Paid Members Jazzy G Posted August 14, 2025 Paid Members Posted August 14, 2025 I don't know whether this makes me a traitor or not, but when Vicky and I were out with her friend the other weekend we used the "add me" setting on my phone. I took a picture of Vicky, her friend and the dog posing between these cardboard cutouts of Wallace & Gromit in Sherwood Forest, then my phone but a border around the edges of the photo I'd just taken. I went and stood next to Wallace and Vicky took the photo, after which the phone managed to insert me into the picture. I was highly impressed. I like these kinds of gimmicks for what I presume is some kind of AI, but I don't think it'll be all that good at creating original pieces. Although having said that some of the stuff I like to listen to sounds like somebody connecting to the Internet 30 years ago with a beat, so what do I know? Quote
Paid Members JNLister Posted August 29, 2025 Paid Members Posted August 29, 2025 https://arstechnica.com/ai/2025/08/why-its-a-mistake-to-ask-chatbots-about-their-mistakes/ is a useful reminder of how LLMs work and how easy it is to get fooled into thinking differently. Quote
The Gaffer Posted February 20 Posted February 20 Diverging from the current Wrestling Books thread discussion on Bret Hart perhaps not being the Excellent of Elocution, I had a rubbish AI encounter in my dispassionate day job this past Wednesday. Essentially, a rather well-meaning, eager pleaser on my small team of people has been working on another AI model to add to the company's dizzying arsenal of models they now use for just about everything. Our lead's delighted with this as the company philosophy (to their credit) has always been to allow for free range side products etc but (as one might imagine) in recent years this has become much, much more "We expect you to start working with AI in these areas" themed. My team essentially has a lot of oversight into maintaining, editing, improving etc the public help centre documentation for the product we support. This new model is essentially a way of searching through that documentation so that we can find what we need, swarm it and simply pull knowledge from it or identify areas to improve. Great, only: - The actual help centre documentation as it is already provides a way to do this through a decent search function which predates. It's literally the job of what it is. - Both the existing search function and just using Google to search keywords - like the most web naive, tech phobic person at your library - are demonstrably faster than this new tool. The model requires a bunch of cookies, access tokens etc to even get up and running. You then need to partake in its training and up-skilling as it constantly shits the bed in pursuit of being able to manage a task most people with internet access have been able to do with minimal fuss for years. We awkwardly sat through a 45 minute team meeting where the person who put this together fumbled their way through walking us through the setup of this. A few kind souls - the sort who still react to AI output like a magician doing a trick - did the clap emoji a few times on the call. Our lead's thrilled. They want us actively thinking about how we can use this going forward. It's a collective madness. Fuck allllll the way off. Quote
Paid Members LaGoosh Posted February 20 Paid Members Posted February 20 My works internal LLM just added image generator and everyone is very excited. Calling each other over to each other's desks to look and laugh at the latest boring unfunny shit they've generated. Quote
DavidB6937 Posted February 20 Posted February 20 And of course the latest craze of asking it to create a picture of you and your life based on everything it knows about you. Oh brilliant! Why are we thinking that's a good idea?! Quote
Paid Members Lorne Malvo Posted February 20 Paid Members Posted February 20 (edited) I used it for the first time ever a few weeks back, out of boredom and curiosity. To be honest, I used it in the hope it would fail so I could show my AI-friendly workmates how useless it is, but it even exceeded my expectations. I asked it 'can you tell me the nearest showtimes for The Chronology Of Water?' It replied 'no, but you can find local cinema listings on Google.' Fucking rubbish! Edited February 20 by Lorne Malvo Quote
Paid Members LaGoosh Posted February 20 Paid Members Posted February 20 5 minutes ago, DavidB6937 said: And of course the latest craze of asking it to create a picture of you and your life based on everything it knows about you. Oh brilliant! Why are we thinking that's a good idea?! At the pub last night one of my colleagues just kept off handedly referring to "ChatGPT" as "Chat" and it quickly became clear this woman runs every aspect of her life through ChatGPT including health advice and her diet. She also lives alone with her dog so maybe she's just lonely I dunno'. Quote
DavidB6937 Posted February 20 Posted February 20 Just now, LaGoosh said: At the pub last night one of my colleagues just kept off handedly referring to "ChatGPT" as "Chat" and it quickly became clear this woman runs every aspect of her life through ChatGPT including health advice and her diet. She also lives alone with her dog so maybe she's just lonely I dunno'. A surprising amount of people do that. It's something I've been learning more about during mental health and Samaritans training. A lot of it is loneliness or not wanting to talk to anyone 'real' about certain things, but at the same time it's a massively dangerous thing to use in certain circumstances. I mean we all know whatever symptoms you type into Google you end up with death anyway so this is no different. Quote
Paid Members tiger_rick Posted February 20 Paid Members Posted February 20 1 hour ago, The Gaffer said: did the clap emoji a few times on the call. Get that in the minor annoyances thread. Only thing worse is people sending hearts. Quote
Paid Members BomberPat Posted February 20 Paid Members Posted February 20 1 hour ago, DavidB6937 said: And of course the latest craze of asking it to create a picture of you and your life based on everything it knows about you. Oh brilliant! Why are we thinking that's a good idea?! The Venn diagram of the people who post this shit, and the people who ten years ago posted a bullshit disclaimer telling Facebook they didn't consent to them using their data, is a flat circle. It's maddening. 1 hour ago, Lorne Malvo said: I used it for the first time ever a few weeks back, out of boredom and curiosity. To be honest, I used it in the hope it would fail so I could show my AI-friendly workmates how useless it is, but it even exceeded my expectations. I asked it 'can you tell me the nearest showtimes for The Chronology Of Water?' It replied 'no, but you can find local cinema listings on Google.' Fucking rubbish! People really do need to recognise the sheer uselessness of it. I went to an LGBTQ+ local history talk earlier this week, and they talked about the Drill Hall theatre doing a lot for gay rights and queer representation in the arts, so during the Q&A somebody asked the speaker (a bloke in his 70s, who's a retired history lecturer and a London tour guide, who has spent years working on this stuff) if there was a book specifically about that theatre. The speaker said that he wasn't aware of one, and then a couple of minutes later this obnoxious bloke in the front row spoke over to him to say that actually, yes, there was a book, and it was called whatever. The name of the book rang alarm bells for me, as to how unwieldy the language was, and I looked it up after the event was over and, sure enough, it doesn't exist. He just asked ChatGPT if a book existed, and because it would always rather give you an answer than not, it just made one up. Who does that help? What purpose does that serve? Similarly, a mate of mine has been working on a book, and he liked his opening line so much that he wanted to double-check that he hadn't inadvertently stolen it from somewhere. So he Googled it, and the AI response told him that it was a line from a specific poem. He looked up that poem, and the line didn't appear anywhere in it. He replied to the AI pointing that out, and it "apologised", and said that actually it's not from the poem, but it's a popular interpretation of the poem. So he looked up interpretations of the poem, and that was bollocks too! Something I find instructive when someone is adamant that this shit works is to "ask" Google the same question multiple times in a row. People who think it works never do this, because they trust its output, but when you do it, it often gives completely different results each time, because it's not actually answering the question, it's scraping data to create something that looks like an answer to that question, and doing it afresh every single time. I was on a podcast where they pitted me against Google AI with questions about old wrestling shows, and it was just hallucinating whole matches and angles that never happened, and the three of us on the call could ask it the same question at the same time and get completely different answers. Quote
Chili Posted February 20 Posted February 20 This morning I saw online a comedian (who is fucking minted through their club and touring) posting about an event they run using AI. The fact that this person works in the arts, isn't scraping for a few quid, but will not get a designer in to pay to make their poster because it's easier to slop it through AI just really irritated me. Quote
Paid Members BomberPat Posted February 20 Paid Members Posted February 20 Yeah, a friend of mine is a really good musician and songwriter, but all of his album art and social media posts are AI generated images. He's really fascinated by what it's capable of, and I can get that, but I don't see how he doesn't recognise that the images look shit, and make him look shit. He did music for a friends' podcast, and every time I've spoken to them they got increasingly concerned that he was using it on the music as well, because he was coming up with new tracks so quickly. I don't think that he was, but even with people pointing it out to him, it's never occurred to him that using AI for the imagery would make people think that he was using it on his music as well - to him, it's just freeing up time for him to make more music because he doesn't have to use that time creating the art. Even though two of the people he's been making music for are professional graphic designers, and one of them pointed out to him that when he was tagged in an Instagram post of this guy's AI artwork, he lost a measurable number of followers - presumably because they assumed that he was using AI for his graphic design work, and wanted no part of it. That's what people aren't seeing - the time and money you save by using AI for your graphics is never going to offset the potential business you lose by people who see AI art and naturally assume that you're lazy with no sense of quality control, and assume the amount of effort you put into that graphic reflects the effort you put into whatever it is you're selling. Quote
TibBo Posted February 20 Posted February 20 It gives people with no creative ability the ability to be creative. Problem is, as they have no creative ability/training/knowledge, they don’t understand why some AI generated crap doesn’t work from a design perspective. Much like how they didn’t understand why something did work when it was designed by a human in the first place… I’m sure AI could have worded this better, but if it had created an image it would have given it 16 fingers Quote
Paid Members Devon Malcolm Posted February 20 Paid Members Posted February 20 I think we need to rename this thread, lads. Quote
Members SuperBacon Posted February 20 Members Posted February 20 5 minutes ago, Devon Malcolm said: I think we need to rename this thread, lads. I think it's quite apt as it's the sort of thing ChatGPT would come up with Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.