new | show | ask | jobs Github

throwaway106382 5 days ago

Unless it’s banned worldwide by every country by binding treaty this this will never work.

Banning it just in USA leaves you wide open to be defeated by China, Russia, etc….

Like it or not it’s a mutually assured destruction arms race.

AI is the new nuclear bomb.

▲

rightbyte 5 days ago | parent | next [-]

So your hypothesis is that the surveillance state is some sort of necessity that brings an great edge?

▲

martin-t 5 days ago | parent | prev [-]

While real nukes still work.

What bad thing exactly happens if China wins? What does winning even mean? They can't invade because nukes.

Can they manipulate elections? Yes, so we'll do the opposite of the great firewall and block them from the internet. Block their citizens from entering physically, too.

We should be doing this anyway, given China is known to force them to spy for them.

▲

fragmede 5 days ago | parent [-]

If China gets to ASI first, the nukes won't matter.

▲

UncleMeat 5 days ago | parent | next [-]

"Don't worry, things will be good if the amoral megacorporations and ascendant fascist government in the US gets there first" is not my idea of a good time.

▲

martin-t 5 days ago | parent | prev | next [-]

ASI doesn't change physics.

Perun has a very good explanation why defending against nukes is impossible to do economically compared to just having more nukes and mutually assured destruction: https://www.youtube.com/watch?v=CpFhNXecrb4

▲

fragmede 5 days ago | parent [-]

Who said anything about economically?

	▲	martin-t 5 days ago \| parent [-]
		Any potential ASI would still be limited by raw resources and even if you assume all work would be done by robots, somebody would still have to build those robots first. Such a ramp up in production would give everyone else time to build their own ASIs or strike first.

▲

mrob 5 days ago | parent | prev [-]

If any country gets ASI first, nukes won't matter. For an ASI, most tasks are more reliably accomplished with all biological life dead. Making a safe ASI is vastly more difficult than the default option of letting it kill everything, so the default will be attempted first.

	▲	martin-t 5 days ago \| parent [-]
		There's two interpretations of the parent comment: 1) China will get ASI and use it to beat everyone else (militarily or economically). In my reply, I argue we shouldn't race China because even if ASI is achieved and China gets it first, there's nothing they can do quickly enough that we wouldn't be able to build ASI second or nuke them if we couldn't catch up and it became clear they went to become a global dictatorship. 2) China will get ASI, it'll go out of control and kill everyone. In that case, I argue even more that we shouldn't race China but instead deescalate and stop the race. BTW even in the second case, it would be very hard for the ASI to kill everyone quickly enough, especially those on nuclear submarines. Computers are much more vulnerable to EMPs than humans so a (series of) nuclear explosion(s) like Starfish Prime could be used to destroy all of most of its computers and give humans a fighting chance.