Comment on idk

• Alaknar@sopuli.xyz ⁨1⁩ ⁨day⁩ ago

  I don’t get why you have to go to such extremes here.

  AI is an extremely broad spectrum of tools. Some of them, yes, use stolen graphics to generate derivative graphics. Some of them attempt writing code.

  But others let you create things that would normally require hundreds of thousands of dollars while still retaining the necessary creative input from the author.

  If you are against such tools as the one used in the linked video, I think you should also stand very much against Photoshop allowing people to paint without using actual pigments and oil.

  source

  • Senal@programming.dev ⁨8⁩ ⁨hours⁩ ago

    Weak comparisons help no-one, photoshop is nothing like LLM’s

    All of the big commercial LLM’s (without exception afaik) have been trained on a large corpus of data that has been obtained by various sketchy and illegitimate means. (some legitimate as well).

    That’s the major difference between the two.

    If you are using a model that has only been trained on legally obtained data, disregard this point.

    I’m not even against competent tool use of LLM’s but please use better arguments.

    source

    • Alaknar@sopuli.xyz ⁨3⁩ ⁨hours⁩ ago

      Weak comparisons help no-one, photoshop is nothing like LLM’s

      Then people need to specify that they’re against generative LLMs, like Chat-bots or slop-generators, not “all AI”.

      There was just a thread on Twitter where a company showcased an amazing tool for animators - where you, for example, prepare your walking/sitting/standing animations, but then instead of motion-capturing or manually setting the scene up, you just define two keyframes - the starting and the ending position of the character… and then their AI picks the appropriate animations, merges between them and animates the character walking from one position to the other.

      It’s a phenomenal tool for creatives, but because the term “AI” appeared, the company got shat on by random people.

      All of the big commercial LLM’s (without exception afaik) have been trained on a large corpus of data that has been obtained by various sketchy and illegitimate means

      No. All generative graphical slop AIs and generic chat-bot LLMs have been trained on large corpus of data that has been obtained by various sketchy and illegitimate means.

      THAT’S the major difference.

      If you are using a model that has only been trained on legally obtained data, disregard this point.

      I’m not even against competent tool use of LLM’s but please use better arguments.

      And yet, the guy I was responding to wrote:

      So I’m gonna execute the code of someone who doesn’t know the first thing about coding on my computer? Great!

      I’d rather have AI art and human code.

      So, he basically says something that directly contradicts what you’re saying - he prefers the generative slop machines, than tools that actually help developers or artists.

      source
  • Potatar@lemmy.world ⁨1⁩ ⁨day⁩ ago

    Art cannot destroy my system. Bad code which might ask for elevated access can…

    source