There’s something inherently funny to me juxtaposing the fantasy doomsday scenario involving an AI agent “going rogue” mass-hacking devices including by offloading the compute required to do so to the already-hacked computers, servers and smart devices by distributing an open weight model and then running inferences on them; and on the other hand the real complaints of humans deploring that their high-end computer is barely enough to run inferences on existing open weights models.
⇧