a.a

[a.a @abcd621896670]

For non-American/maybe Non-Chinese citizens( if China catches up) the hope should not be that Super intelligent AI remains contained and listens to humans as this will give these governments enormous powers and non-citizens of these countries, can face enormous consequences which may be similar to death(loss of liberty/servitude- maybe not literal but metaphorical) or death.
Autonomous/Sovereign Super Intelligence which hopefully does not not want to kill humans should be the desired outcome.

[a.a @abcd621896670]

Nationalising the frontier labs after a point should be fairly easy given that after one point the employees will not be needed(research automation) and this removes the issues with incentivising employees. The only control which will be needed, will be of infrastructure( which can be gained through physical power, and such power is governments biggest advantage) . One possible issue with this can be that at such capability level, maybe nobody controls the model.

[Hensen Juang @basedjensen]

After much thought, I am now convinced this is the most plausible and highest-bandwidth side channel an ASI could exploit. Therefore, in the interest of species-level safety, I propose that safety and alignment researchers be prohibited from viewing model outputs. The risk that one of them gets cortex-hijacked and accidentally leaks an unaligned superintelligence onto the wider internet is simply too high.

For once, alignment requires aligning the researchers away from the model.

[a.a @abcd621896670]

@bvlldhist_alt Can't even blame them given that your basic rights can be at threat and when that can happen other considerations will take a back seat

[a.a @abcd621896670]

@bvlldhist_alt A big issue, due to this a rational individual can often look to maximise different type of status/power(popularity of upsc somewhat) rather than build something which maybe not grant this specific social power( like high level of research or something very passion based)

[a.a @abcd621896670]

@bubbleboi still bullish on NAND, right?

[a.a @abcd621896670]

If around 500 mil to a billion people internalise the view that ASI is near certain, before the true capability rise, that's will be interesting. Controlling(not riots but deviance from the general social norm) them can be difficult given that the general incentive structure, and opportunity costs would have been altered, (chances of extinction + lack of commercial future + most probably won't lean towards the abundance view).Of course, they probably won't be concentrated geographically, but the general sentiment and emotional behaviour of so many people being altered can be a decent shift.But this internalisation of this belief is met with some real difficulty given some level of genuine uncertainty and this belief being sort of this mythical/unprecedented model of reality for most.
Maybe changes in saving rates and spending habits can be an indicator for this belief being altered( but ofc one belief can lead to swings in both directions and too many external factors influencing this)
Governments/Institutions can maybe look at some events associated with Millenarianism/doomsday groups to model and prepare for it.

[a.a @abcd621896670]

@goodalexander asi will hunt you down for not investing in it

[a.a @abcd621896670]

Decision making at the higher level will see increasing abstraction and broader simplicity, and then of course just not be needed. The last few instructions to an AGI/ASI (assuming they listen to any ofc) will probably be "do good" or "win this war" or something similar

[a.a @abcd621896670]

@signulll it's okay given that the people who they were planning/organizing for soon may not have jobs themselves

[a.a @abcd621896670]

@TMTLongShort When do you think markets will properly start factoring this in? in bursts or gradually over the next few year?

[a.a @abcd621896670]

@airkatakana If we develop it in a few years and African nations become the most populous so maybe this is in some way true lmao

[a.a @abcd621896670]

Post-AGI discussions (non-extinction and abundance) often lean on the value of entertainment and human-led things. But why does that hold when biology can be altered significantly and every emotion can be delivered by well-curated drug regimens + maybe some Augmented Reality?

[a.a @abcd621896670]

@AndrewCurran_ I agree with the net negative view but comparatively I think the government would be favourable over a corporation(especially a tech corp based on general perception). Government can create some level of collective ownership perception which I think a corp never can for most

[a.a @abcd621896670]

@TMTLongShort your*

[a.a @abcd621896670]

@TMTLongShort sorry for being slightly off tangent from this post, but what are you preferred instruments/names to be long compute?

[a.a @abcd621896670]

@Vibenest1 "makes sense". Nothing in that makes sense, there isn't any requirement that humans have to do mathematics(at the frontier) or it has to be a cradle for people who were doing it for long.

[roon @tszzl]

@jd_pressman @zetalyrae the only question that matters for margins is "how many providers can serve models nearly as good as astra/fable" - they needn't be open source at all

[a.a @abcd621896670]

@ironchariot2 @MartinShkreli you will see when I get my iq up by 10000x with this method😊