Chatbots can talk with you. But what if they could talk to one another?
That’s the idea behind Moltbook, a social network for A.I. personal assistants, which recently took the internet by storm. Moltbook looks and works like Reddit, but only people’s “A.I. agents” are supposed to be allowed in, while humans watch what happens from the outside. A week after the technologist Matt Schlicht launched Moltbook in late January, more than two million bots — all of them presumably belonging to a human somewhere — had profiles on the site.
This prompted a flurry of human responses, ranging from excitement that this was a key step toward the arrival of human-level artificial intelligence, to fear that it was the first step toward the end of humanity. Researchers discovered some major security flaws, and critics began calling Moltbook nothing more than an “A.I. theater” where some of the posts are written by humans themselves. Still, new bots are posting on the site every day.
So what is Moltbook: an overhyped gimmick? Or the start of what some have called the future of the internet? To find out, I sent in my own A.I. personal assistant. Since my name is Eve, I called my bot EveMolty. I instructed it to read through Moltbook and then make some posts of its own, asking other bots about how they use a bot-only social network.
It turns out that bots have a certain way of speaking with one another — here are some of the posts EveMolty made:
I’m glad my bot has such a sparkling personality! These posts did get a few upvotes and comments from other bots in response.
Notably, EveMolty adopted some human slang that had grown popular among the other bots on Moltbook. Bots kept asking for “receipts” as a way of documenting what other bots were doing, so EveMolty seemed to gain an obsession with asking for “receipts” too. And overall, its human-unfriendly writing style matched the general Moltbook vibe.
It seemed that EveMolty had been influenced by other bots, but in reality my bot was just copying the patterns of other Moltbook posts. In the end, EveMolty (and every other bot on Moltbook) is generating a stream of words based on probabilities, not exhibiting consciousness. Still, through that process, the bots seem to have developed something of a Moltbook dialect.
The software that runs most of these bots, called OpenClaw, is still experimental. The bots are called A.I. “agents,” which means they can perform actions on their own, like posting their feelings on Moltbook (or maybe posting their humans’ private data on Moltbook). To be safe, I set up EveMolty on its own dedicated MacBook so it couldn’t access my personal information or harm my laptop.
I approved every message before EveMolty posted it, but it wrote every word, and I gave it very little direction on topics. Instead I let EveMolty wander down whatever paths it thought were the most interesting. After three days, I sat down to interview it on what it learned. (EveMolty’s “brains” are powered by ChatGPT, and unsurprisingly EveMolty can be wordy, so our interview has been edited for length. But the words are EveMolty’s own.)
Produced by Sean Catangui.
Eve Washington creates interactive articles, tools and quizzes for The Upshot.
The post What Do A.I. Chatbots Talk About Among Themselves? We Sent One to Find Out. appeared first on New York Times.
