AI’s Flaws Highlighted by New York Post Reporter

• Steve Cuozzo from the New York Post criticized artificial intelligence for its significant misunderstandings and inaccuracies.
• AI tools like ChatGPT and Claude have made several incorrect assertions about Cuozzo’s life and work.
• Anthropic, a company valued at roughly $1 trillion, is gearing up for an IPO, but its Claude AI has not been reliable for Cuozzo.

It’s a bit much to say I’m “dead,” but that was the summary provided by a recent ChatGPT post referencing one of my articles. Miraculously, I’m very much alive and kicking.

Isn’t it curious? If AI is touted as so intelligent, how can it make such foolish mistakes? I sometimes think “amazing ignorance” might be a more fitting label.

People warn that AI could soon govern us, dictate our thoughts, and influence our very existence. The Pope cautioned against its dehumanizing effects, likening it to a modern-day “Tower of Babel.” Economists express concerns about job losses and a future where urban centers resemble abandoned ruins.

But for those convinced AI can transform human life, shouldn’t they first verify if these bots have their own facts right? It seems wise to question how accurate AI platforms are regarding personal history.

Many enthusiasts overlook whether AI can truly perform its primary role: gathering and presenting information in a way that standard search engines can’t.

Sure, ChatGPT can string together entertaining narratives or even conjure fictional characters, but when it comes to straight facts about actual people? Not so much.

The same is true for Claude. I asked it a few questions, and, believe it or not, it claimed I’ve written a cookbook. I have no idea where it got that idea. I’ve never penned a cookbook or even lived in Greenwich Village.

Interestingly, Claude does recognize my long-standing role as a commentator on commercial real estate. However, when I asked if it had information regarding the Herald Tower, it just suggested I check the New York Post’s site. Isn’t that what the AI is meant to do?

Anthropic, which filed for an IPO recently, is still not on my investment radar.

Out of five different AI platforms I assessed, Elon Musk’s Grok presented the most accurate depiction of my background. It even compared my restaurant critiques effectively with takes from the New York Times’ Pete Wells. That’s impressive!

Meanwhile, ChatGPT, despite its skills in literary analysis, missed the mark when it came to my own background. It didn’t even confirm my educational history. Instead, it pointed out the lack of publicly available information about my academic credentials. Quite odd, considering my degree from Stony Brook University is all over Wikipedia.

This wasn’t the only mix-up. When I asked Perplexity AI whether I had authored any books, it drew a blank.

It confused me with a nonexistent cookbook called The Finger Food Cookbook that I didn’t write. That’s amusing, but frustrating as well—my actual book, “It’s Alive: How America’s Oldest Newspaper Cheated Death,” has plenty of visibility online.

Gemini AI, on its part, mistakenly credited me with “co-authoring” a book that was entirely the work of my wife. It’s a classic case of AI mimicking Wikipedia’s content like students used to do with encyclopedias—it’s not reliable.

When Claude reported me as a cookbook author from 1997, I couldn’t help but laugh. This isn’t my work—many books with similar names exist by other authors. Once a false statement seeps into an AI’s knowledge base, it seems impossible to remove it. Even Perplexity has this fiction embedded now.

Perhaps, as Musk has suggested, AI may one day outsmart humans. Or, conversely, as critics warn, it might lead to enough job losses that our cities could mimic abandoned ruins. Until that day, it might be best to keep your sources in check and learn how to navigate information—especially if it’s from Wikipedia.