Sam Altman: People Are Starting to Talk Like AI

OpenAI CEO Sam Altman mentioned on Monday that folks might have picked up AI’s quirks and are beginning to sound like giant language fashions on social media.

Altman wrote in a publish on X that he had the “strangest expertise” trying on the flurry of on-line discussion board posts about Codex, OpenAI’s new agent tool for builders.

“I assume it is all faux/bots, despite the fact that on this case, I do know codex progress is absolutely robust and the pattern right here is actual,” Altman mentioned on X.

Representatives for Altman at OpenAI didn’t reply to requests for remark from Enterprise Insider.

Altman posited just a few causes for the surge in such content material on social media. He mentioned similarities in writing type may very well be cropping up as a result of “actual folks have picked up quirks of LLM-speak.” Altman added that the “Extraordinarily On-line crowd drifts collectively in very correlated methods.”

One more reason steered by Altman was that the character of the hype cycle round AI instruments “has a really ‘it is so over/we’re so again’ extremism.” The rise of such content material may very well be pushed by “optimization stress from social platforms on juicing engagement and the associated approach that creator monetization works,” he continued.

“However the web impact is in some way AI twitter/AI reddit feels very faux in a approach it actually did not a yr or two in the past,” Altman wrote.

Altman made an analogous remark in an X publish he printed final week.

“I by no means took the lifeless web principle that severely nevertheless it looks as if there are actually a variety of LLM-run twitter accounts now,” Altman wrote in a publish on Wednesday.

Paul Graham, the cofounder of startup incubator Y Combinator, responded to Altman’s publish, saying that he seen the identical pattern when he used X.

Graham mentioned AI-generated content material wasn’t simply coming from “faux accounts run by teams and international locations that wish to affect public opinion,” it was additionally being printed by “a variety of particular person would-be influencers.”

Altman and Graham aren’t the one ones sensing an increase in “AI slop,” low-quality content material made with little effort utilizing AI, on the web.

Substack CEO Chris Best mentioned in an episode of “The a16z Podcast,” which aired final week, that “refined AI goon bots” will saturate the media panorama with engagement bait.

“We’ll reside in a world the place you could possibly have a bunch of AI slop that retains dumb folks clicking,” Greatest mentioned.