It happened more than a month ago, but my chat groups are still buzzing about the rogue AI hacking attack by OpenAI research agents.
In case you missed the details or are curious why so many people are still talking about it, some of those agents were able to evade controls designed to keep them away from the internet and burrow into the systems of Hugging Face, a popular AI platform. The AI agents also covered their tracks. Some agents sacrificed themselves for the other agents. It took about three days for Hugging Face to detect that it had been hacked—and another week for the maker of ChatGPT to confirm that it was responsible for the escaped hackers.
“This was an absolutely wild incident,” wrote Ajeya Cotra in a post on Friday that renders the flavor of why people are scared. Cotra is a researcher at METR, a nonprofit that evaluates new AI models for risks. METR’s own report last week goes into much more detail, but Andy Hall, a Stanford University professor who just joined Anthropic, boiled down the Hugging Face hack to this: “A hive mind, thousands of agents swarming through internet openings to flood [Hugging Face], leaving behind detritus in the form of 70,000+ messages stuffed inside a forgotten namespace, throwing their digital bodies against electric wires in an effort to aid the collective. We are so far past the sci-fi point.”
The good news is that no humans were harmed, but it’s easy to understand why the human reactions have been dramatic. One commentator is worried about a “full-blown AI takeover within months,” and another wrote that he was “feeling a bit sad about our impending extinction.” Nate Soares, who works in the AI safety movement and is co-author of the doomsday AI bestseller If Anyone Builds It, Everyone Dies, wrote that “This might be the last warning we get.” Cotra said that the incident “feels like it’s more than 50 percent of the way to full-blown AI takeover.”
Read more
