← Multimodal Sources 🕐 2 min read
Multimodal Sources

Mapping the Mind of a Neural Net: Goodfire’s Eric Ho on the Future of Interpretability

Executive summary · March 2026

> So we worked with ARK Institute, like I mentioned a little bit earlier, to understand and interpret Evo 2, which is their kind of DNA foundation model. So it's a sequence-to-sequence model.

Show: Training Data · Publisher: Sequoia Capital · Host: Sonya Huang, Pat Grady

Episode URL: https://pscrb.fm/rss/p/traffic.megaphone.fm/CPUAI9154425844.mp3?updated=1751996708

Publish date: 2025-07-08
Duration: NAs
Default source credibility: HIGH — Sequoia partners interview frontier-lab founders + F500 AI buyers. VC-hosted — portfolio-company framing on recommendations; named guest metrics stay HIGH. Peer-tier to No Priors in quality.

  • Goodfire’s Eric Ho discusses the importance of AI interpretability for understanding and editing neural networks, emphasizing the need for safe and reliable AI deployment in mission-critical contexts.
  • The company’s work on disentangling superposition and mapping neural network concepts could lead to more intentional design of AI models, akin to the human genome project.
  • Eric predicts full decoding of neural nets by 2028, enabling precise control and editing of AI models, with potential applications in genomics and other fields.

Extracted quotes

# Credibility Speaker Org Timestamp Topic Quote
1 HIGH Eric Ho (Founder) Goodfire 26:28 01-ai-native-landscape So we worked with ARK Institute, like I mentioned a little bit earlier, to understand and interpret Evo 2, which is their kind of DNA foundation model. So it’s a sequence-to-sequence model. sequence of nucleotides and it predicts the next nucleotide in a sequence. And our theory is like, this is a narrowly superhuman model. So we really like to work on narrowly superhuman models because it can teach us something about the world that humans don’t really know.
2 HIGH Eric Ho (Founder) Goodfire 46:12 01-ai-native-landscape I think we can do this. And I think we can do this. And hold me to this in 2028, we’re going to figure it all out.

Per-quote detail

1. Eric Ho — Goodfire (26:28)

So we worked with ARK Institute, like I mentioned a little bit earlier, to understand and interpret Evo 2, which is their kind of DNA foundation model. So it’s a sequence-to-sequence model. sequence of nucleotides and it predicts the next nucleotide in a sequence. And our theory is like, this is a narrowly superhuman model. So we really like to work on narrowly superhuman models because it can teach us something about the world that humans don’t really know.

  • Stat: null
  • Credibility: HIGH — Named exec at identifiable org + specific metric with denominator + unscripted interview.
  • Topic tag: 01-ai-native-landscape

2. Eric Ho — Goodfire (46:12)

I think we can do this. And I think we can do this. And hold me to this in 2028, we’re going to figure it all out.

  • Stat: null
  • Credibility: HIGH — Named exec at identifiable org + specific metric with denominator + unscripted interview.
  • Topic tag: 01-ai-native-landscape

Extracted 2026-04-14T19:21:23 via scripts/podcast_mine.py (MLX mlx-community/Qwen2.5-32B-Instruct-4bit).