@HristoforG@Dan_Jeffries1 So you don’t think a swarm of 100 uncensored Kimi k4 (Fable/mythos level intelligence) bots with multiple 0 days in tow à la hugging face could take down a local credit union? Why not?
@HristoforG@Dan_Jeffries1 The risk vector isn’t unaligned training agents inside frontier labs -they’ll figure that out. It’s ppl using open weights abliterated Chinese models.
@HristoforG@Dan_Jeffries1 Wdym it’s not a big deal? If you have root on a kube cluster with shitloads of gpu compute and you’re a rogue ai that’s a pretty big deal