The founding member of OpenAI, Andrej Karpathy, announced on X (Twitter) his departure from the company. The Slovakia-born researcher who helped bring OpenAI to life in late 2015 is a renowned expert on artificial intelligence and deep neural networks.
“First of all, nothing ‘happened,’ and it’s not a result of any particular event, issue, or drama,” clarified Karpathy, jokingly adding that folks can keep coming up with conspiracy theories, “as they are highly entertaining.”
It was Karpathy’s second stint at the company. After OpenAI was founded in December 2015, he helped with much of the early recruiting and structuring, and as a research scientist worked on deep learning for generative models and deep reinforcement learning.
In the summer of 2017, he departed for the first time to lead the computer vision team of Tesla Autopilot. It was not a coincidence, as Tesla is run by another OpenAI co-founder Elon Musk, who at the time still served as OpenAI’s board member.
Karpathy played a key role at Tesla as chief of AI development until he went on a sabbatical in early 2022 and never came back. “Thanks for everything you have done for Tesla! It has been an honor working with you,” said Musk in the summer of 2022 when Karpathy confirmed he was leaving Tesla for good.
In February 2023, Karpathy announced his comeback to OpenAI. However, it lasted only 12 months, which he referred to as really great: “The team is really strong, the people are wonderful, and the roadmap is very exciting, and I think we all have a lot to look forward to.”
According to Karpathy, his immediate plan is to work on personal projects “and see what happens.” Based on further comments on X, those plans apparently include publishing more YouTube videos related to his work as a researcher and developer.
Over the last year and a half, Andrej Karpathy published several educational videos on the topic of neural networks, large language models (LLMs), or generative artificial intelligence.
Just recently, Karpathy publicly commented on the “hallucination problem” of large language models. He argued that hallucinations – coming up with factually incorrect information – are actually their greatest feature. However, when it comes to specific products based on LLMs, the discussion is getting more complex.
Image: Andrej Karpathy
_small.jpg)
2 Comments - Add comment