fretmeister Posted February 18 Posted February 18 (edited) I asked an AI to do me an image of 2 of my favourite bass players: Duck Dunn and Bootsy Collins. Instead of generating them separately the AI went all BrundleFly... Edited February 18 by fretmeister Spelling! 16 Quote
Happy Jack Posted February 18 Posted February 18 4 minutes ago, fretmeister said: I do find it puzzling that the AI is capable of doing so much in these images, but then hits a brick wall when it comes to things like the number of strings / tuners / pegs etc. 6 2 Quote
chris_b Posted February 18 Posted February 18 The Delta plane in Toronto was using AI auto pilot to land. 1 Quote
BigRedX Posted February 18 Posted February 18 23 minutes ago, Happy Jack said: I do find it puzzling that the AI is capable of doing so much in these images, but then hits a brick wall when it comes to things like the number of strings / tuners / pegs etc. AI appears to very stupid when it comes to many things that human beings can see and easily understand. I was recently asked to use several tiny photographs of buildings in a publication that needed to appear at full page width which involved enlarging them by over 400%. I have several AI-based tools for image enlargement that were initially bought to use with headshots for contributors (for another publication) that had been taken from their organisation's website or their LinkedIn profile. For this task it is very good indeed. Unfortunately in the case of the photographs of the buildings it rendered all the regular brickwork as something between avant-garde architecture and dry stone walling. As a human being I can instantly look at the photographs and recognise that those buildings are made of bricks that need to be arranged in straight rows. AI has yet to be able make that deduction. 1 Quote
fretmeister Posted February 18 Author Posted February 18 11 minutes ago, SICbass said: I foresee a flurry of new BC profile pictures. Duck Dunn and Cliff Burton.... 1 Quote
Cato Posted February 18 Posted February 18 (edited) As I understand it Generative AI works by scanning a source ( usually the internet but it can be a specific database) for information/ images that fit the keywords in the description it's provided and combining information from those sources to create a new image. At no point does it actually understand any of that information itself. ie. It doesn't know what a bass guitar is, let alone how it's components work, hence the obvious innacuracies in many of the images. Edited February 18 by Cato 2 Quote
ezbass Posted February 18 Posted February 18 35 minutes ago, Happy Jack said: I do find it puzzling that the AI is capable of doing so much in these images, but then hits a brick wall when it comes to things like the number of strings / tuners / pegs etc. The strings are never straight either, the easiest thing to do, you would have thought. Quote
tauzero Posted February 18 Posted February 18 37 minutes ago, ezbass said: The strings are never straight either, the easiest thing to do, you would have thought. In the case of the second picture in the OP, they can even merge. Quote
MacDaddy Posted February 18 Posted February 18 The fail is possibly in the prompt and not the output. GIGO. Quote
MacDaddy Posted February 18 Posted February 18 This is my first attempt, could probably get it better if I spent more time on it. I used the free version of DeepAI, this was the prompt; Highly detailed illustration of two legendary bass players, Bootsy Collins and Donald 'Duck' Dunn, standing side by side on a stage. Bootsy Collins is wearing a vibrant, glittery outfit with his iconic star-shaped sunglasses, a wide-brimmed hat, and holding a futuristic, star-themed bass guitar. Donald 'Duck' Dunn is depicted in a casual 1970s outfit: a button-up shirt, jeans, and a pipe in his mouth, holding a classic Fender Precision Bass. The scene includes realistic lighting with stage spotlights and a subtle audience in the background. Focus on accurate facial features and distinctive styles for both musicians. Quote
Dood Posted February 18 Posted February 18 16 minutes ago, MacDaddy said: The fail is possibly in the prompt and not the output. GIGO. A basic prompt that I have used has been over 800 "words" even before I have asked a question or fed it any details to work from. Quote
prowla Posted February 18 Posted February 18 I think AI is rubbish; I commented today that it's a bit like "Military Intelligence": two words which don't belong in the same sentence. Quote
ezbass Posted February 18 Posted February 18 6 minutes ago, MacDaddy said: This is my first attempt, could probably get it better if I spent more time on it. I used the free version of DeepAI, this was the prompt; Highly detailed illustration of two legendary bass players, Bootsy Collins and Donald 'Duck' Dunn, standing side by side on a stage. Bootsy Collins is wearing a vibrant, glittery outfit with his iconic star-shaped sunglasses, a wide-brimmed hat, and holding a futuristic, star-themed bass guitar. Donald 'Duck' Dunn is depicted in a casual 1970s outfit: a button-up shirt, jeans, and a pipe in his mouth, holding a classic Fender Precision Bass. The scene includes realistic lighting with stage spotlights and a subtle audience in the background. Focus on accurate facial features and distinctive styles for both musicians. At last I understand why Duck Dunn was so awesome, check that right hand . 1 Quote
Happy Jack Posted February 18 Posted February 18 1 hour ago, MacDaddy said: This is my first attempt, could probably get it better if I spent more time on it. I used the free version of DeepAI, this was the prompt; Highly detailed illustration of two legendary bass players, Bootsy Collins and Donald 'Duck' Dunn, standing side by side on a stage. Bootsy Collins is wearing a vibrant, glittery outfit with his iconic star-shaped sunglasses, a wide-brimmed hat, and holding a futuristic, star-themed bass guitar. Donald 'Duck' Dunn is depicted in a casual 1970s outfit: a button-up shirt, jeans, and a pipe in his mouth, holding a classic Fender Precision Bass. The scene includes realistic lighting with stage spotlights and a subtle audience in the background. Focus on accurate facial features and distinctive styles for both musicians. And as you can see, AI has managed to capture the very essence of Duck there. Oh no ... wait ... 1 Quote
Burns-bass Posted February 18 Posted February 18 Ask it to create you an image of a glass of wine full to the brim. Literally doesn’t understand it as there’s no reference to go on. Quote
MacDaddy Posted February 18 Posted February 18 1 hour ago, Happy Jack said: And as you can see, AI has managed to capture the very essence of Duck there. Oh no ... wait ... If the sh1t fits... Quote
Norris Posted February 19 Posted February 19 At Christmas I generated a few themed Teams backgrounds for online meetings. In one of them a guitar (being played by a Christmas elf) had two necks - at opposite ends of the body 😂 Quote
tauzero Posted February 19 Posted February 19 2 hours ago, Norris said: At Christmas I generated a few themed Teams backgrounds for online meetings. In one of them a guitar (being played by a Christmas elf) had two necks - at opposite ends of the body 😂 And? 1 Quote
Downunderwonder Posted February 19 Posted February 19 16 hours ago, Cato said: As I understand it Generative AI works by scanning a source ( usually the internet but it can be a specific database) for information/ images that fit the keywords in the description it's provided and combining information from those sources to create a new image. At no point does it actually understand any of that information itself. ie. It doesn't know what a bass guitar is, let alone how it's components work, hence the obvious innacuracies in many of the images. Can you explain how it comes up with so many three banger Fenders? Quote
zbd1960 Posted February 19 Posted February 19 As a now retired IT guy I get frustrated with the use of the term "AI". The media, inevitably, misuses the term, and PR/marketing departments hype it death. In most cases, AI is a series of algorithms backed up by a large database and a rule set. It is not 'intelligent' it's just quick at searching. We've got the same issue with AI that we've had with previous technology shifts. In the 50s everything was 'atomic' or 'rocket', in the 60s we start to see "Twenty-first century", in the 80s everything had to be labelled 'digital', in the late 90s it was "e-...." and then with Apple it became "i-....". Now it's "AI". The stuff it's good at is pattern recognition and it's going to be a big help with processing scans for cancers etc. The big area of concern at the moment is no-one has worked out how or what needs regulating about it. As with any technology there is potential for harm as well as good. The issue at the moment is 'deep fakes'. 3 Quote
42Hz Posted February 19 Posted February 19 (edited) 19 hours ago, fretmeister said: Duck Dunn and Cliff Burton.... The new Escher bass.😀 Edited February 19 by 42Hz Quote
prowla Posted February 19 Posted February 19 2 hours ago, Downunderwonder said: Can you explain how it comes up with so many three banger Fenders? It doesn't see strings - it sees lines, so the binding might confuse it. Quote
BigRedX Posted February 19 Posted February 19 9 minutes ago, prowla said: It doesn't see strings - it sees lines, so the binding might confuse it. Most AI can't even see lines properly as my building brickwork example proved. 1 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.