(C) Daily Kos This story was originally published by Daily Kos and is unaltered. . . . . . . . . . . Open thread: Failures of natural intelligence [1] ['Alonso Del Arte', 'Dan K', 'Nanny Ogg', 'Sabrina Haake', 'Chitown Kev'] Date: 2026-09-10 You must use A.I. or get left behind. Without A.I., you won’t be able to innovate new ways to mess up tasks for which we already have reliable tools and methods. It still amazes me that people can look at artwork created by A.I., see its obvious flaws, and still trust A.I. to do things where the flaws are harder to see, such as software development. It also surprises me that people trust A.I. for things they once knew how to do without any problem before the A.I. craze started. First example: Spreadsheets Walking downtown, I overheard a couple of men talking about some list or other. “Can A.I. sort these by dollar amount?” asked one of the men. “Yeah, probably,” the other man replied. I don’t know what they were referring to by “these” and I didn’t want to look like I was actively eavesdropping on their conversation, so I tried to keep my pace. My first guess is that they were talking about retail products. In which case they might have been looking at a retailer’s website that provides ways to sort items by price, star rating, whatever criterion the products can be sorted by. Or maybe they were talking about expenses from an expense report. Whatever “these” were, I think “yeah, probably” was exactly the right answer. Not “yes, absolutely!” but “yeah, probably.” Yes, A.I. can probably sort those items by dollar amount. Whether it actually will do that depends as much on a pseudorandom number generator as it does on the actual items being sorted. A.I. could sort the items by dollar amount or it could just as easily insert a few “hallucinated” items and fail to sort them into any discernible order. It could even break the relations between the columns, turning a straightforward expense report into one of those games in which you have to draw lines between columns. It would be much easier and more reliable to put those items into a spreadsheet, like Microsoft Excel or Google Sheets. And then not only can you get the items sorted by dollar amounts, you can do other things like reckon totals and averages, per diem and annualized costs, etc. Now, getting the items into the spreadsheet might be time-consuming and error-prone for you to do by hand. Maybe you could use your phone’s optical character recognition (OCR) program to scan the items and put them into the spreadsheet. OCR is a form of A.I., isn’t it? But please, A.I., don’t do anything other than transcribe the items into the spreadsheet. Second example: Headlines and lead lines in newspaper articles A newspaper has lots of articles. I can’t read all of those articles, nor do I want to. Some articles I want to read only partially. How do I select the articles I want to read in full? How do I select the articles I want to read partially? If only there was some kind of A.I. solution! Now it seems every USA Today article on the newspaper’s website has an A.I. overview provided by Taboola’s DeeperDive GenAI. Those overviews come with a little information button, which gives the following blurb: DeeperDive is a GenAI Answer Engine built for trusted publishers. It delivers clear answers and real-time content based on what people are reading now — showcasing various stories from the USA TODAY Network that highlight not just what happened, but why it matters. This is currently a BETA being quality assessed by human beings. Mistakes may occur. Please reference surfaced articles to validate AI summary The overviews seem to be accurate. But… the first line of each A.I. overview that I looked at feels kind of bland to me compared to the lead lines human reporters write. Professional reporters put a lot of effort into crafting punchy headlines. As students they also learned the inverted pyramid structure, in which the most important details are placed close to the beginning of the article and the less important details are placed later in the article, where they may be safely chopped off if the paper’s layout doesn’t have enough room for the whole article. Since I have the ability to read A.I.-generated summaries, I also have the ability to read headlines and lead lines, and to decide if an article just doesn’t interest me enough to read the whole thing. A.I. summaries of newspaper articles introduce a layer of uncertainty for a problem that real journalists had already devised solutions for decades ago. Third example: Just talk to someone Bok Creative posted on Instagram a new ad for Google Gemini in which a young woman arrives at her college dorm for the first time and worries that her style will clash with that of her roommate, who has already moved her things in, including bedding. I suppose that this is a genuine problem some young people have. Last time I had a roommate, the style had already been decided for us: the bedding, for example, featured drab green blankets. I believe he later attained a certain rank and was allowed to style his quarters more freely, but then also he no longer would have had a roommate. So anyway, the young lady in the ad whips out her phone and asks Google Gemini for advice. The roommate has “a clean girl vibe,” according to the young woman, who characterizes herself as a “maximalist.” Google Gemini saves the day by suggesting new bedding for the young woman to buy, and apparently the young woman had not already brought bedding in. But here’s a crazy thought: talk to the roommate. I’m not the first to have this crazy thought. YouTuber “One Minute Off” Abby also suggests in a recent video looking at that ad and a couple other Google Gemini ads that the young protagonist should just wait to have a chat with her new roommate before rushing off to buy more stuff. Who knows, maybe the roommate is also a “maximalist” who worried about her style being too much for her roommate. The open thread question: What other examples can you give of people using A.I. to do things where simpler, more reliable tools or methods already exist? [END] --- [1] Url: https://www.dailykos.com/stories/2026/9/10/800074825/community/open-thread-failures-of-natural-intelligence/ Published and (C) by Daily Kos Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified. via Magical.Fish Gopher News Feeds: gopher://magical.fish/1/feeds/news/dailykos/