Editor: Momoko So sleepy
Resignation has become the norm for OpenAI. One of the O1 inference model contributors has just announced his departure to start his own business. At the same time, in order to fill the vacancy of the post-training leader, OpenAI appointed one of the seven leaders of o1 to lead the team.
Just now, another veteran officially announced his resignation from OpenAI.
Luke Metz, one of the contributors to the o1 inference model, posted, "I am leaving OpenAI to end this amazing journey of more than two years".
与此同时,Information爆料称,此前出走的后训练团队负责人Barret Zoph,现有了新的继任者——William (Liam) Fedus。
He is also one of the seven leaders of the O1 model.
It's worth mentioning that both of today's protagonists were previously Google researchers; After coming to OpenAI, he co-participated in the research and development of ChatGPT, GPT-4 and O1.
One of the seven leaders of o1, the person in charge of training after promotion
Fedus接任的职位此前由Barret Zoph担任,后者在两周前与首席技术官Mira Murati和研究主管Bob McGrew同时离职。
Barret Zoph
Notably, Fedus is also listed as one of the contributors to the new o1 inference model, along with dozens of other researchers.
Prior to that, he worked at Google Brain, where he focused on improving the performance of efficient neural networks through Mixtures of Experts.
Fedus received his B.S. in Physics from MIT, B.S. in Physics from the University of Cambridge, M.Sc. in Elementary Particle Physics from UCSD in '16, and Ph.D. in Computer Science from the University of Montreal in 2017.
ChatGPT veteran resignation
Luke Metz, who was a former colleague of Zoph and Fedus in Google's AI research group, is now leaving OpenAI after two years at the company.
Zoph, Fedus, and Metz are all alleged to be major contributors to ChatGPT and have held various post-training roles at the company.
Currently, Zoph plans to start a new company, but it's unclear if Metz or Murati will join.
In his official tweet, he said that he was one of the original team members and participated in the development of an "early research preview" product along with co-creator John Schulman, post-training VP Barret Zoph, Liam Fedus, and many other colleagues.
It eventually evolved into ChatGPT today.
While we were all passionate about the project, no one could have predicted that it would get to where it is today in such a short period of time – with hundreds of millions of users. It's a rare privilege to be able to witness its growth.
I learned a lot in the process. Many thanks to Sam Altman, Greg Brockman, Mira Murati, Ilya Sutskever, and Bob McGrew for giving me this opportunity. OpenAI is really a place like no other.
Now, it's time for something new!
If you look at the introduction page of ChatGPT at the beginning of its release, you can see that Luke Metz is one of the contributors.
Even the O1 model, which has just been launched, has his contribution.
He introduced on his personal homepage, "I am a researcher who is passionate about bringing artificial intelligence technology into the world. In my free time, I enjoy researching robotics, programming languages, electronics, 3D printing, and graphics processing.
Prior to that, Luke was a research scientist at Google Brain.
He earned a bachelor's degree in engineering from the Franklin · Orin School of Engineering in 2015.
One of the most cited papers in all of his Google Scholar studies was an unsupervised learning representation of CNNs published as an undergraduate graduate.
Resources:
https://www.theinformation.com/briefings/openai-names-new-head-of-post-training-while-another-researcher-departs?rc=epv9gi