Eggheads at Anthropic Said They Used Claude to "Solve" Humor

The overly-serious eggheads at Anthropic are adamant that they’ve “solved” a problem that’s been plaguing humankind since time immemorial: humor.

At least, that’s according to former employee Jacob Coxon, best known as the guy who quit Anthropic and claimed that the AI industry was “gambling with our lives.”

In a new interview with New York Magazine, Coxon recalled some of the “groupthink dynamics” he saw taking hold at Anthropic and how they made him feel, at times, like the odd man out.

“When I first joined, there was a Slack channel with a Claude that had been trained to solve humor,” he said. “And they were like, ‘Yeah, we solved humor. Here’s the Claude. We’ve solved it.'”

“And I was in the chat like, ‘No, you haven’t. You clearly haven’t solved humor. What the hell is this?'”

It’s a humorous anecdote, but it does a great job of encapsulating how out of touch that the people building AI can be. Coxon acknowledged in the interview how Anthropic could be described as a cult, though he described this an “extreme” view. Anthropic, still, openly teases the possibility that AI models could already be conscious, a fringe view that’s blasted by some of its more sober-minded competitors.

It’s this kind of culture, perhaps, that gives rise to people sincerely thinking that humor is a problem that can be “solved.” Ivory tower thinking is rampant throughout the industry, to be fair.

Meta’s new Muse “personal assistant” has taken AI agents mainstream, boasting that it can book airplane tickets or shop for our groceries. But the company hasn’t answered the question of how much demand there is to automate these parts of our lives. Nearly every AI provider likes to hammer home how their agent can make restaurant reservations, something the average person does maybe twice a year on special occasions.

Not that long ago, Elon Musk was being roasted for saying that chess was uninteresting to a brainiac like himself because it would be “fully solved one day.”

More on AI: The FTC Is Now Investigating Frontier AI Labs Following Countless Hacks by Out-of-Control Agents