Those claiming AI training on copyrighted works is “theft” misunderstand key aspects of copyright law and AI technology. Copyright protects specific expressions of ideas, not the ideas themselves. When AI systems ingest copyrighted works, they’re extracting general patterns and concepts - the “Bob Dylan-ness” or “Hemingway-ness” - not copying specific text or images.

This process is akin to how humans learn by reading widely and absorbing styles and techniques, rather than memorizing and reproducing exact passages. The AI discards the original text, keeping only abstract representations in “vector space”. When generating new content, the AI isn’t recreating copyrighted works, but producing new expressions inspired by the concepts it’s learned.

This is fundamentally different from copying a book or song. It’s more like the long-standing artistic tradition of being influenced by others’ work. The law has always recognized that ideas themselves can’t be owned - only particular expressions of them.

Moreover, there’s precedent for this kind of use being considered “transformative” and thus fair use. The Google Books project, which scanned millions of books to create a searchable index, was ruled legal despite protests from authors and publishers. AI training is arguably even more transformative.

While it’s understandable that creators feel uneasy about this new technology, labeling it “theft” is both legally and technically inaccurate. We may need new ways to support and compensate creators in the AI age, but that doesn’t make the current use of copyrighted works for AI training illegal or unethical.

For those interested, this argument is nicely laid out by Damien Riehl in FLOSS Weekly episode 744. https://twit.tv/shows/floss-weekly/episodes/744

  • CeeBee_Eh@lemmy.worldEnglish
    45·
    2 months ago

    Like fuck it is. An LLM “learns” by memorization and by breaking down training data into their component tokens, then calculating the weight between these tokens.

    But this is, at a very basic fundamental level, how biological brains learn. It’s not the whole story, but it is a part of it.

    there’s no actual intelligence, just really, really fancy fuzzy math.

    You mean sapience or consciousness. Or you could say “human-level intelligence”. But LLM’s by definition have real “actual” intelligence, just not a lot of it.

    Edit for the lowest common denominator: I’m suggesting a more accurate way of phrasing the sentence, such as “there’s no actual sapience” or “there’s no actual consciousness”. /end-edit

    an LLM would learn “2+2 = 4” by ingesting tens or hundreds of thousands of instances of the string “2+2 = 4” and calculating a strong relationship between the tokens “2+2,” “=,” and “4,”

    This isn’t true. At all. There are math specific benchmarks made by experts to specifically test the problem solving and domain specific capabilities of LLM’s. And you can be sure they aren’t “what’s 2 + 2?”

    I’m not here to make any claims about the ethics or legality of the training. All I’m commenting on is the science behind LLM’s.

    • Eccitaze@yiffit.netEnglish
      34·
      2 months ago

      Get a load of this maroon, they think LLMs are actually sapient! Thanks, I needed that laugh.

      • CeeBee_Eh@lemmy.worldEnglish
        42·
        2 months ago

        Get a load of this maroon, they think LLMs are actually sapient!

        I guess reading comprehension is as bad here as it’s ever been on the internet.

        • Eccitaze@yiffit.netEnglish
          22·
          2 months ago

          Fine, you win, I misunderstood. I still disagree with your actual point, however. To me, Intelligence implies the ability to learn in real-time, to adapt to changes in circumstance, and for self-improvement. Once an LLM is trained, it is static and unchanging until you re-train it with new data and update the model. Even if you strip out the sapience/consciousness-related stuff like the ability to think critically about a scenario, proactively make decisions, etc., an LLM is only capable of regurgitating facts and responding to its immediate input. By design, any “learning” it can do is forgotten the instant the session ends.

          • CeeBee_Eh@lemmy.worldEnglish
            11·
            2 months ago

            Fine, you win, I misunderstood.

            It’s not a competition, but I genuinely respect you for saying you misunderstood.

            Once an LLM is trained, it is static and unchanging until you re-train it with new data and update the model.

            Absolutely! I honestly think this is the main thing (or at least one of the main things) that prevent human-level intelligence or even sentience in LLM’s.

            Think about how our minds work. From the moment we’re born (really, it’s way before that) our brains are bombarded with input and feedback from every sense. It takes a person many months of that to start recognizing things. That’s also why babies sleep so much, their brains are kinda “training” and growing fast. Organizing all the data into memories.

            Side bar: this is actually what dreams are. Dreams are emotions, thoughts, ideas, or whatever concept a neuron or group of neurons are associated with getting triggered. When we dream it’s our brain taking the days inputs and building new connections. The neural connections in our brains are very much like weights and feed-forward process of neural activation is near identical to how artificial neural networks function. They aren’t called “artificial neural networks” for no reason.

            Here’s a useful graphic that shows things that make up “intelligence”

            A very basic definition of intelligence is “the ability to solve problems or make decisions”.

            I think the term is just often misused in common parlance so often that people start applying in a scientific setting incorrectly. Kinda how people used to call an entire computer the CPU, which like the word intelligence everyone understands what’s being said, but it’s factually wrong.

            Same thing today when people say “I bought a new GPU” when they should say “I bought a new video card” as the GPU is just a component.