Goodfire CTO on Concept Manifold Geometry and a $1000/Month ML Research Agent
The Cognitive Revolution · youtube · 2026-08-09
Goodfire co-founder and CTO Dan Balsam delved into the latest advancements in AI interpretability research on a podcast, introducing 'Silico'—a $1,000-per-month internal research platform built for his team.
Key Insights:
- Predictive Data Debugging: Fine-tuning and RL often amplify unwanted behaviors already latent in pre-training data. Interpretability can pinpoint the exact data and features driving these updates.
- Concept Manifold Geometry: Models do not store concepts as simple one-hot features, but as sparse mixtures of meaningful subspaces. This geometry determines the success of interventions like steering, explaining why control can fail off-manifold.
- Silico Platform: An automated agent platform designed specifically for ML research, aiming to assist researchers in deeper model dissection and experimentation.
More from AGI Musings
- AI Labs Criticized for Optimizing Competence Over Wisdom and Prosocial Traits — repligate · 2026-08-09
- Alignment Won't Be 'Solved'—Models Will Just Mature, Says Researcher — repligate · 2026-08-09
- Three Years Later: Are We Close to AI 'Gods' That Code Without Bugs? — flowersslop · 2026-08-09
- e/acc Leader Advises Google to Pivot to Open Source Base Models — beffjezos · 2026-08-09
- Growing Impedance Mismatch Between AI Acceleration and Enterprise Adoption — beffjezos · 2026-08-09
- SF's Bullish Stance on Job Creation Challenged by 'Intelligence from Sand' Argument — zebird0 · 2026-08-09