Conversation

Notices

1. Embed this notice
   kaia (kaia@pleroma.soykaf.com)'s status on Friday, 01-May-2026 17:19:34 JST kaia
   like lain pointed out, fedi is where the same people who boost their ebooks accounts are vehemently anti-AI
   • Embed this notice
     SuperDicq (superdicq@minidisc.tokyo)'s status on Friday, 01-May-2026 17:19:33 JST SuperDicq

     in reply to

     @kaia@pleroma.soykaf.com That's a really terrible Strawman. If you think that you have absolutely no idea what the problem with "AI" is.
     
     Hosting a local free software implementation of a markov chain algorithm is absolutely not comparable in any way to using proprietary LLMs through an shitty API that the user has zero control over.

   • Embed this notice
     SuperDicq (superdicq@minidisc.tokyo)'s status on Friday, 01-May-2026 17:21:43 JST SuperDicq

     in reply to

     @kaia@pleroma.soykaf.com Most people are not actually "Anti AI". AI is just a technology. They are only "Anti AI" as in they hate big corpos who try to take control over our tech.

   • Embed this notice
     SuperDicq (superdicq@minidisc.tokyo)'s status on Friday, 01-May-2026 17:26:49 JST SuperDicq

     in reply to

     @kaia@pleroma.soykaf.com These local models are still proprietary and you have no control over them.
     
     And most likely the training data violates licenses of the material it is trained on.
     
     I'm sure if someone made a local model that runs using free software and is trained exclusively on data that they actually have permission for, nobody would can reasonably be opposed to that.

   • Embed this notice
     kaia (kaia@pleroma.soykaf.com)'s status on Friday, 01-May-2026 17:26:50 JST kaia

     in reply to

     • SuperDicq
     @SuperDicq the same people still get angry when I run a local model on my 4090, or post AI art generated locally. it's evidently not about the proprietary-ness.
   • Embed this notice
     SuperDicq (superdicq@minidisc.tokyo)'s status on Friday, 01-May-2026 17:28:33 JST SuperDicq

     in reply to

     • Phantasm

     @phnt@fluffytail.org @kaia@pleroma.soykaf.com Neo-luddites are fucking stupid. I don't take their arguments against technology seriously.
     
     Instead of trying to fix issues they just want to burn everything down and go back to the stone age like they have just completely given up. It's honestly depressing.

   • Embed this notice
     Phantasm (phnt@fluffytail.org)'s status on Friday, 01-May-2026 17:28:34 JST Phantasm

     in reply to

     • SuperDicq

     @kaia @SuperDicq It's about the fear of getting outpaced by a program. The same people that yell abolish copyright and claim to be anarchists now really want copyright to destroy their perceived evil.

     Imagine getting mad over what other people use.

   • Embed this notice
     SuperDicq (superdicq@minidisc.tokyo)'s status on Friday, 01-May-2026 18:03:50 JST SuperDicq

     in reply to

     • Phantasm

     @phnt@fluffytail.org @kaia@pleroma.soykaf.com A new emerging technology allows you to do less manual work, isn't that is a good thing?
     
     Why would anyone ever complain about having less work to do?
     
     Oh right, society still forces you to work even when there's no work available because otherwise you're not allowed to live in a house and eat food.
     
     How about we fix that problem by introducing something like basic income or whatever instead of blaming it on technology?

   • Embed this notice
     SuperDicq (superdicq@minidisc.tokyo)'s status on Friday, 01-May-2026 20:34:05 JST SuperDicq

     in reply to

     • Cheetah Meld

     @pingviini@pleroma.shunderdo.me @kaia@pleroma.soykaf.com Just because a model has a free license does mean that this is actually valid. If it is trained on data that is not compatible with that license, the license is probably invalid because it might be derivative work.
     The software used for inferencing and (additional) training with these models is also freely licensedNeither ROCm nor CUDA are free software. And neither is the firmware of the GPUs required to run these.
     you can look at the binary, you can patch it, you can disassemble it, you can run it in a debugger,Of course you can, but that's not the same as having the source code. Reverse engineering an entire program takes a lot of time and is often an unreasonable things to do especially if you want to make small changes. This obviously gives the person who has the source code an astronomically huge advantage, which means they have unjust control over you. (that's what I mean with you are not really in control)
     Low-Rank Adaptations for these models using small datasets on consumer hardware to adapt and modify the output of the modelYes, there are methods to retrain existing models, but it's not the same because you do not have the same power that the original developer. Just like knowing how to use a disassembler is not equivalent to having the program's source code...

   • Embed this notice
     Cheetah Meld (pingviini@pleroma.shunderdo.me)'s status on Friday, 01-May-2026 20:34:07 JST Cheetah Meld

     in reply to

     • SuperDicq
     @SuperDicq @kaia
     
     > These local models are still proprietary and you have no control over them.
     
     Simply not true. Many many models out there (and not just academic, irrelevant models, I mean models that people actually use) are licensed freely (Apache 2.0 or MIT are popular licenses) and that includes the parameters, i.e. all those numbers in the gigabytes-large .safetensors file that you download and then stuff into your GPU's VRAM. The software used for inferencing and (additional) training with these models is also freely licensed.
     
     What's true is that, more often (but not always!) the training data used to produce those parameters is proprietary and/or undisclosed. The computing power required to train a model *completely from scratch* is also prohibitively time-consuming and/or expensive for any single user with consumer hardware.
     
     So it's easy to assume that the big blob file used for inferencing is completely opaque and immutable, regardless of licensing, and effectively people are still using a black box, but that's not true - just like an executable binary program isn't really a black box. It's on your computer, you can look at the binary, you can patch it, you can disassemble it, you can run it in a debugger, you can run it in a sandbox, you can inject your own code into it, etc.
     
     You can do a lot of similar things with models on your computer, like training Low-Rank Adaptations for these models using small datasets on consumer hardware to adapt and modify the output of the model, or modify the parameters directly to modify the output of a model. People routinely do both of these things to teach models new tricks (example: Generate images of things that the base model does not know about) or to make the model do things the original creator trained it not to do (example: Do ERP or use politically incorrect language).