Researcher

Chris Olah

Chris Olah is an Anthropic co-founder and researcher who studies how neural networks represent information and produce behavior. His path began with independent projects and a Thiel Fellowship in 2012, followed by self-directed machine learning with help from mentors such as Michael Nielsen. At Google Brain, he worked with colleagues on ways to inspect visual networks, including DeepDream. He later formed OpenAI’s Clarity team around interpretability research. After leaving with colleagues, he helped establish Anthropic, where model building and interpretation developed together. The team’s 2024 Golden Gate Bridge experiment showed how changing a detected internal feature could change Claude’s answers. Later work studies interactions between features and internal computation. These results explain parts of a model, with much of its behavior still to investigate.

• Interpretability
• Neural networks
• AI research