Posts
Comments
These frontier models could still be vulnerable to stealth (e.g. “sleeper agent”) attacks, specialist models, and stealth attacks by specialist models. The balance depends on the ability gap – if the top model is way ahead of others, then maybe defence dominates attack efforts. But a big ability gap does not seem to be playing out, instead there are several frontier models near-frontier, and lots of (more or less) open source stuff not far behind.
Seems like a stark case of contrast between Bayesianism and the way a frequentist might approach things. I.e. do not reject the null hypothesis of no significant probability until convinced by evidence, either formal arguments or by seeing real-life mishaps. Labeling something as having P(x)~0 probably helps to compartmentalize things, focus to other tasks at hand. But can lead to huge risks being neglected, like in this case of AI Alignment.
Edit: "premortem" seems like a useful exercise to align mind & gut