Shulman Continues to Focus on AI Alignment
Regarding Shulman's departure, Altman posted on social media X: "Thank you for everything you've done for OpenAI! You are an outstanding researcher with deep thoughts on products and society, and most importantly, you are a good friend to all of us. We will miss you very much and make you proud of this place.
(In 2015, I first met John at a cafe in Berkeley. He said, "On one hand, it seems ridiculous to talk about AGI now, but on the other hand, I think it's very reasonable, which is why I think it's important to talk about it." Then he listed important parts of OpenAI's initial strategy. This took about 15 minutes, and then we awkwardly chatted for another 45 minutes :))"
Shulman pushed for the so-called "post-training" process within OpenAI, which involves fine-tuning large language models that power products like ChatGPT. Additionally, he took over the remaining members of the safety team focused on preventing AI from causing societal harm - the "superalignment" team. These personnel changes occurred against the backdrop of rapid business development, potential significant losses, and cyclical staff turnover.
For example, OpenAI recently reassigned another safety lead, Alexander Madry, to other positions. Company co-founder Ilya Sutskever, who made significant contributions to the company's major breakthroughs, recently left to start a competing startup. Another co-founder, Andrej Karpathy, left in February and subsequently launched his own educational startup. OpenAI also hired its first Chief Financial Officer and Chief Product Officer, whose additions may have impacted Peter Deng's responsibilities.
OpenAI is currently in a transition period, potentially shifting from a non-profit organization to a for-profit company similar to Anthropic, and is training its next-generation flagship AI model. Additionally, OpenAI recently launched a search engine and is developing an AI aimed at automating computer tasks for customers.
Shulman announced his departure on social platform X. He stated that he wanted to further focus on AI alignment, return to actual technical work, and decided to pursue this goal at Anthropic. He also emphasized that his departure from OpenAI was not due to a lack of support for alignment research.
Here is the full text of Shulman's post on social media X:
Today, I shared the following with my colleagues at OpenAI:
I've made the difficult decision to leave OpenAI. I'm making this choice because I want to focus further on AI alignment and start a new chapter in my career where I can return to doing hands-on technical work. I've decided to pursue this goal at Anthropic, where I believe I can gain new perspectives and do research on the topics I'm most interested in alongside people who are deeply engaged with these subjects.
To be clear, I'm not leaving because of a lack of support for alignment research at OpenAI. On the contrary, the company leadership has been very committed to investing in this area. My decision is a personal one based on how I want to focus my efforts in the next stage of my career.
Almost 9 years ago, I joined OpenAI as part of the founding team right out of grad school. It's been the first and only company I've worked at, aside from internships. It's also been a lot of fun. I'm grateful to Sam and Greg for bringing me on board in the beginning, and to Mira and Bob for their trust in me, which brought enormous opportunities and helped me successfully navigate various challenges. I'm proud of what we've accomplished together at OpenAI: building an extraordinary, unprecedented company with a mission for the public good.
I'm confident that OpenAI and the teams I've been involved with will continue to thrive without me. Post-training is in very good hands with an amazing depth of talent on the bench. I've gotten too much credit for ChatGPT; Barrett has done an incredible job building the team into an unquestionable powerhouse.
I've been very pleased to see the alignment team come together with some promising projects. Under the leadership of Mira, Boaz, and others, I believe the team will be in very capable hands.
I'm incredibly grateful to have had the opportunity to be part of such an important piece of history, and I'm proud of what we've accomplished together. I'll still be rooting for you all, even as I work elsewhere.
Brockman to Continue Leave Until Year-End
Recent unreported leave and departure events, while seemingly unrelated, reflect that OpenAI's leadership has not fully stabilized since CEO Altman was fired and rehired last November. Brockman played a crucial role in the company, translating major discoveries by AI researchers into large-scale AI models and products like ChatGPT. He was also a key supporter of Altman, especially during the latter's efforts to return to the company. Brockman told employees that he plans to return to work after his extended leave.
Brockman responded on social media X: "I plan to take a year-long leave until the end of this year. This is the first time I've had the opportunity to fully relax since co-founding OpenAI nine years ago. However, our grand mission is not yet complete - we still need to focus on developing safe and reliable AGI technology."
However, fans don't seem to buy Brockman's explanation. Some mentioned the well-known AI scientist and OpenAI co-founder Andrej Karpathy, who left after a long vacation, "We know what happens after a long vacation... ahem."
Other netizens compared him to Musk: "Imagine if Musk had told everyone: I'm taking a 4-month vacation. The world would really end."