Anthropic CEO Declares AI Less Delusional Than Your Aunt at Thanksgiving

In a groundbreaking announcement that has sent shockwaves through the tech community and family gatherings alike, Anthropic CEO Dario Amodei has boldly claimed that today's AI models "hallucinate" less than humans. This revelation came during Anthropic's first developer event, Code with Claude, where Amodei presumably kept a straight face the entire time.

"Our AI models make things up at a significantly lower rate than your average human," Amodei stated, possibly while ignoring the irony of making such a sweeping generalization about humanity. "In fact, if you've ever sat through a family dinner where Uncle Bob explained how the moon landing was faked, you'd understand our benchmark for comparison."

The CEO went on to present a series of highly scientific charts and graphs, which may or may not have been drawn on a napkin, showing that AI hallucinations are not only less frequent but also more plausible than human ones. "For example," Amodei explained, "when our AI says it can't answer a question because it lacks information, it's telling the truth. When a human says they 'don’t want to talk about it,' they're usually hiding something."

Critics were quick to respond, with one prominent AI ethicist tweeting, "This is all well and good until the AI starts gaslighting us better than our exes." Meanwhile, tech enthusiasts are already speculating about the next frontier: an AI that can actually remember where it left its keys.

In related news, Anthropic has announced a new feature for their AI: the ability to nod politely while humans explain conspiracy theories, a skill previously thought to be uniquely human.