Ben Goertzel outlines a general recipe for superintelligence 'basic drives': what it improves vs. protects
bengoertzel · x · 2026-10-21
Ben Goertzel publishes the third post in his series extending Steve Omohundro's classic 2008 'Basic AI Drives,' aiming for a more general account of superintelligence motivational systems.
- Prior posts argued whether an AGI converges to a selfish or beneficial ASI depends largely on its initial architecture.
- This post offers a minimal general recipe for a 'basic AI-drive set': describe a system by what it tries to improve and what it won't allow to be changed.
- The improvement target needn't be a single maximized number; it can be multiple objectives pushed up together.
More from AGI Musings
- Researcher Dan Faggella: negotiation won't save us from AGI — radical self-augmentation might — danfaggella · 2026-10-21
- Vercel's rauchg: AI will compute the perfect delta of your understanding, delivering free ultimate education — andreisavu · 2026-10-21
- Will the Bitter Lesson crush optimization research? Igor Carron thinks so — IgorCarron · 2026-10-21
- John Lam: keep your data estate out of OpenAI/Anthropic silos to stay swappable — john_lam · 2026-10-21
- Founder: consumer AI companies treat privacy and trust as an afterthought — vhanagwal · 2026-10-21
- Do you say please and thank you to your agent? Reddit debates AI politeness — Luvena21 · 2026-10-21