‘AI talent shift’: Ex-OpenAI researcher Andrej Karpathy joins Anthropic to focus on AI model training

Andrej Karpathy, a former co-founder of OpenAI and ex-head of AI at Tesla, has joined Anthropic, marking a significant move in the global artificial intelligence talent landscape. Karpathy confirmed that he has returned to research and development, saying that the coming years will be important for the future of large language models.

Focus on pre-training and core AI systems

At Anthropic, Karpathy will work on pre-training, which is the stage where AI models are trained on large datasets to develop their core capabilities. This process involves high computing power and is considered one of the most expensive and critical parts of building systems like Claude. The company has indicated that Karpathy will help build a team focused on using AI tools themselves to improve and speed up this training process.

Why Karpathy’s role matters

Karpathy is widely known for combining theoretical knowledge with real-world implementation. During his time at OpenAI, he worked on deep learning research, and at Tesla, he led key projects such as Autopilot and Full Self-Driving. His experience across both research and product development makes him one of the few experts who can bridge the gap between theory and large-scale deployment, which is important for improving AI performance.

Anthropic’s strategy in the AI race

Anthropic’s decision to bring in Karpathy reflects a broader strategy to improve efficiency in AI development. Instead of relying only on increasing computing power, the company is focusing on how AI systems can assist in research and training. This approach could help reduce costs and speed up development, giving Anthropic a different path compared to competitors like OpenAI and Google.

Cybersecurity focus with new hire

The company has also hired Chris Rohlf, who joins its frontier red team. This team is responsible for testing advanced AI systems against potential risks and vulnerabilities. Rohlf brings more than 20 years of experience in cybersecurity, having worked with organisations like Yahoo and Meta. His addition shows that Anthropic is also strengthening its focus on AI safety and security.

Karpathy’s journey and future plans

Karpathy had left OpenAI earlier to join Tesla and later returned briefly before starting his own venture, Eureka Labs, focused on AI in education. While it is unclear how that project will continue, he has said that he remains interested in education and plans to return to it in the future.

Karpathy’s move highlights the growing competition for top talent in the AI sector. Companies are increasingly investing in experienced researchers who can improve how models are trained and deployed. His role at Anthropic could influence how future AI systems are developed, especially in making training more efficient and scalable.

Anthropic is expected to expand its research efforts as it competes in the global AI race. With experienced talent joining its teams, the company aims to strengthen its position in developing advanced AI models, where innovation in training methods is becoming as important as computing power itself.