Andrej Karpathy, Co-Founder of OpenAI, Joins Anthropic’s Pre-Training Team
Andrej Karpathy, a distinguished AI expert known for his roles at OpenAI and spearheading AI initiatives at Tesla, has officially become a part of Anthropic.
“I’ve joined Anthropic,” Karpathy announced on X this Tuesday. “The upcoming years at the cutting edge of LLMs promise to be incredibly transformative. I’m thrilled to be part of the team and to return to R&D.”
This week, Karpathy embarked on his role at Anthropic, concentrating on pre-training under the guidance of team leader Nick Joseph. This stage involves comprehensive training runs aimed at equipping Claude with essential knowledge and capabilities, as the company has stated. It is recognized as one of the most resource-demanding phases in the development of a leading model.
An Anthropic spokesperson revealed to TechCrunch that Karpathy will head a team aimed at leveraging Claude to propel pre-training research forward.
Karpathy is notable for being one of the few researchers adept at merging LLM theory with practical large-scale training. His appointment to establish this team highlights Anthropic’s conviction that AI-driven research, alongside computational power, is crucial for maintaining competitiveness with OpenAI and Google.
While at OpenAI, Karpathy made significant contributions to deep learning and computer vision until he departed in 2017 to join Tesla. He oversaw Tesla’s Full Self-Driving (FSD) and Autopilot programs until his exit in 2022.
He briefly returned to OpenAI for a year before launching Eureka Labs in 2024, a startup focused on deploying AI assistants in education.
Karpathy has not shared many updates about Eureka Labs since its inception, leaving it uncertain whether he will continue with the startup. He has also taught an online course titled Neural Networks: Zero to Hero, helping students build neural networks from scratch, and runs a YouTube channel where he frequently shares lectures on LLMs and AI.
“I remain deeply passionate about education and plan to resume my work on it in due course,” Karpathy noted.
TechCrunch has reached out to Karpathy for additional insights.
In related news, Anthropic has welcomed Chris Rohlf to its frontier red team, which evaluates advanced AI models against significant threats. Rohlf brings over two decades of cybersecurity experience, previously contributing to Yahoo’s esteemed cybersecurity team known as “The Paranoids,” and spent the last six years at Meta before joining Anthropic. He was also a fellow at Georgetown’s Center for Security and Emerging Technology, where he was involved in the CyberAI project.
“We have a genuine opportunity before us to significantly enhance cybersecurity with AI,” Rohlf expressed in a post on X. “I can’t imagine a better company or team to join at this critical juncture.”