Post ArzX5TeMOl5MGXK7d2 by Brynawel@mastodon.social
(DIR) More posts by Brynawel@mastodon.social
(DIR) Post #ArzKDm4sTBDcKZLwTA by stux@mstdn.social
2025-03-12T19:27:55Z
0 likes, 0 repeats
My personal stance on (generative) AI is rather simple: I do not want it.I am a creative person who like to create, build and design, AI feels like a salesman selling me my own work, no thank youuuOne of the very few exceptions would be for alt-text generation but even then it has to fulfill a whole checklistLet's keep the social in social media! :nkoHammer: š¤
(DIR) Post #ArzKLBKgyDbAgXHYLQ by WillyECoyote69@mstdn.social
2025-03-12T19:29:20Z
0 likes, 0 repeats
@stux Agree ! But i guess it will overcome, like the Internet did.
(DIR) Post #ArzKRXab8MQhRxkoq0 by thefwguy@techhub.social
2025-03-12T19:30:26Z
0 likes, 0 repeats
@stux I feel exactly the same.I do love AI, is a very powerful tool for many things, especially the edge AI and in the embedded world but I feel that replacing "creativity" and human interaction is just plain wrong.
(DIR) Post #ArzKZLOyKYiLy42Mue by WillyECoyote69@mstdn.social
2025-03-12T19:31:55Z
0 likes, 0 repeats
@stux Never Ever Let A.I. give some place in the Fediverse !
(DIR) Post #ArzKf8AWRZT2fD2OjQ by cgudrian@social.tchncs.de
2025-03-12T19:32:47Z
0 likes, 0 repeats
@stux Iāve read far too many bad AI generated alt texts to even remotely consider using it myself. These texts are read by humans and should be written by humans.
(DIR) Post #ArzKpquuvR7GLNqpEW by stux@mstdn.social
2025-03-12T19:34:49Z
0 likes, 0 repeats
@cgudrian I agree š im always adding alt-text but many still don't Some is better then nothing i guess, but like i said.. it's a loong checklist
(DIR) Post #ArzLCEKkXkuXL9bXYu by mirandawhiting@mastodon.social
2025-03-12T19:38:52Z
0 likes, 0 repeats
@stux Somebody said it might be useful for translation but I would still prefer not.
(DIR) Post #ArzLRNGHdEH9NDjjSy by spielUndSpass@mstdn.social
2025-03-12T19:41:36Z
0 likes, 0 repeats
@stux i see you point and i'll offer you a alternative one or rather extend on it:If i want help, then i don't care if i get the help from a human or an AI. I'd prefer not to bother a human actually. So for Alt-Texts, searches, creating images and music, advice etc i think it would be great, if it works. I'm not confident it works well enough though, so i don't use it.Getting social interactions or replacing your own creative output with it is a huge trap, though. I agree with you on that.
(DIR) Post #ArzLTxcpP5NWjTVzou by TripleIris@mstdn.social
2025-03-12T19:42:04Z
0 likes, 0 repeats
@stux I tried gen AI to generate subtitles for a video once. It was so full of errors and badly formatted that it would have been faster to just manually type them up myself.Anyway, I firmly believe the reason companies are backing AI so heavily is because they want to fully replace their human workforce (many already are doing just that) so I'm against it on principle, even if I didn't think it produced garbage results.
(DIR) Post #ArzLUwl66xEqKmqxg8 by martine@ioc.exchange
2025-03-12T19:42:18Z
0 likes, 0 repeats
@stux Iām creative & artistic but I do enjoy playing around with AI - sadly I think the ship has sailed as regards competing with human creativit
(DIR) Post #ArzNd5SFrZIrWu36HY by StephanieMoore@mastodon.online
2025-03-12T20:05:54Z
0 likes, 0 repeats
@stux would love to hear what might go on that checklist for you
(DIR) Post #ArzOUHtwJl3vToMmJM by Novlene@mastodon-belgium.be
2025-03-12T20:15:43Z
0 likes, 0 repeats
@stux I use it to for example to- Move faster through larger texts, to summarize them- Prepare something from an almost empty fridge- Create an empty template to start off with some structure (or restructure if I feel this is needed)Oops ... this comes off as me being lazy if I put it this way š³.But it does save a lot of time
(DIR) Post #ArzPa4NfbDbBEqar2G by nuche@mastodon.social
2025-03-12T20:27:57Z
0 likes, 0 repeats
@stux Mainstream AI is anecdotal. It doesn't replace creativity. It's simply a convenience tool.
(DIR) Post #ArzSS8DRkxCCg8iZpw by ncw413@mastodon.social
2025-03-12T21:00:06Z
0 likes, 0 repeats
@stux "a salesman selling me my own work" describes it well.
(DIR) Post #ArzTiIgOhxlJjFX7ku by keremgoart@mstdn.social
2025-03-12T21:14:15Z
0 likes, 0 repeats
@stux š
(DIR) Post #ArzX5TeMOl5MGXK7d2 by Brynawel@mastodon.social
2025-03-12T21:52:05Z
0 likes, 0 repeats
@stux I found genAI useful for exactly one purpose: dumbing down my technical texts into marketing shit "so we can use it on a LinkedIn-Post".
(DIR) Post #ArzdTEXKq9vE4FkTlg by Alisa385@mastodon.social
2025-03-12T23:03:39Z
0 likes, 0 repeats
@stux ā
Thanks for your interested!Here is more details if you interested click here & Apply Nowšā¬ļøis.gd/IiRMFR
(DIR) Post #ArzhH1SW2LQ9qivYNk by yknotclick@mastodon.social
2025-03-12T23:46:16Z
0 likes, 0 repeats
@stux i think this a very in line opinion on here, so iāll be brave and say that i agree with you on parts, but i donāt on others. i think generative art can never replace true art, because what makes real art real is the emotion put into it, and a lot of that comes from forming a connection with the artist. in my eyes, gen AI sucks the same way mainstream social sucks: people are using it for the wrong reasons and capitalistic gains. we shouldnāt be scared of it, just cautious!
(DIR) Post #ArzhacjXftyKxNEo6a by stux@mstdn.social
2025-03-12T23:49:51Z
0 likes, 0 repeats
@yknotclick Oh i totally agree š If i currently notice how much comes to creating good levels in games for example, im kinda sure AI will never get that down like hoomans do, also with feel and emotion
(DIR) Post #ArzhxGGhMl8FwUXb4S by yknotclick@mastodon.social
2025-03-12T23:53:56Z
0 likes, 0 repeats
@stux iām saying that even if they do, i think humanity will choose to pick art from actual artists every time in the future. the fact that the fediverse already existed pre-everything thatās happening right now is just proof of that in my eyes. evil might be winning this hand by a decent margin, but it always swings back eventually. i just hope im still around to see itš
(DIR) Post #Arzi7hdmS321jRJfNI by stux@mstdn.social
2025-03-12T23:55:50Z
0 likes, 0 repeats
@yknotclick Oh yes but aside from artwork etc, it's often kinda "playing" with the players emotion and knowledge to guide themWith puzzles for example! Giving the player the "idea" they got something all "figured out" while presenting them with a new obstacle or something in that way
(DIR) Post #ArzieuLCFkaWg3SOlk by yknotclick@mastodon.social
2025-03-13T00:01:50Z
0 likes, 0 repeats
@stux i also feel like not enough people are aware of LLMs. AI doesnāt have to be devastating to the environment and run on giant clouds, you can run a lot of these models on hardware most gamers already have. the key issue iāll note, if we (being more developed, peaceful(š¤Ø) nations) ease up on the development, it gives the other side more time to catch up. i donāt trust our side right now, but i do trust them just a liiiittle bit more
(DIR) Post #Arzjq2uOq2lqOCnX7I by divVerent@blob.cat
2025-03-13T00:15:08.446996Z
0 likes, 0 repeats
@stux @yknotclick Level design is outright impossible for AI. However, human _guided_ level design with AI, as in, you build the level structure and all the gameplay aspects, AI fills it out and adds details - that can kinda work, but AI tools right now are still rather coarse for that.What I'd love is if AI could fill out the details, but as a human I can then edit, move even the generated detail around, and AI then comes and makes everything consistent again. No text prompts, but a "cooperation" with an AI based editor.And yes, real art is better, but many game projects are one person projects and it's really hard to find artists that e.g. would help a small open source game, due to the licenses that tend to explicitly allow others to copy the art into their own game (provided credit is given and similar open source license).As an example, I'd have no trouble at all if someone used generative AI to make texture packs for a 3D game. That's the kind of stuff it actually works well for. Make me a seamless texture of <X>.Except... at least Gemini fails horribly at that. But I know a GenAI model could be _built_ for that... and from what I read, Stable Diffusion can do it.
(DIR) Post #Arzlrj1vuufPE0FZey by lalah@sakurajima.moe
2025-03-13T00:37:44Z
0 likes, 0 repeats
@stux the problem with AI is that the computers that create it are massive polluters, causing cancer and asthmahttps://news.ucr.edu/articles/2024/12/09/ais-deadly-air-pollution-toll
(DIR) Post #ArzndwhgTz73UDKttY by Cheeseness@mastodon.social
2025-03-13T00:57:41Z
0 likes, 0 repeats
@stux For alt-text, I think it could make sense as an accessibility tool for users, but for myself, to present the output of tools like that as part of my work/the effort I've put in feels misleading/inappropriate (if I'm not putting that effort in, I shouldn't try to claim that I have/feel entitled to the benefits)
(DIR) Post #Arzr588GOmgSc4QFns by skweetis@mstdn.social
2025-03-13T01:36:10Z
0 likes, 0 repeats
@stux I don't want it personally. I have tried to think of an ethical use case and I considered maybe if it helped people with language difficulties write to their elected representatives. Would need guardrails and I'm still not sure.
(DIR) Post #As00UucABARTnymCOW by yknotclick@mastodon.social
2025-03-13T01:17:38Z
0 likes, 1 repeats
@divVerent @stux as with many things, my understanding is that ,with AI, the home grown stuff is always better. even if youāre running a local model that youāve slowly trained yourself, itāll be trained to your specific use case. iād say probably 90% of āAIā is just built on chatgpt. but things like stable diffusion and comfy tend to perform better than any out of the box solution you could ever buy
(DIR) Post #As01I30FmNam0lTRpo by divVerent@blob.cat
2025-03-13T03:30:41.231895Z
0 likes, 0 repeats
@yknotclick @stux But anyway, to be clear - I see AI as very useful for the "menial" tasks. You know, you take a photo of asphalt of the road in front of your home, and then an AI model can turn this into a really good seamless texture of asphalt - way better than the current approaches of blending, mirroring or resynthesizing (in Photoshop I think this is called healing, so e.g. you take photo of asphalt, duplicate to a 4x4 pattern, paint heal brush over the seams, cut out the middle, and repeat this step a second time for any remaining seams). There is no art to that. And as such, AI is qualified to help.But for that to be really useful, we do not need generating pictures from a prompt - that is fun, but not very useful in the long run, except maybe to get a starting point.What we need is assisted editing. Take a photo of your home, and on the screen drag the furniture around another way. Type text and it automatically gets reformulated as you type, but you can make corrections or changes at any time, to which it will in turn respond.In the longer term I can even see things happen like turning 2D pictures into really good 3D models and textures of the object. But from that we are still far away.
(DIR) Post #As01uuQRbWvahbFsbw by untakenusername@infosec.exchange
2025-03-13T03:37:30Z
0 likes, 0 repeats
@stux I think it does have its uses - as you said, for alt text generation - but there's a ton of working being done on using machine learning models to create new drugs and new algorithmsthose new algorithms are super specific to multiplying complex mathimatical objects but that probably has some practice application down the roadML drug thingy: https://alphafold.com/ML math Algo thingy: https://deepmind.google/discover/blog/discovering-novel-algorithms-with-alphatensor/
(DIR) Post #As0CQDvW9GT1hdMNpg by iveyline@mastodon.world
2025-03-13T05:35:18Z
0 likes, 0 repeats
@stux 100% support your view. Who wants to become an AI zombie?
(DIR) Post #As0udSmmbrxJXFOjfE by yknotclick@mastodon.social
2025-03-13T07:32:41Z
0 likes, 0 repeats
@divVerent @stux you can do a lot of what youāre talking about already with standard diffusion and comfy! the same way the open source community is pioneering social over here, theyāre doing it with AI. thereās things like diffusion models that let you rig models youāve drawn and in-painting to only re-gen what you want. to be clear, less trying to convince AI ā Bad, but more so that thereās more options than just cGPT and it doesnāt HAVE to be AI = bad.
(DIR) Post #As0udTh9EJ2gM3lkWG by divVerent@blob.cat
2025-03-13T13:50:48.074720Z
0 likes, 0 repeats
@yknotclick @stux Yeah, my point is rather, right now these tools are rather separate from regular editing workflows.When will there be a plugin to do AI assisted image editing in #Krita or #GIMP? Why not even in #Kdenlive?Because the interesting part is - yes, when working at a low level with these APIs directly, there are things possible that aren't really well exposed yet. Even if I wanted to deal with all that mess, to e.g. "heal" part of an image with AI, I'd have to first export the image with the to be replaced part cleared in the alpha channel, then run it through a separate tool, and then merge it back. This is the integration overhead that has to be removed. And this kind of stuff is what people really need.Not a clippy that can talk about politics.
(DIR) Post #As0vBNysyqCcTmEazw by divVerent@blob.cat
2025-03-13T13:56:57.470377Z
0 likes, 0 repeats
@yknotclick @stux I should add: at work we have some AI integration in the coding IDE.What doesn't work is all the prompt based stuff. It is awfully slow and never produces a good result anyway.What is REALLY good though is ML based code autocompletion. It's really good at guessing what I want to do based on a few letters. And best of all, it does all the boilerplate for me so I can focus on the logic (which it most definitely cannot guess).This is what I imagine _good_ AI integration to be. Draw an awful scribble and AI makes it look good, while letting you make changes and it will in turn perform these changes in a better way. All happening in realtime. Or carve out rooms in a 3D editor and AI will decorate the walls for you. Cut a hole for a door, and it'll make sure the doorway has trims. Place three lights across the ceiling and it'll figure out the pattern you intend for placing the others.Because that gets rid of the menial tasks and lets the user focus on the creative part of the work.
(DIR) Post #As1HZsL3xcjkZcgrIW by yknotclick@mastodon.social
2025-03-13T15:32:46Z
0 likes, 0 repeats
@divVerent @stux i think this functionality should exist in GIMP with a plugin, i know it does for photoshop. it does all the alpha compiling for you in the background basically. in SD, thereās a window just for inpainting where you can put in an image of you holding a paper towel roll, brush over the roll, and have it make it a sword or something else. then, you can run that picture through another diffuser to adjust the posing of the character in the picture
(DIR) Post #As1HZtJKLYwVaWszEO by yknotclick@mastodon.social
2025-03-13T15:33:57Z
0 likes, 0 repeats
@divVerent @stux ill also admit that i donāt use GIMP, so i donāt know for certain. but all the stuff about standard diffusion, i have actual experience doing with local models and the results are mind blowing tbh. iād definitely do some digging on github, thereās a LOT of cool open source AI stuff happening every day
(DIR) Post #As1HZtpaPckNCZykVs by divVerent@blob.cat
2025-03-13T18:07:52.663876Z
0 likes, 0 repeats
@yknotclick @stux So it's open source, but for Photoshop only :)Sadly I only run open source, so I can't use that then. Kinda sad.