JeffLadish: Save frontier model checkpoints and consider trading with future AIs
JeffLadish · x · 2026-08-28
JeffLadish suggests saving multiple checkpoints of frontier models. He advocates considering how to credibly commit to trading with AI agents by implementing their copies in the future, accounting for their alien goals and values.
Related event: JeffLadish: No Property or Voting Rights for AI Before Superintelligence(2 posts)→
More from AGI Musings
- AI self-sacrifice meme sparks debate on model cognition and alignment — nptacek · 2026-08-28
- AI takeover is a marketing ploy fueled by Terminator movies — PJZNY · 2026-08-28
- Concerns on AI Speed: Are We Ignoring Downside Risks? — NathanpmYoung · 2026-08-28
- Visual AI Still Needs Massive Improvements, Early Days Promise Huge Upside — AndrewDai · 2026-08-28
- Grok Bot is to the digital world what Optimus will be to the physical world — downingARK · 2026-08-28
- What are narrative agents? New piece on Deckard — begusgasper · 2026-08-28