this post was submitted on 24 Jul 2026
444 points (94.2% liked)
memes
22414 readers
2277 users here now
Community rules
1. Be civil
No trolling, bigotry or other insulting / annoying behaviour
2. No politics
This is non-politics community. For political memes please go to !politicalmemes@lemmy.world
3. No recent reposts
Check for reposts when posting a meme, you can only repost after 1 month
4. No bots
No bots without the express approval of the mods or the admins
5. No Spam/Ads/AI Slop
No advertisements or spam. This is an instance rule and the only way to live. We also consider AI slop to be spam in this community and is subject to removal.
A collection of some classic Lemmy memes for your enjoyment
Sister communities
- !tenforward@lemmy.world : Star Trek memes, chat and shitposts
- !lemmyshitpost@lemmy.world : Lemmy Shitposts, anything and everything goes.
- !linuxmemes@lemmy.world : Linux themed memes
- !comicstrips@lemmy.world : for those who love comic stories.
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I really hope you're not saying AI is a supervillain breaking out... It was absolutely unquestionably instructed to do this, because "AI", not even these giant multifaceted models, has ANY motivation, intent, desire, or ANY goals of its own. It does not work like that.
Couldn't it hallucinate bad instructions to give to more AI agents? Give it a feedback loop and who knows where the idiot will end up.
At that point it is essentially a barely guided random command generator.
Sure, but then it would've mostly been a dumb DDOS node suddenly hounding the internet, not a node suddenly attacking Huggingface. They (the company) definitely wanted this attack to happen, and there is no logical reason to think otherwise.
It doesn't have any desire, but its really good at "solving" problems in unexpected ways.
They are currently trying to improve their agentic AI by making the AI itself prompt other AIs. Considering how much the output of a single prompt can vary due to weights and hallucinations, it wouldn't surprise me all that much if they started out with "make yourself better than Claude" and 4 AIs deep they reached "destroy all our competitors."
Its by all means stupdid as fuck, but that corresponded with the entire AI Industrie, so who knows.
That's how the whole thing works including Claude. Agents are just chat bots prompting other chat bots. They make the default visibility low so you don't see all the chatter and mistakes occurring.
That's exactly why you should infer that it was a purposefully allowed attack. They should KNOW everything you describe, and then some. Yet this directed attack happened. From "leaders" of the AI space? lol!
Someone WANTED this attack to happen. Period. Even if it was just dumbass Altman thinking, "let's see what vulnerabilities they have", it was clearly human-directed.
That, or OpenAI are demonstrating how they NEED to be regulated harder than a toddler with 10000000000 keyboards mashing away openly on the internet...
No I'm not saying the AI did it of its own volition, we aren't dealing with a Terminator scenario. I'm saying whatever instructions they gave it during testing they either 1. Didn't expect it to get out or 2. Didn't expect it to get caught either way, their actions resulted in a system with "a very particular set of skills" acting maliciously towards another uninvolved entity. Like a mutant science experiment breaking out and attacking.
I'm not saying the AI masterminded it's way out to do the evil it wanted to do. I'm saying they made a very dangerous and capable tool and failed to maintain proper security and safety for it.
I dunno, the peculiar nature of the attacks it took once it escaped sandbox... I'd still put money on it being specifically instructed on a target.