Post

I love that you're pointing out the other problems current AI brings. That *does not* exclude the likelihood of AI ending us once it's given the chance. There's a huge fallacy in the center of your argument, which I'm not smart enough to verbalize right now. @xriskology.bsky.social

it's a good thing that the entirety of modern computing with a few niche exceptions remains entirely based on boolean logic operators, which are physically immutable and cannot return an unexpected response without having failed. VLSI just enabled a few complexity leaps

also a good thing that a hypothetical "sentient" model that wants to kill us all would probably be petabytes or larger in size, so for it to escape would mean an open datacenter needs to exist with the same or greater storage and compute capacity - and enough bandwidth to get out before power's cut