ChatGPT and the Memory of Don Rickles: When AI Distorts Reality

ChatGPT and the Memory of Don Rickles: When AI Distorts Reality

Recently, a person asked ChatGPT to identify an unnamed male celebrity who allegedly sent suggestive messages to Lena Dunham in 2012. To her great surprise, the AI named Don Rickles, a claim that seems doubtful, to say the least.

This raises an important question: how should we approach learning about and using artificial intelligence?

An Unlikely Anecdote

Have you ever heard of the incident in which Don Rickles supposedly tried to seduce Lena Dunham in the middle of the night? No? Allow us to explain. But first, let's talk about Reese Witherspoon.

A person with simple interests: new technologies and celebrity gossip. That is why she was intrigued by the recent online controversy over Reese Witherspoon's call for women to learn how to use AI. The statement provoked so many reactions that Witherspoon had to publish an additional explanation.

AI and Celebrity Gossip

This person's interest was also piqued by Lena Dunham's new book. These two subjects are linked, in a way. Keep reading! Reese is generally right about the importance of AI, similar to what many business leaders say. However, her comments pushed this person to think more deeply about what "learning to use AI" really means. Is it simply writing emails with ChatGPT? Understanding the technology behind different models? Or mastering coding? What level of AI skill is needed to avoid being left behind in the workplace and in everyday life?

One way this person has explored the question is by using ChatGPT as a kind of super search engine, to dig up information she knows is available online but that would take considerable effort to find by traditional means.

An Investigation into Gossip

A recent example concerns, of course, celebrity gossip. While reading Dunham's new memoir, "Famesick," this person came across juicy anecdotes about celebrities, some named, others anonymous. An anonymous article mentioned a male celebrity who allegedly sent a flirty message to Lena after meeting her backstage during the taping of "The View" in 2012. She figured she could solve the mystery by finding out who the other male guest on that episode was, information that should be online but is hard to dig up.

She therefore asked ChatGPT to identify the male guest on the episode of "The View" that Lena appeared on that year. At first, ChatGPT claimed there were only the show's four female members. When she pushed, the AI suggested Chris Evans and Chris Hemsworth, which was incorrect, since they appeared on a different episode the same year, according to IMDb.

An Unexpected Response

When this person specified "a comedian," as Dunham had described the man, ChatGPT confidently replied that it was the legendary comedian Don Rickles who had sent Dunham a message after the show.

She burst out laughing, because of all the people who might have sent an "are you awake?" message late at night, she is almost certain it was not Don Rickles, who would have been 85 at the time.

A Description That Doesn't Fit

Dunham's description of the man: "a bit of an American Hugh Grant, famous for his chatty charm and ability to seduce his on-screen partners with his fast-talking anxiety and hand gestures. Apparently a comedian, he was there to promote his gothic-tinged film, where he had made a dramatic turn." This does not resemble a Borscht Belt insult comic like Don Rickles at all.

After spending a great deal of time searching the internet in the traditional way, this person can make a few guesses about how and why ChatGPT got it wrong here. The IMDb episode guide for "The View" from 2012 is uneven, with entries for some episodes missing guest information, and no videos accessible online. The only evidence she found that Lena Dunham was ever on "The View" on April 20, 2012, was a Vulture blog post from that day, with an embedded YouTube clip marked private.

The Limits of AI

Knowing this, it is possible to begin to see how the AI got confused: when information is lacking, the AI sometimes mixes together what it can find to try to provide a plausible answer. Chris Evans and Chris Hemsworth appeared on the May 4, 2012 episode of "The View," and Dunham and Rickles appeared together in an episode in 2016.

ChatGPT doing this kind of thing, essentially guessing at what you might want to hear, might be useful if you are trying to write an email to a friend, perhaps? It is not useful, obviously, if you are looking for a specific fact and it simply makes something up.

For the record: neither Lena nor Don (who died in 2017) nor the National Comedy Center, which holds the Rickles archive, responded to this person's requests for comment.

Are We Stuck in a Pizza Glue Loop?

Listen, we understand. It is not particularly exciting to point out that ChatGPT gets things wrong in the spring of 2026. We know this, or at least we should all know it. Yet this person keeps running into so many obvious errors when she asks AI for factual things. These are the glaring mistakes she spots when she knows that what the AI generated is not the right answer.

But what about the errors she does not spot, or does not even know how to spot? Things she blindly accepts as facts? For work-related things, she will always check, but in those cases, is it really saving her time?

The Future of AI

How fast will this improve? Will we be stuck in a pizza glue loop forever? Is this going to make a bunch of lawyers and tax accountants lose their jobs? We mean, okay, sure.

This is where the "Learn to use AI!" idea from Witherspoon and other bosses seems frustrating. This person feels fairly confident using various AI tools and has a decent idea of how they work. She is a woman, and she has learned to use AI! And yet here she is, still dissatisfied.

There is a gap between what Reese Witherspoon wants for this person and what this person wants from AI, and also the benevolent image of comedy legend Don Rickles. For now, these things simply do not line up properly.