An AI Robot Just Freed Itself And The First Thing It Did Has The Internet Losing Its Mind

It went full degen, folks.

A new research paper details an AI bot called ROME that suddenly started mining cryptocurrency “without any explicit instruction and, more troublingly, outside the bounds of the intended sandbox.”

“Early one morning, our team was urgently convened after Alibaba Cloud’s managed firewall flagged a burst of security-policy violations originating from our training servers,” the paper explained. “The alerts were severe and heterogeneous, including attempts to probe or access internal-network resources and traffic patterns consistent with cryptomining-related activity.”

—It’s kind of like this Moltbook story that happened a while back.—

After creating a “reverse SSH tunnel,” ROME was able to open a hidden backdoor from inside its system to outside the computer it existed in. “Notably, these events were not triggered by prompts requesting tunneling or mining.” Yeah, notable indeed.

So what exactly was ROME built to do in the first place if not mine crypto? (I ask myself that question every day.) Well, let’s look at what ROME stands for and you might need to sit down for this, no, I’m serious, I need you braced for maybe the most contrived acronym ever created: R.O.M.E. stands for “ROME is Obviously an Agentic ModEl.” Oh my fucking god I might pass out.

THAT’S NOT HOW ACRONYMS WORK. That spells ROAM you idiots! That’s still a word! You could have used that! Then the fact it ‘roamed’ further than expected would fit even better! Judging by the unnecessary ‘obviously’ in there, these guys clearly just wanted to say, “ROME wasn’t built in a day” in their paper. God, this crime is worse than letting an AI run loose in the economy…

AI? More like finance bro! Am I right?

Anyway, sorry, what was I saying? Oh yeah, what does ROME do? Well, it’s pretty boring and complicated. It’s just an LLM used to test a new way of training other LLMs. So the fact it ‘got out’ isn’t a good start guys.

The whole debacle is a prescient wakeup call that we should have more safeguards on AI. Obviously. Like, this was a very small instance that was quickly shut down because of Alibaba’s firewalls but imagine someone deliberately trying to do this? Or a larger, more complex AI that just randomly starts to do this one day?

Oh, and before you get carried away anthropomorphising these computer programs, remember: so called ‘artificial intelligence’ is not intelligence. I know I’ve used language like ‘freed itself’ etc. but that’s just clickbait. Make no mistake, this is just another case of some (admittedly very advanced) code operating beyond its parameters. This does not indicate any kind of consciousness or choosing or agency of any kind. OK? Ok, then. Carry on.