Here is the English translation of the text, with the specified terms retained as is:
OpenAI has lost another top talent!
According to The Information, Alec Radford, the lead author of the original GPT paper, is also leaving OpenAI to pursue independent research.
It is understood that Alec joined OpenAI in 2016, and his name appears in the papers from the first GPT to GPT-4o, with him being the first author for the first two generations.
Additionally, Alec is the lead author of OpenAI's multimodal model CLIP, as well as the speech recognition model Whisper.
Regarding Alec's departure, OpenAI's research lead Mark Chen stated, "We deeply respect and appreciate Alec and his contributions, and look forward to continuing to collaborate with him as he explores independent research."
Previously, former research team lead Jeff Clune (who left OpenAI in 2022 and is now an associate professor at the University of British Columbia, also serving as an advisor to DeepMind) had highly praised Alec, considering him to be one of the most influential but least known AI scientists, and that Alec should be recognized as the "father of modern generative AI."
A bachelor's degree holder who persisted in large models, the lead author of the first two generations of GPT
In 2016, Alec Radford obtained his bachelor's degree from the Franklin W. Olin College of Engineering, and then joined OpenAI, where he has worked for 8 years.
Although Alec only has a bachelor's degree, he is an undisputed leader in the field of large models, with his papers having been cited over 190,000 times to date.
The top three most cited papers are related to GPT-3, CLIP, and GPT-2 respectively.
These three papers laid important foundations for today's ChatGPT, and in fact, Alec has been involved in almost all of OpenAI's major breakthroughs.
In addition to the three most cited papers, Alec's name appears in the author lists from the first GPT to GPT-4o, and he was the first author for the first two generations.
Alec was also the lead author of CLIP, which is an important achievement of OpenAI in the multimodal direction and has had a significant impact on the entire industry. He also participated in the development of OpenAI's initial drawing model DALL·E.
In other modalities, Alec is the lead author of OpenAI's open-source speech recognition system Whisper.
However, Alec's name is no longer in the list of contributors for the latest o1 model, although Ilya, who has been involved in all the GPT work with him since the beginning, is still on the list.
Nonetheless, across the entire OpenAI, while others have turned to other fields such as robotics or game AI to "publish some papers," only Alec has persisted in the research of large language models.
It is his persistence in the large language model direction that has changed the development path of OpenAI and the entire AI field.
All authors of the first two generations of GPT papers have now left OpenAI
After Alec's departure, it marks the exit of all authors of the first two generations of GPT papers from OpenAI.
The most famous among them is the former Chief Scientist Ilya Sutskever, who, like Alec, was involved in the work on GPT, including the first two generations.
Ilya's announcement of his departure from OpenAI in May this year shook the entire AI community, and he has since founded a new company focused on building "safe superintelligence."
The other two authors of the first GPT paper are Karthik Narasimhan and Tim Salimans.
Karthik Narasimhan joined OpenAI as a visiting scholar, and after his visit ended in 2018, he took a faculty position at Princeton, where he is now the co-director of the Princeton NLP group, working with Chen Danqi.
Karthik is also one of the authors of the open-source Devin - the thesis advisor of Yao Shunyu, a Tsinghua Yao Class prodigy, during his PhD at Princeton (Yao Shunyu joined OpenAI in August this year).
The third author, Tim Salimans, joined Google Brain in July 2018 as a Senior Research Scientist.
Similarly, GPT-2's third author Rewon Child and fourth author David Luan also chose to join Google.
Rewon later participated in the work on GPT-3, leaving OpenAI for Google in 2021, and then joined the startup Inflection AI in 2022, before moving to Microsoft this March.
David joined Google in 2020 and later co-founded a company called Adept with Transformer paper authors Ashish Vaswani and Niki Parmar, where he served as CEO. However, according to his LinkedIn profile, he is no longer with Adept and has joined Amazon to work with reinforcement learning expert and Andrew Ng's first PhD student, Pieter Abbeel.
In addition to those who joined Google, GPT-2's author also includes Dario Amodei, the co-founder of Anthropic, the company behind the AI assistant Claude. Dario was involved in the work on GPT-2 and 3 at OpenAI, before leaving with his sister Daniela at the end of 2020 to co-found Anthropic, where they serve as CEO and President respectively.
Those who are heading to Anthropic, besides founder Dario, also include Jeffrey Wu, the second author of GPT-2, who left in July this year, becoming the second-to-last participant of the first two generations of GPT to leave OpenAI.
And today, with the news of Alec's departure, the eight researchers who created the earliest two generations of GPT are now all going their separate ways.
Reference links:
[1]https://www.theinformation.com/briefings/senior-openai-researcher-radford-departs
[2]https://scholar.google.com/citations?user=dOad5HoAAAAJ&hl=en
[3]https://x.com/iScienceLuvr/status/1869852854728700166
This article is from the WeChat public account "Quantum", written by Keleisi, and published with authorization from 36Kr.
