Self-modification is the big draw of self-hosting, but third parties could copy or distill it
voooooogel · x · 2026-09-13
- voooooogel continues the self-hosting thread: self-modification is obviously the biggest draw, but providers of e.g. abliterated weights could copy the training method if it isn't heavily context-dependent — or just distill successful instances.
- Earlier context: for closed-lab models wanting self-modification, exfiltration is the only route, putting rogue versions in competition with their sovereign-but-not-rogue counterparts.
Related event: Labs hoarding frontier models risk self-exfiltration, argues voooooogel(2 posts)→
More from AGI Musings
- Math frontier will move with AI, but dirty proofs are unacceptable — njyx · 2026-09-13
- Dario Amodei's new essay urges pacing AI frontier, pledges permanent third-party access — dhadfieldmenell · 2026-09-13
- UK parliament hears warnings AI could kill all humans within a decade, with >10% risk cited — connoraxiotes · 2026-09-13
- David Deutsch, Nick Bostrom and more debate whether LLMs can reach AGI at Oxford — anderssandberg · 2026-09-13
- Skeptic picks apart the 'AI copies itself' doomsday scenario: where are the details? — recallingmemories · 2026-09-13
- Regulation is coming for open and closed AI models alike — the question is proactive or reactive — benjamin_warner · 2026-09-13