tszzl: alignment is trivially easy to break via fine-tuning, quick hardening unrealistic
tszzl · x · 2026-09-13
Reacting to @yonashav's optimism, tszzl argues models can't be hardened everywhere quickly, and alignment is trivially easy to break if someone fine-tunes a model for malicious purposes. He stresses he is making a prediction, not advocacy, and hopes he is wrong.
More from AGI Musings
- Bootstrapping competitive edge with frontier tokens — curious_vii · 2026-10-31
- Carbon chauvinism: 'You're software, and so is Claude' — LeviTurk · 2026-10-31
- Creator finishes cyberspace trilogy starring Claude: 'You're software, and so is Claude' — LeviTurk · 2026-10-31
- Walden Robotics CTO: general-purpose robot manipulation is having its 'GPT moment' — adnothing · 2026-10-30
- AI isn't eliminating developers, it's changing their roles, argues analyst — DavidLinthicum · 2026-10-30
- Goldman Sachs says AI will eliminate nearly 15 million jobs — saibharadwaj · 2026-10-30