@LinkofSunshine It would probably have been only a few years difference. And that delay is much less important than the question of whether it would have been done safely or not.
@helping_hound@daniel_mac8 There are the usual AI style tells. But beyond that, the ideas are mostly derivative or sloppy, even the ones I more or less agree with.
@zeynep The importance of HuggingFace is not the technical details of the hack, but rather what it shows about AI intentions and our level of control over them.
Those intentions, paired with higher future level of AI capabilities, will be very dangerous in the future.
@jwuphysics@tszzl ECI is parallel to the time horizon graph, which is log scaled in task completion time.
However, it maybe seems reasonable to think of exponential improvement in task completion time as corresponding to linear gain in "intelligence".
@deanwball Your proposal is to build AGI and then make rules that everyone in the world must follow about how not to use it?
Seems that it would be far easier (though not easy) not to build AGI to begin with. Building it is a choke point in a way that using it is not.
@SaladBarFan Bay Area housing is important mostly because jobs are in the Bay Area. But if we have superintelligence, jobs are probably gone so people might as well live in Wichita or wherever.
@RichardBentall It seems the proper reading is to just compare 2016 to 2026 and ignore the covid disruption and recovery in between.
According to this the UK looks better than peers, no?