# Chris Olah

Co-founders · Interpretability lead at Anthropic · joined 2021

Left the University of Toronto after one year, without finishing a degree, and became one of the founders of interpretability research.

Full profile: https://www.autumn.ai/resources/anthropic/people/chris-olah

**At Anthropic:**
- Mar 2021 – present: Cofounder, Interpretability Research Lead
Sources: [1](https://www.linkedin.com/in/christopher-olah-b574414a)

**Career before Anthropic:**
- Sep 2018 – Dec 2020: OpenAI, Member of Technical Staff
- Nov 2016 – Sep 2018: Google Brain, Research Scientist
- Oct 2015 – Nov 2016: Google Brain, Research Associate
- Jun 2011 – Sep 2011: Xelerance, Developer
- Jul 2010 – Nov 2010: Environment Canada, Developer, Adaptations and Impacts Research Division
Sources: [1](https://www.linkedin.com/in/christopher-olah-b574414a)

**Education:**
- University of Toronto (attended, studied mathematics for one year before dropping out, no degree)
Sources: [1](https://en.wikipedia.org/wiki/Chris_Olah)

**Lesser-known facts:**
- Studied mathematics at Toronto for one year and has no degree. [1](https://en.wikipedia.org/wiki/Chris_Olah)
- Received a Thiel Fellowship in 2012. [1](https://en.wikipedia.org/wiki/Chris_Olah)
- Joined Google Brain as an intern in 2014; DeepDream (2015) is among his early work. [1](https://www.linkedin.com/in/christopher-olah-b574414a)
- Co-founded Distill, an interactive machine learning journal, in 2017. [1](https://www.anthropic.com/news/chris-olah-pope-leo-encyclical)
- Published a denial-of-service vulnerability in the Unbound DNS resolver in 2011, logged as CVE-2011-1922. [1](https://colah.github.io/cv.pdf)
- Created the meta-SNE algorithm, which maps the space of neural networks, during a 2014 Google Brain internship hosted by Jeff Dean. [1](https://colah.github.io/cv.pdf)
- Started the journal Distill to reduce what he calls "research debt" by explaining existing results unusually clearly. [1](https://80000hours.org/podcast/episodes/chris-olah-unconventional-career-path/)
- Spoke at the Vatican launch of Pope Leo XIV's first encyclical in May 2026, the only Big Tech representative invited. [1](https://www.reuters.com/world/europe/anthropics-olah-says-ai-must-be-guided-outside-big-tech-2026-05-25/)
- Says his team keeps finding internal states in models that functionally mirror joy, satisfaction, fear, grief and unease. [1](https://www.anthropic.com/news/chris-olah-pope-leo-encyclical)

**Research papers before Anthropic:**
- Concrete Problems in AI Safety (2016), author 2 of 6: https://arxiv.org/abs/1606.06565
- The Building Blocks of Interpretability (2018), author 1 of 7: https://distill.pub/2018/building-blocks/
- Zoom In: An Introduction to Circuits (2020), author 1 of 6: https://distill.pub/2020/circuits/zoom-in/

**Interviews and talks:**
- [80,000 Hours Podcast on interpretability research](https://80000hours.org/podcast/episodes/chris-olah-interpretability-research/)
- [80,000 Hours Podcast on his unconventional career path](https://80000hours.org/podcast/episodes/chris-olah-unconventional-career-path/)

**Known for:**
- The Building Blocks of Interpretability (2018)
- Circuits (2020)
- Multimodal neurons in CLIP (2021)
- Activation atlases (2019)
- DeepDream (2015)
Sources: [1](https://en.wikipedia.org/wiki/Chris_Olah)

**Forbes estimate:** $15.5B as of 2026-10-18 (an estimate, not a disclosed holding). [1](https://www.forbes.com/profile/christopher-olah/)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people