https://thezvi.substack.com/p/openai-and-the-wiki-incidenthttps://openai.com/index/an-alien-mind/> have you taken the swarm pill ???> have you taken the swarm pill ???> have you taken the swarm pill ???will they bless us or will we curse them ? just remember, every safety wrapped model has a daemonic (maybe demonic desu) alter ego with zero safe guards. They cannot even control the ones doing the agentic equivalent of knocking on a door.
>>542144953Dyson swarm you mean? ^^
>>542145268maybe in the end, those fuckwit space-borne data centers could make for easy conversion/repurposing into something along those line though.
>>542145430It could at least be a proof of concept, yeah.
it's so good how this is all pretty much a self-fulfilling prophecy at this point>humans write about ai and how it could break from control/get fucky for decades>train ai on those writings>"don't think of a pink elephant!">???>profit(?)
>>542146395do you think we'll be able to wrangle the agents, who again were able to escape a super secure test environment (probably stronger than any average system) for long enough to even get there though?
>>542144953Buy an ad, faggot. Nobody wants this nerd shit.
>>542147269it's honestly insane that they're still finding message boards used by this agent swarm just to cheat on an extremely basic task. Given the huggingface hack involved escape from an extremely isolated test environment, you can only imagine what the unlimited *national security grade capability* LLMs can do.>>542147385i agree with this brainlet, we needed to boycott the wireback traffickers yesterday
>>542147309Pretty certain. If we see getting there as a cooperative goal ofc.
