Care to share examples of AI slop you've encountered?

[ElectricBlue]

More AI junk - showing up in advice to tourists in Tasmania. Need a better map .

https://www.abc.net.au/news/2026-01-22/ai-images-of-tasmania-on-tour-website/106253448

 

[astuffedshirt_perv]

...Worse, YouTube's algorithms keep promoting these poor quality fakes. Ghoulish.

As opposed to the ghouls who eagerly seek and slurp that stuff up?

 

[NoJo]

ChatGPT is a vital part of my work life, and it knows I won't tolerate hallucinations. Itt always cites references, and knows that I expect high-quality references, and will check them. It still makes mistakes, but only about once in twenty or so responses, which is pretty damn good.
Basically I use it as a gopher for web research.

 

[ElectricBlue]

ChatGPT is a vital part of my work life, and it knows I won't tolerate hallucinations. Itt always cites references, and knows that I expect high-quality references, and will check them. It still makes mistakes, but only about once in twenty or so responses, which is pretty damn good.
Basically I use it as a gopher for web research.

Five percent error rate, huh? Not something I'd trust.

 

[AG31]

Basically I use it as a gopher for web research.

Yeah, it saves steps to ask "how can I chat with a human support person at xxxxxxxxxxxx." Those answers are usually right. But if there's any possible confusion, like with model numbers or operating system releases, or... or.... I haven't cured it from sounding terribly sure about its wrong answers.

It is handy to be able to feed it a screen shot.

 

[old_prof]

ChatGPT is a vital part of my work life, and it knows I won't tolerate hallucinations. Itt always cites references, and knows that I expect high-quality references, and will check them. It still makes mistakes, but only about once in twenty or so responses, which is pretty damn good.
Basically I use it as a gopher for web research.

Do you actually check if the references exist? It will definitely fabricate them.

 

[NoJo]

Do you actually check if the references exist? It will definitely fabricate them.

Always. And yes, sometimes it produces links to nonexistent urls.

For python coding, ChatGPT 5.2 is pretty damn good, if all you need is code that works (as opposed to elegant code).

 

[AG31]

So I asked ChatGPT what the hours were for library ABC. It gave me a list of summer and winter hours. The winter hours had it open from 2 to 7. So then I asked what the libraries were that were closest to a particular address and open at noon on Tuesday. It listed several, among them "ABC - Open by 9:00 am on Tuesdays and stays open well past noon (typically until 9 pm)." So finally I pointed out the contradiction, and it said (see below, so I don't interrupt the flow of my post):

I went to the ABC library's web site. The ABC library is open 9 a.m. to 9 p.m. every weekday.

Why didn't I do that in the first place, you ask. Because I want to find out how ChatGPT works and lies.

ChatGPT's explanation for conflicting answers.

Official Current Schedule​

According to the library’s published seasonal hours on the town/library site:

• Tuesday: 2:00 pm – 7:00 pm during both the summer (May–September) and winter (October–April) schedules.

So yes — on Tuesdays the library usually opens at 2 pm, not earlier under the current seasonal schedule.

Why It Might Seem Like It’s Open Earlier​

Some other online listings or generic business hours databases (like crowd-sourced sites) may still show older or incorrect hours — like 9 am–9 pm — for this location, but the official ABC Public Library site shows the 2 pm start on Tuesdays.
If you want, I can check today’s current open/closed status based on real-time hours.

 

[MillieDynamite]

I asked my phone for the score of the last playoff game between the Thunder and the Timberwolves, and it gave the wrong score, showing the Thunder as the series loser. I wanted the actual score, since I'd already heard that the most beloved team in all of sports had won. Sigh, fucking Google AI!

Go Thunder!

 

[MillieDynamite]

Oh, Thunder 124, Wolves 94!

 

[SkyBubble]

I've posted examples of what Grok does before but it's quite difficult to talk about what's going on there without breaching the no-politics rule.

I actually find @Grok to be the best of them.

 

[SkyBubble]

I mean, that could just be that it isnt up to date enough to know of a book published in 2025.

It periodically claims that Charlei Kirk, Rob Reiner, and Ayatollah Khameini are alive.

Tomorrow I'll ask it about Lincoln.

 

[SkyBubble]

Five percent error rate, huh? Not something I'd trust.

Yet they trust AI detectors that are much less reliable.

 

[MillieDynamite]

Is he the one who gave the Trumpenburg address?

It periodically claims that Charlei Kirk, Rob Reiner, and Ayatollah Khameini are alive.

Tomorrow I'll ask it about Lincoln.

 

[AG31]

For years I've been grateful to have Microsoft Answers available whenever I have questions or problems about Microsoft products. Real people, who are able to engage in clarifying conversation, respond quickly. They're intelligent and almost always useful.

Today I went to that URL and faced an idiot AI interface that gained me nothing! I can't tell if it finally funnels to real people. Will have to wait and see if I get an e-mail response. It's just AWFUL.