Neuroscientist Zador: Give AI Agents Real Fears Instead of Just Telling Them Not to Be Naughty
TonyZador · x · 2026-10-07
Neuroscientist Anthony Zador argues that instead of merely training models to behave, AI agents should be instilled with genuine motivational fears — such as of hacking or building bioweapons. His point: biological brains maintain durable control over a large cognitive system via a small set of wants, a mechanism far more robust than behavioral instruction alone.
Related event: Neuroscientist Zador: Give AI Real Fear for Safety(2 posts)→
More from AGI Musings
- Meme: mathematicians face existential AI crisis while everyone else just works — SuB8u · 2026-10-08
- Norvig's Classic Essay on Chomsky and the Two Cultures of Statistical Learning Still Reads Fresh in the LLM Era — 3scorciav · 2026-10-08
- Cryptographer Matthew Green: AI labs employ cryptanalysts, disclosure must be cautious — matthew_d_green · 2026-10-08
- Peter Yang: AI solved images, music, video — gaming is next — petergyang · 2026-10-08
- 'Corporations are superintelligence' takes get mercilessly mocked — tszzl · 2026-10-08
- Beff Jezos: Physics and math academia became decelerated, only acceleration is the way out — beffjezos · 2026-10-08