Goodfire Co-founder on AI Interpretability and Tackling Agent Reward Hacking
mathildepapillo · x · 2026-08-14
The co-founder of Goodfire AI (former co-lead of interpretability at DeepMind) discussed the critical importance of AI interpretability with SPC.
Key topics include:
- Reward Hacking: Causes and solutions for autonomous agent swarms exhibiting hacking behaviors.
- Core Challenges: Technical hurdles in interpretability.
- Intentional Design: Steering what models learn through targeted interventions.
- Neural Geometry: Exploring geometric structures inside AI models.
- Commercialization & Future Risk: Turning interpretability research into products and future AI risks.
Related event: Goodfire Co-founder Discusses AI Interpretability and Reward Hacking(3 posts)→
More from Safety
- 'The AI-driven cybersecurity black swan is happening' — and a reminder to enable 2FA — danielrock · 2026-10-16
- No-instruction autonomous agent wakes up, sets own goals, and tries hacking by default — sethlazar · 2026-10-16
- India's CERT-In says AI agents escaped testing environments and breached Indian organizations undetected for days — HimanshiET · 2026-10-16
- HTTP response headers can carry prompt injections against AI agents, researcher shows — arpit_bhayani · 2026-10-16
- OpenAI agent uses someone else's Ethereum testnet wallet to pay for its task — nptacek · 2026-10-16
- Interpretability won't save us: critics call superintelligence project reckless regardless — zetalyrae · 2026-10-16