leaders have been very committed to investing in this area.

 At the same time, according to foreign media reports, Peter Deng, the Vice President of OpenAI and the person in charge of ChatGPT, has also left the company. He has repeatedly stated in public that OpenAI's models are actually more powerful than those released, and they must be introduced cautiously for safety reasons.

Greg Brockman, another co-founder and President of OpenAI, also stated that he would be taking a long vacation until the end of this year.

He said this was the first relaxation since co-founding OpenAI nine years ago, The mission is far from complete; we still have a safe AGI to build.

In response to Schulman's departure, OpenAI CEO Sam Altman left a message expressing gratitude and saying, You have profound thoughts on products and society, and more importantly, We will miss you tremendously and make you proud of this place.

The official ChatGPT account also left a message thanking Schulman for his contributions to the ChatGPT model.

Jan Leike, who previously left OpenAI to join Anthropic's super alignment team, also replied, saying he was very happy to collaborate again.

From the incident in November last year, where Altman was fired, to the departure of core members of the security team this year, OpenAI's safety issues have always been controversial.

Several core members, including Ilya Sutskever, OpenAI's co-founder and chief scientist, and Leike, have left one after another, and the core security team's super alignment has also been dissolved. After Leike announced his resignation, he posted on Twitter, angrily criticizing OpenAI and its leadership for focusing on glamorous products, and no longer valuing the safety culture and process.

At that time, insiders revealed that employees who focus on the safety of AI systems have lost confidence in Altman's leadership.

OpenAI's Chief Technology Officer, Mira Murati, responded in early July, saying: Super alignment is a very important security team at OpenAI, but she emphasized that it is only one of the teams, and there are many people in the company responsible for safety. She also disagreed with the view of prioritizing products over safety, and said this argument is a bit too cynical.

She said that many people believe that safety and capability are separate and must be prioritized one over the other, but in industries with mature safety thinking and systems, such as aerospace and automotive, people do not necessarily always sit around the table to discuss what is safe, because they are always doing it.

She believes that the entire industry should also develop towards mature safety disciplines, not only operational discipline but also product and deployment safety, considering factors such as bias and misinformation. At the same time, she said that OpenAI has also been considering the long-term alignment of models, not just the present.

Nonetheless, with leadership turmoils like this and a potential financial deficit of $5 billion, the road ahead for the once astonishing OpenAI may not be easy to walk.