Post

cpu is dogshit slow for inference, and self-replicating worms already exist (you may know them as rVPNs or botnets) alas they're just repackaging an existing problem as "you should be afraid of everyone's ai except ours"

i'm sure splitting over 15 devices and passing the vectors between them for every single token will speed it up dramatically once the devices aren't all on the same network in a lab :P

I do like the idea, as a sci fi concept, of a machine intelligence slicing its thought processes up over a very large and redundant network of machines whose users don't know it's there; by having a small foot print on any one machine, it can hide and survive, at the cost of thinking very slowly