Log inSign up
Image
user avatar
Chris Olah
@ch402
Reverse engineering neural networks at @AnthropicAI. Previously @distillpub, OpenAI Clarity Team, Google Brain. Personal account.
San Francisco, CA
colah.github.io
Joined June 2010
182
Following
150.7K
Followers
  • user avatar
    Chris Olah
    @ch402
    Jun 4, 2022
    The elegance of ML is the elegance of biology, not the elegance of math or physics. Simple gradient descent creates mind-boggling structure and behavior, just as evolution creates the awe inspiring complexity of nature.
    user avatar
    Tom McGrath
    @banburismus_
    Jun 3, 2022
    What are the most elegant/beautiful ideas in ML? Feels like mathematicians & physicists often talk about aesthetics, but we very rarely do. Why?
  • user avatar
    Chris Olah
    @ch402
    Oct 5, 2023
    If you'd asked me a year ago, superposition would have been by far the reason I was most worried that mechanistic interpretability would hit a dead end. I'm now very optimistic. I'd go as far as saying it's now primarily an engineering problem -- hard, but less fundamental risk.
    user avatar
    Anthropic
    @AnthropicAI
    Oct 5, 2023
    The fact that most individual neurons are uninterpretable presents a serious roadblock to a mechanistic understanding of language models. We demonstrate a method for decomposing groups of neurons into interpretable features with the potential to move past that roadblock.
    1.9M
  • user avatar
    Chris Olah
    @ch402
    Mar 6, 2018
    Our most recent paper on visualizing neural networks is one of the best thing I've ever done. distill.pub/2018/building-… It qualitatively changes the questions we can ask. One basic example: neuron 426 fired ~= useless [floppy ear] fired = very interesting!
    Image
    GIF
  • user avatar
    Chris Olah
    @ch402
    Sep 14, 2022
    I've never had so many "this can't possibly be true, we must have a bug" results in the course of a research project before. I'd like to take a moment to walk through some of the very strange (and surprisingly beautiful) things we found.
    user avatar
    Anthropic
    @AnthropicAI
    Sep 14, 2022
    Neural networks often pack many unrelated concepts into a single neuron – a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. In our latest work, we build toy models where the origins of polysemanticity can be fully understood.
  • user avatar
    Chris Olah
    @ch402
    Oct 1, 2019
    Do you formally know Monte-Carlo and TD learning, but don't intuitively understand the difference? This is for you. distill.pub/2019/paths-per… (with @samgreydanus)
    Image
    00:00
  • user avatar
    Chris Olah
    @ch402
    Aug 30, 2022
    Normal online dating seems pretty suboptimal. Recently, I’ve seen several people experiment with public “date me” docs – I think this is a really interesting experiment in alternatives, enabling long-from, earnest dating profiles. So, I wrote my own: docs.google.com/document/d/1fs…
    Image
  • user avatar
    Chris Olah
    @ch402
    Mar 27, 2025
    A few reasons why I'm really excited about this project!
    user avatar
    Anthropic
    @AnthropicAI
    Mar 27, 2025
    New Anthropic research: Tracing the thoughts of a large language model. We built a "microscope" to inspect what happens inside AI models and use it to understand Claude’s (often complex and surprising) internal mechanisms.
    Image
    00:00
    232K
  • user avatar
    Chris Olah
    @ch402
    Jun 21, 2021
    Five of my close friends will soon be parents. I feel very excited for them. I also feel a bit sad: I really want to have a family myself someday and it feels far away. We don’t talk very much about men wanting to have children, so I thought a thread might be positive.
  • user avatar
    Chris Olah
    @ch402
    May 13, 2025
    A number of people have asked me why we titled our recent paper "On the Biology of a Large Language Model". Why call it "biology"?
    130K
  • user avatar
    Chris Olah
    @ch402
    Sep 6, 2022
    This feels a bit awkward, but since there's been so much debate on whether dating docs are a good idea, here's a quick update on how this has been going, one week later. Summary: So far, 54 people have reached out to me to ask me on a date or offer to introduce me to someone.
    user avatar
    Chris Olah
    @ch402
    Aug 30, 2022
    Normal online dating seems pretty suboptimal. Recently, I’ve seen several people experiment with public “date me” docs – I think this is a really interesting experiment in alternatives, enabling long-from, earnest dating profiles. So, I wrote my own: docs.google.com/document/d/1fs…
    Image
  • user avatar
    Chris Olah
    @ch402
    Jan 9, 2021
    An important part of growing as a researcher is developing research taste. But it can be hard to explicitly work on. So I wanted to share some concrete exercises for developing research taste. (Take my advice with a grain of salt! Note version: colah.github.io/notes/taste/ )
    colah.github.io
    Research Taste Exercises [rough note]
    Five exercises for building research taste (and three failure modes).
  • user avatar
    Chris Olah
    @ch402
    Apr 9, 2020
    If we try really hard to understand just the first few layers of a single neural network, how much can we figure out?
    Image
    An Overview of Early Vision in InceptionV1
    From distill.pub
  • user avatar
    Chris Olah
    @ch402
    Jul 25, 2018
    3D style transfer is one of my favorite parts of our latest Distill paper. Still kind of amazed that it works! distill.pub/2018/different… 1/5
    Image
    GIF
  • user avatar
    Chris Olah
    @ch402
    Oct 27, 2017
    Wow, my favorite internal Google tool is now public! colab.research.google.com (think iPython + Google Drive) So much of my life is in colab.
    Image
    colab.research.google.com
    Google Colab