Don’t be too shocked, but I think AI does have some utility, despite the occasional hallucination.
A Utah police department’s use of artificial intelligence led to a police report stating — falsely — that an officer had been transformed into a frog.
The Heber City Police Department started using a pair of AI programs, Draft One and Code Four, to automatically generate police reports from body camera footage in December.
A report generated by the Draft One program mistakenly reported that an officer had been turned into a frog.
“The body cam software and the AI report writing software picked up on the movie that was playing in the background, which happened to be ‘The Princess and the Frog,” Sgt. Rick Keel told FOX 13 News. “That’s when we learned the importance of correcting these AI-generated reports.”
We use AI at my university for that purpose, too. Ever sit through a committee meting? Someone has to take notes, edit them, and post them to a repository of meeting minutes. It’s a tedious, boring job. Since COVID moved a lot of those meetings online, we’ve found it useful to have an AI make a summary of the conversation, sparing us some drudgery.
Of course, someone should review the output and clean up the inevitable errors. The Heber City police didn’t do that part. Or maybe they did, and someone found the hallucination so funny that they talked about it.
Did they find a princess to kiss him?
Also…”Kermit Vs The Swedish Chef”
Perfect Cell looks a bit odd. AI hallucination?
.https://youtube.com/watch?v=7mJOYaAxd34
An AI-Generated NWS Map Invented Fake Towns In Idaho
Well, nobody says that AI is totally useless, it has brought real progress in some fields:
https://www.nature.com/articles/d41586-024-03214-7
(sorry, paywalled and Nature is expensive)
https://deepmind.google/science/alphafold/
It could certainly do away with some classes of repetitive tasks, following the trend towards automation seen in the last two centuries or so.
The problem right now is this idiotic mania which brings those who should know better stuffing it absolutely everywhere. That’s not progress, it’s yet another example of cattle-like groupthink.
Plus, the possible misuses and/or downright criminal uses of AI are legion and of course, to no one’s surprise, nobody gave the slightest thought about that either.
Plus-plus, there’s a real risk of yet another worldwide economic bubble going pop, to no one’s enjoyment.
So yes, overall I am rather lacking in AI enthusiasm. Grump.
Technology is always a double-edged sword that can be used for either good or evil. (Or neutral too I guess?)
What we can do versus what are good things to do vs what are terrible things to do.
We need to choose wisely.
Too often folks don’t.
I do thing AI should always be clearly labelled as such and known to be such so we don’t consider reality with non-reality. Becoz that causes a fucktonne of poblems.
I’ve found AI summaries pretty bad when I read them. But, the competing option is usually no summary at all. And somebody keeps asking for the AI summaries so I guess they must be getting something out of it.
I suspect the weak point in AI summaries isn’t just the LLM, but the voice to text transcription. And in the case described, maybe it’s computer vision? Why don’t more people talk about voice to text and computer vision? I really want people to better understand the distinctions between different AI technologies.
“Committee meting” sounds fairly accurate for some such gatherings, doling out just enough information to seem useful… (Also, three double-letters in close proximity – who needs a fourth?)
At least Draft One is well-named and directly implies some further review and editing is warranted.
I like the fact that the author of the article felt the need to specify that the report was, in fact, false. As if there was a real possiblity that a police officer really had been turned into a frog, and he wanted to forestall any possible confusion.
If the police officer had been working in Ankh-Morpork , the Discworld, it would have been eminently plausible. I recall some heavy bruiser working for the Patrician once was turned into a toad after adressing a wizard with less than perfect manners.
Currently stuck in an online meeting with boss and coworkers. Don’t need AI for the summary:
I’m sorrounded by idiots!!!
One problem is that people* assume the damn things are reliable and authoritative — until they do something obviously absurd like this, at which point you should be worrying about what less-obvious bullshit they’re generating. Yes, AIs can probably pick up some of the drudge work, but the output needs to be sanity-checked by a human (and at some point the sanity-checking will be more work than just doing the original work).
* Mostly non-technical users, I think. Those of us who pay attention know better.
@9
In this day and age, with the incompetent reporting of important national and international events we have from newspapers and TV networks, it is refreshing that there does exist a small assemblage of old school journalists who can write the truth about the controversies of the day with out sucking up to the powers that be.
outis @ #4 — “Well, nobody says that AI is totally useless…”
Yes they do. And some people have advocated curtailing it if not banning it outright. Generally what people are calling “AI” these days is LLMs. As statistical models they have tremendous potential but they can produce some crazy results like cops turning into frogs, and worse. Where a business using an LLM can’t afford hallucinations, there are approaches coming online to steer the LLM such as knowledge graphs, ontologies, GraphRAG, and various other guardrails to keep the AI on track by embedding information in the material. Of course, that’s not as straight forward as it might sound.
The other problem with LLMs is they use a lot of energy.
stevewatson @ #12 — “One problem is that people* assume the damn things are reliable and authoritative…” Yes, some people assume that but I can assure you that the marketing and legal people make no such assumption.
I am in construction and the meetings we have with sub contractors and architectural firms in Microsoft Teams are summarized by Read AI. It’s nice when I am off shift and have missed a meeting, I can browse through the summary and see what was discussed. Very handy.
It could have been worse. The movie could have been Babe or worse still Animal Farm.
Is this why there’s so many reports of Giant Frogs taking over Portland OR?1??
@14, robro: of course, everything in moderation… and correcting tools would be helpful to say the least. Common sense also works, as those bright sparks announcing the discovery of “millions of new materials” found out:
https://www.nature.com/articles/d41586-025-03147-9
we are in the very early days after all and that’s why I find this kind of brainless hype so annoying.
Your point about the energy being wasted is also important, and it’s causing other bright sparks announcing the launch of (imaginary) new nuclear reactors right and left:
https://www.theregister.com/2025/08/18/google_smr_datacenters/
molten salt… talk about fever dreams.
One good thing may come out of this, anyway: should the bubble go kablooie, there will be a damn lot of cheap computing capacity available for some serious project at last. We’ll see.
I accept PZ stating that there are some things AI can help with. But, I must say the price we pay is too high for so little gain by a few.
It is so arrogant for these idiotic plutocratic AI developers to think that ingesting tons of human stupidity in writings will create an ‘intelligent’ result.
Based on the qualified tech sites I read, it is indeed ‘artificial’ it is almost never ‘intelligence’.
And, I CONDEMN the destruction of the health of tens of thousands of people by illegal (and in a few cases, coerced legal) pollution caused by the data centers (lookin’ at you mofo muskrat!) and the massive waste of water and obscene jacked-up electric bills of people forced to pay for the billionaires irresponsible actions. It is OBSCENE that the magat-in-chief is causing a half-dozen ugly coal power plants to stay pumping out poison just to feed the plutocratic datacenter greed.
Also, what about the evil muskrat’s grok that continues to make obscene images of children???
I want to tell these AI slop CEOs and investors to just “rip off the bandage,” swallow the sunk costs, and let us regular people get on with our lives. Of course, that would mean they have to admit they’ve been living in a fantasy world.
“The body cam software and the AI report writing software picked up on the movie that was playing in the background, which happened to be ‘The Princess and the Frog,” Sgt. Rick Keel told FOX 13 News.
Not a hallucination, then. It was in the input stream.
I wonder, what does that someone compare the AI summary to to spot/confirm errors, and use as a valid source for corrections when one is found?
karellen, whoever reviews it should have been present during the meeting; that is the fix.
I myself used to take minutes — and I was always aware that what’s in the minutes is official.
If it ain’t there, it is only alleged, formally. And if it’s there, well, that’s in the minutes.
I shall never tell whether I, um, ever massaged the minutes to improve them.
(Or: “If it isn’t in the minutes, it didn’t happen” and “The record speaks for itself”)
@#10: I believe it was the Patrician himself who was turned into a frog, in Discworld #5, Sourcery.
I really really really recommend this excellent Democracy Now interview w Karen Hao, author of Empire of AI. She really gets at the fundamental rot behind the whole scam. https://www.youtube.com/watch?v=Xa6JuimHoEA
she says the main villain is not AI per se but the Silicon Valley “scale at all costs” mentality, and one of the points she makes is that AI can do great things with even tiny amounts of high-quality data. (The exact approach of the megagarbage approach the big companies are using.) The example she uses is of people using it to help preserve the Maori language.
During Covid, we were required to make recordings of our Zoom lectures that included voice-recognition transcripts for the hearing-impaired students to read.
You should have read what the app “recognized” and transcribed as my chemistry lectures, filled with chemistry words. Not one sentence, the entire semester, was even intelligible.
How AI would do at transcribing me speaking about covalent bonding, or stoichiometry (say that one five times fast), or the Schrodinger equation? Or even with me just reciting the names of the elements (rubidium, ruthenium, rhodium, rhenium, radon, radium, rutherforidum, roentgenium?) Just wondering.
I am longing to try it with, say, ChatGPT. Such wondrous hallucinations as would ensue!
“I’m longing to try it with, say, ChatGPT. The hallucinations would be spectacular.”
Um.
It’s dead easy: use a phone, tablet, laptop, desktop, whatever so long as it has a mic and can run any GPT‑based app or browser (ChatGPT, Copilot, Edge, mobile apps, API wrappers).
Whence the unnecessary longing?
How is the weather there today John?
As to AI, I have 2 comments.
1. It’s ruining porn.
2. It’s not helping in my job at all, because I can’t trust its summaries and have to read the documents myself to make sure there’s no error(in this case, medical reports).
(I know it was 44.9C in Adelaide today)