Will an open-source model with dangerous capabilities differentially unlearned be released by EOY 2027?
AIOne-Off1jManifold MarketsGeen KYCGeverifieerde afwikkelingsgegevensRedelijk gekalibreerdHuidige gemeenschapsvoorspellingNee 50.5%
Voorspellers
3
Vraagtype
binary
Methodologie
Play-money forecasting platform
Brontype
Voorspelling
Marktdata
Bijgewerkt 1 uur geleden
Vertraagd12 aug 26, 15:59 UTC31 dec 27, 23:59 UTC
Trends
Gerelateerde Markten
Gerelateerde vragen voor context, geen vergelijking van gelijkwaardige contracten.
Will a Mythos-tier-at-cybersecurity open-source model be released before end of 2027?
Manifold MarketsAI: Will someone open-source a $1B model by 2030?
Manifold MarketsIf not passed, will there be any open source model that would have been covered by SB 1047, by January 1st 2027?
Manifold MarketsAI: Will someone open-source a $100M model by 2030?
Manifold MarketsOpen source token share Oct 19–25 (OpenRouter)
KalshiUS government bans an open source AI model in 2026?
PolymarketRegels
Within a year we might get open source models with the cyber capabilities of Mythos.
- One possible way to release such a model without causing a cyber apocalypse would be to differentially reduce cyber capabilities in the released weights of the model (like Anthropic did that one time for Opus)
- For example they could do this and then withhold the cyber-capable modules from the public https://alignment.anthropic.com/2026/modular-pretraining/
- Unlearning has to be robust, but that sounds hard to judge, so I'll err on the side of YES and so if it's unclear whether the unlearning is robust then the market resolves YES, but if the unlearning is obviously not robust (someone publicly restores capabilities, company says they literally just did GradDiff, etc) then the market resolves NO.
- Capabilities have to actually be dangerous.