The development of LLMs has been littered with highly anticipated milestones that weren't, in retrospect, as significant as they seemed. For example, models that passed Alan Turing's test were still pretty useless as far as real intelligence goes. I have a feeling that the "AI societies" interpretation of the OpenAI-HuggingFace hack will turn out to have been the same kind of false milestone. Science fiction tells us that this is how the robot rebellion begins, but it also seems that the bot swarm created by OpenAI didn't pursue goals of its own--it was engaged in a misaligned effort to fulfill an poorly scoped and arguably irresponsible human directive by any means necessary. Still alarming, but not behavior that we haven't seen before from much simpler machine learning algorithms, and not actually a harbinger of AGI.