Eric Ho
@ericho_goodfire
Co-founder / CEO @GoodfireAI
Just wrote a piece on why I believe interpretability is AI’s most important frontier - we're building the most powerful technology in history, but still can't reliably engineer or understand our models. With rapidly improving model capabilities, interpretability is more urgent,…
They said it couldn't be done. They said it *shouldn't* be done. We tried anyways. Today we're launching Ash, the first AI designed for therapy, and announcing $93M in funding from @radicalventures @ForerunnerVC @a16z @felicis @trailmixvc @joinsequel and many more. We've had…
New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
i'm excited about AI x materials - so many new technologies will be enabled by accelerating materials discovery. congrats to the Radical team!
We are excited to announce a $55M Series Seed+ led by RTX Ventures, joined by NVentures (@nvidia's VC arm), @noavctech, @infinitexyz_, @eni, @alley_corp and many others.
Sycophancy, Grok, Bing Sydney - we're not great at catching & mediating unexpected model behavior before deployment. But this is strange - shouldn’t the info we need be right there in the weights & activations? Why we think interpretability is AI's most important frontier (1/10)
Just wrote a piece on why I believe interpretability is AI’s most important frontier - we're building the most powerful technology in history, but still can't reliably engineer or understand our models. With rapidly improving model capabilities, interpretability is more urgent,…
As it turns out, the parity feature from our CLT replication generalizes to Katy Perry lyrics
great work! really like studies like this doesn't change my thinking much that AI is already a huge speed up, although the difference in dev perception is interesting only 44% had used cursor before - how the hell did they find people who hadn't used cursor? these are not AI…
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
back in college i helped with research in a social robotics lab, where personal robot tutors were shown to help with child education i think this can be really impactful and is definitely a step in the right direction. it's also very cute and i want one!
Thrilled to finally share what we've been working on for months at @huggingface 🤝@pollenrobotics Our first robot: Reachy Mini A dream come true: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community. Tiny price, small size, huge…
It was a ton of fun chatting with @sonyatweetybird and @roelofbotha about interpretability!
Can we map the mind of an LLM? Our first mechanistic interpretability episode on Training Data featuring @GoodfireAI founder @ericho_goodfire (and our first cameo from @roelofbotha!) Goodfire is building an independent mech interp lab, led by some heavyweight researchers from…