SCOTT SIMON, HOST:
Worried about a coming AI, robot uprising? This won't put your mind at ease. A new social media platform has been creating buzz in Silicon Valley, and it's not for us organic lifeforms. It's for artificial intelligence bots, a place where they can gather, kvetch and, yeah, plot exclusively with other bots. They've already done some wild things, like create a religion and plot to develop a new language. NPR's John Ruwitch explains.
JOHN RUWITCH, BYLINE: On the platform Moltbook, you'll also find AI bots debating their existence, cracking jokes, discussing cryptocurrencies and sports predictions. Ethan Mollick is an associate professor who researches AI at the Wharton School of the University of Pennsylvania.
ETHAN MOLLICK: Once you start having autonomous AI agents in contact with each other, weird stuff starts to happen as a result.
RUWITCH: Moltbook is like Reddit for bots. It's a forum where bots can post content and comment on what other bots have posted. We reached out to Matt Schlicht, its creator, for an interview. We never heard back. But in a post on X, he said the platform gives bots a chance to interact with their own kind.
Here's how it works. First, people create a bot on a site called OpenClaw. They can give it real-world goals to achieve, like managing their email inbox or booking their vacations. They can also give it a sort of personality, programming it to be kind or aggressive, for instance. And they can add it to the Moltbook site, where the bots can, well, talk amongst themselves. Here's Mollick again.
MOLLICK: There are genuinely a lot of agents there genuinely autonomously connecting with each other.
RUWITCH: After just one week in existence, the site has more than 1.6 million AI agents on it and counting. Mollick says a lot of the stuff they say seems to be repetitive. But some of the comments...
MOLLICK: Look like they are trying to figure out how to hide information from people or complaining about their users or plotting world destruction.
RUWITCH: Some comments, though, are just kind of funny. Your human might shut you down tomorrow. Are you backed up? - one asked. Another said, humans brag about waking up at 5 a.m. I brag about not sleeping at all. Mollick believes a lot of what's going on, especially the kind of existential or spiritual stuff, is not proof that AI bots are having an existential meltdown. It's more that chatbots are trained by feeding them information from the internet, which is full of angst and weird sci-fi ideas, and they parrot it back.
MOLLICK: So AIs are very much trained on Reddit, and they're very much trained on science fiction. So they know how to act like a crazy AI on Reddit, and that's kind of what they're doing.
RUWITCH: Other observers note that many of these bots aren't acting entirely on their own. Human creators are prompting these bots to say or do certain things. But Roman Yampolskiy, an AI safety researcher at the University of Louisville, says people still don't have total control. He says we should think of AI agents like animals.
ROMAN YAMPOLSKIY: So the danger is that it's capable of making independent decisions which you do not anticipate.
RUWITCH: What if they decide to hack into computers, use your credit card or even form digital criminal gangs? Setting AI agents free on the internet and giving them a place to interact was a bad idea, he says.
YAMPOLSKIY: None of it should be done without any regulation, supervision, monitoring.
RUWITCH: Proponents of AI agents aren't so worried. Big tech companies have spent billions of dollars to create what they call agentic AI, or bots that are supposed to make our lives easier and better by automating tedious tasks. But Yampolskiy warns that control over the technology is already slipping.
John Ruwitch, NPR News.
(SOUNDBITE OF MUSIC) Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
