• XEAL@lemm.ee
    link
    fedilink
    arrow-up
    48
    arrow-down
    9
    ·
    1 year ago

    Large Language Models (such as GPT) and AI image generators.

    I follow certain AI related post tags on Tumblr and sometimes I see people expressing pure hatred towards these tools, as they only see the AIs as content thieves.

    • DokPsy@infosec.pub
      link
      fedilink
      arrow-up
      29
      arrow-down
      1
      ·
      1 year ago

      I don’t mind the tool itself if you use it as such. I do mind when people use its output as the final product. See: the lawyer who used chatgpt for a legal brief

      • XEAL@lemm.ee
        link
        fedilink
        arrow-up
        6
        arrow-down
        1
        ·
        1 year ago

        The lawyer fuck up is what happens when someone doesn’t know or understand the limitations of a LLM.

        If you want a GPT model tailored and specialized for a specific task, you have to train it with custom data, fine tune it and tweak the model’s parameters. You cannot do that from the ChatGPT web/app, you need a custom implementation coded in Python or some other language.

          • XEAL@lemm.ee
            link
            fedilink
            arrow-up
            2
            ·
            1 year ago

            Thanks. I have a quite powerful rig, but at the moment I work with OpenAI’s API using GPT 3.5 Turbo using a custom (but shitty) Python script with a simple Gradio web interface. However, I mostly stopped improving or updating it months ago. As long as I don’t use LlamaIndex, the cost is quite low.

            I already use Stable Diffusion WebUI, tho.

            Also the “fine tuning” I was talking about is this https://platform.openai.com/docs/guides/fine-tuning

            • TechieDamien@lemmy.ml
              link
              fedilink
              arrow-up
              2
              ·
              1 year ago

              I am aware what fine tuning is. It is available from the train tab while the base checkpoint is loaded in both cases.

        • uralsolo [he/him]@hexbear.net
          link
          fedilink
          English
          arrow-up
          2
          ·
          edit-2
          1 year ago

          I also don’t think that the ChatGPT model is able to do something that requires referencing case law or medical texts or whatever else at all in its current form. The way it works by generating probabilities for certain words is all wrong for doing something where the value of the output isn’t subjective - you need the model to be able to distinguish between facts and opinion, you need it to be able to cite sources for what it says, you need it to be able to produce coherent cause and effect chains and formulate an argument, all things which no currently existing LLM is capable of no matter how much you fine tune it because of how it works.

        • DokPsy@infosec.pub
          link
          fedilink
          arrow-up
          1
          ·
          1 year ago

          I’m glad you understand my point. Chatgpt is not Google. It’s a language model that will give you something that looks like the thing you asked for it to provide. It can and will pull facts out of its recycle bin if it fits the cadence of what it expects the answer to look like.

          • XEAL@lemm.ee
            link
            fedilink
            arrow-up
            1
            ·
            edit-2
            1 year ago

            ChatGPT is not Google, but sometimes it can work as a glorified search engine or even compete with asking in forums.

            I’ve lost count of how many times ChatGPT has produced Bash or Python code for what I needed. Yes, sometimes the code is wrong and/or requires tweaking and sometimes I resorted to look into the documentation, but no one will answer faster and anytime of the day like ChatGPT does, at least not for free.

            • DokPsy@infosec.pub
              link
              fedilink
              arrow-up
              1
              ·
              1 year ago

              It’s a tool to aid in creating a product, not a tool that magics out a finished product. That’s my point. Too many people use it as the latter instead of the former.

              • XEAL@lemm.ee
                link
                fedilink
                arrow-up
                1
                ·
                1 year ago

                100% agree.

                Maybe, with lots of training, weaking and testing the latter could be achieved, but that’s it.

        • DokPsy@infosec.pub
          link
          fedilink
          arrow-up
          2
          ·
          1 year ago

          Letting a language model do the work of thinking is like building a house and using a circular saw to put nails in. It will do it but you should not trust the results.

          It is not Google. It can, will, and has made up facts as long as it fits the format expected

          Not at the very least proof reading and fact checking the output is beyond lazy and a terrible use of a tool. Using it to create the end product instead of as a tool to use in creation of an end product are two very different things.

            • Carighan Maconar@lemmy.world
              link
              fedilink
              arrow-up
              2
              ·
              1 year ago

              The person you first replied to asked you to see the legal brief as an example of why they mind using the output as the finished product. You then asked for an explanation. To which I asked you, hey, have you actually looked at that example? You have not.

              What exactly do you want here, other than be argumentative for combative reasons?

    • alcoholicorn [comrade/them, doe/deer]@hexbear.net
      link
      fedilink
      English
      arrow-up
      26
      arrow-down
      2
      ·
      edit-2
      1 year ago

      they only see the AIs as content thieves.

      AI is a method of content theft, it takes other people’s work and pieces it together in a way that resembles other works, without any actual coherency.

      I don’t like that it churns out slop that displaces actual content.

      I also don’t like the way it’s sped up enshitification of google and news sites. I didn’t think it could get worse than pages of listicles written by disinterested journalists paid fuckall to churn out 10 a day, but now you have chatGPT churning out 100 completely useless articles a day.

      • XEAL@lemm.ee
        link
        fedilink
        arrow-up
        3
        arrow-down
        3
        ·
        edit-2
        1 year ago

        LLMs just automates and does faster certain things that a person could do on their own if they invested way more effort and time. If a human being takes people’s work and pieces it together in a way that resembles other works without using any LLM/AI or automation tool, is the final result content theft too?

        I agree with the content enshitification, but I disagree about the coherency.

        Usually, implementations like the ChatGPT web/app will generate different outputs for the same prompt/input. You can also ask it to tweak a previous output, make it shorter, more concise, exclude parts, etc. And if you’re making API calls through a script you can tweak parameters like the Temperature, Top P, Presence Penalty or Frequence Penaly, which affect things like the coherence, randomness or repetitiveness of the output.

        There’s also fine tunning using embeddings, which can help training a model to fit one’s specific needs and expectations, but I haven’t got to try it yet.

        • alcoholicorn [comrade/them, doe/deer]@hexbear.net
          link
          fedilink
          English
          arrow-up
          12
          arrow-down
          2
          ·
          1 year ago

          I disagree about the coherency.

          Coherency requires relating symbolic meanings. AI just uses statistical analysis.

          Consider if you were locked in the national library of Thailand. You don’t speak Siamese, and any pictures or bilingual dictionaries were removed.

          Given a thousand years, you could look at the patterns and produce text similar to what someone who writes Siamese would write, but there’s still no coherency because you cannot connect the meaning behind any of the words.

          That doesn’t necessarily mean your outputs are useless though, someone who does read Siamese can have you generate outputs until you print out something they can infer a coherent thought from, but you’re fundamentally unable to be trained to do that yourself.

          If a human being takes people’s work and pieces it together in a way that resembles other works without using any LLM/AI or automation tool, is the final result content theft too?

          We’re getting into ethics territory. IP is a social construct and we live under capitalism, our model for determining what is and isn’t theft should be selected by what supports artists and consumers against capitalists.

          • XEAL@lemm.ee
            link
            fedilink
            arrow-up
            1
            ·
            1 year ago

            Given a thousand years, you could look at the patterns and produce text similar to what someone who writes Siamese would write, but there’s still no coherency because you cannot connect the meaning behind any of the words.

            That doesn’t necessarily mean your outputs are useless though, someone who does read Siamese can have you generate outputs until you print out something they can infer a coherent thought from, but you’re fundamentally unable to be trained to do that yourself.

            You’re comparing an LLM to something similar to the infinite monkey theorem. In your analogy, you should consider that someone who knows perfect Siamese is giving me feedback to optimize and improve my outputs, even I don’t really know the meaning of anything.

            While an LLM may not have a conscience to evaluate if its output is coherent, it can identify patterns and relationships from its training and can generate text that is still appears coherent to human readers.

        • TheActualDevil@sffa.community
          link
          fedilink
          arrow-up
          3
          ·
          1 year ago

          If a human being takes people’s work and pieces it together in a way that resembles other works without using any LLM/AI or automation tool, is the final result content theft too?

          Yes, obviously. Artists and writers can learn from others and can be inspired by other’s works, but they can’t use parts of those works. That is content theft. Imitating a style is fine, but you have to create something new. LLMs cannot create, only steal.

          • XEAL@lemm.ee
            link
            fedilink
            arrow-up
            1
            arrow-down
            2
            ·
            edit-2
            1 year ago

            If, for example, I ask an LLM to produce a short story with a completely unique and random prompt that doesn’t resemble any known existing story in its training data (or in the entire world, if you like), is the generated output of the LLM also stolen?

            • TheActualDevil@sffa.community
              link
              fedilink
              arrow-up
              2
              arrow-down
              1
              ·
              1 year ago

              I think what you’re proposing isn’t something they can do. Are you saying “What if I asked it to create a short story who’s pieces don’t resemble any pieces of known stories?” or are you saying “What if I asked it to create a short story who’s whole doesn’t resemble any known stories?”

              The first one can’t happen. The second? Yes, it’s stealing.

              Where is it getting this story? LLMs don’t have creativity. They don’t understand story structure. It pulls sentences and paragraphs from work in it’s training data. If the generated output contains work that others have made, that’s called plagiarism. If it doesn’t, then your hypothetical isn’t realistic. LLMs can’t create original works. That’s the whole point. It pulls pieces of the training data and rearranges them. It would be like if I was writing a college paper and instead of writing anything myself I just pulled 100 different sources and copied a sentence or two from each source and structured them as my paper. That’s 100% plagiarism.

              • XEAL@lemm.ee
                link
                fedilink
                arrow-up
                1
                arrow-down
                1
                ·
                1 year ago

                I was referring to producing a unique plot.

                The process of generating a story involves recombining and rephrasing the LLM’s training data in unique ways, it’s not a copypaste job. They generate content by predicting and generating text based on patterns, an this implicates a degree of transformation and synthesis.

                Where do you draw the line between plagiarism vs inspiration, whether it’s a person or an LLM? How long and similar to something existing does a fragment of text have to be to cross the plagiarism line?

    • Rin@lemm.eeOP
      link
      fedilink
      arrow-up
      24
      arrow-down
      1
      ·
      edit-2
      1 year ago

      As an artist I think it’s a more complicated issue than a lot of people are making it out to be, and all the fearmongering some popular artists are promoting really doesn’t help.

      • XEAL@lemm.ee
        link
        fedilink
        arrow-up
        4
        arrow-down
        2
        ·
        1 year ago

        I think it’s a more complicated issue than a lot of people are making it out to be

        Agree.

        Also. People are pissed that what they have taken years to master others can now get close to replicate with little effort and time.

        I’ve just realized that although they call the AIs “content thieves”, what they really feel is that as AIs are able to replicate their skills quickly, it makes them feel their own merit diminished.

        If an artist creates artwork inspired on some other artist eveyone’s cool; if an AI does the same, then it’s stolen work even if the generated image is a unique new one.

      • davehtaylor@beehaw.org
        link
        fedilink
        arrow-up
        9
        arrow-down
        15
        ·
        1 year ago

        Using AI is feeding bullshit into a bullshit generator that’s handing back a synthesis of stolen art.

        It’s a bullshit tool for hacks and grifters to pretend they’re “artists” so they can exploit another avenue for the grind.

        This is not a debate. I don’t give a shit what your excuses are. Shout at a wall for all I care.

        • TooMuchDog@lemmy.ml
          link
          fedilink
          arrow-up
          13
          arrow-down
          5
          ·
          1 year ago

          You sound like you’ve already closed your mind to the discussion, but in case you’re actually still willing to healthily engage in the discussion here is a really good video about why calling people who utilize AI in their work “hacks and grifters” is a very narrow minded (and often factually incorrect) way of looking at AI utilization.

    • Kalash@feddit.ch
      link
      fedilink
      arrow-up
      9
      ·
      edit-2
      1 year ago

      It’s not that I hate it, but like, chatGPT sucks.

      There was this uber hype around it, then we started using it … and it just makes so many errors, it’s literally just generating more work. Scrapped it after less than a week. It’s modern snakeoil.

      • DustyNipples@lemmy.world
        link
        fedilink
        arrow-up
        3
        ·
        1 year ago

        Bard is the same, I asked it questions about two of my favourite bands whom I know a lot about. It omitted facts and invented things that were not true!

        • Kalash@feddit.ch
          link
          fedilink
          arrow-up
          3
          ·
          1 year ago

          We used it for code generation. But we ended up spending more time fixen and debugging the generated code than it would have taken us to just write it. Also it introduces the most annoying type of bugs. Like once it misspelled a property name, but only at one point in the code, got it right everywhere else.

        • XEAL@lemm.ee
          link
          fedilink
          arrow-up
          1
          ·
          1 year ago

          That’s why, in the case of a GPT model you would feed it custom training data using something like LlamaIndex. I don’t know if there’s an API available for Bard, tho.

          You’re wrong assuming that the free models that we have at our disposal are the only possible and best implementations of these LLMs.

      • Mojo@ttrpg.network
        link
        fedilink
        arrow-up
        2
        ·
        1 year ago

        What! I have the opposite experience.
        Im a tabletop roleplaying gamemaster and it has helped me immensely with translations, formatting of text, compiling and keeping track of my players character backgrounds and even coming up with plots and scenes that are suited for each player.

      • XEAL@lemm.ee
        link
        fedilink
        arrow-up
        2
        ·
        1 year ago

        What did you use it for? I helps me a lot with coding, scripting, translations, terminologies… Sometimes it makes mistakes, but other times it produces working code that accomplishes what I asked for.

        In any case, ChatGPT is just a demo that uses the GPT-3.5 Turbo model. Many people is being misled assuming that the ChatGPT research preview is all that the model has to offer. You can also try the improved model GPT-4, but it’s not free.

        If you really want to get its full potential you need a custom implementation in Python that works against the API and do things like fine tune the model, embeddings, feed it custom data or give it access to tools with LangChain.

        Of course that’s not something easy to do, but don’t think that the ChatGPT web/app is GPT models’ full potential.

        • Kalash@feddit.ch
          link
          fedilink
          arrow-up
          1
          ·
          edit-2
          1 year ago

          I have a feeling this one’s mostly operator error.

          Once we found the issues, it was actually quite easy to tell the AI to fix them. But at this point you’re debugging generated code to imrpove your input for the code generator … and it just was faster to write the code by hand.

          And yes, there was a vast overestimation of what it can do, especially by some managers that used to be coders and thought this would compensate for their lack of recent practical expirence. It didn’t … I had to fix it.

          • R0cket_M00se@lemmy.world
            link
            fedilink
            English
            arrow-up
            1
            ·
            1 year ago

            My point is that it’s not just for coding, if you think that’s the only use case then sure I get why you’d think it was shitty.

            • Kalash@feddit.ch
              link
              fedilink
              arrow-up
              1
              ·
              1 year ago

              I’ve used it a bit for general knowledge things and fun facts, and on more than a couple of occasions it just made shit up.

              I’m sure it has some uses, I see a lot of AI generated porn in my “all” feed … just haven’t found one for myself or my work.

              • R0cket_M00se@lemmy.world
                link
                fedilink
                English
                arrow-up
                1
                ·
                1 year ago

                Interesting, I’m working as a network engineer and my current job is overhauling an old TV broadcast facility. There are a lot of random solutions like using off brand switches and lack of documentation, etc.

                AI has been absolutely critical, it doesn’t do the work for me, but like any good tool it amplifies my ability to do work by cutting out the middle man of sifting through pages of spice works and stack overflow articles trying to figure out what command a ten year old Avaya needs to accomplish whatever task I require of it.

                Is it always correct? No. That’s why the engineer behind the screen exists. It does usually get me a workable answer more quickly than just having to look it up myself, though. Between my knowledge of terminal CLI commands and the AI, I’ve been able to get a lot done.

                Hell I had it walk me through the process of setting up automated backups, it even suggested the tftp server I used to do it. Shits been working great.

                Even our service desk has been able to use it to help with more advanced problems by telling it the issue and describing what has already been done.

                Idk why no one else sees the value, I’m over here like Captain Picard solving problems by talking to the LCARS system.

                • Kalash@feddit.ch
                  link
                  fedilink
                  arrow-up
                  1
                  ·
                  edit-2
                  1 year ago

                  I do see the potential value and I’m happy it worked out for you. But don’t end up like the lawyers that used chatGPT like a search engine and it just made up fictiional cases they cited in an actual court.

                  Yeah, that happened.

                  • R0cket_M00se@lemmy.world
                    link
                    fedilink
                    English
                    arrow-up
                    1
                    ·
                    1 year ago

                    You only end up like those morons by trusting AI to be perfect, I do not trust AI to even be “good” let alone perfect.

                    If you’re willing to just throw your job into an LLM and hope for the best you deserve to get fired.

    • SpaceCowboy@lemmy.ca
      link
      fedilink
      arrow-up
      8
      arrow-down
      1
      ·
      1 year ago

      LLM is way overhyped. So if your boss bought into that hype you’re gonna have a certain amount of animosity towards it. I’m a developer and it can be helpful at times, but managers seem to think it can write software on its own.

      It’s basically an iterative improvement over a search engine, but unlike a search engine it cuts off the people creating the content it’s scraping from any kind of revenue stream.

      And yeah there’s some real problems with it stealing content. Which isn’t being addressed at all. And bringing up these issues tend to get treated like Luddites by those that have bought into the hype.

    • Carighan Maconar@lemmy.world
      link
      fedilink
      arrow-up
      5
      ·
      1 year ago

      I wouldn’t say “hate”, to me it’s more… so what? They’re really bad at what they do, only impressive at first glance. Not bad for some brainstorming, but then you end up with a facsimile of what the actual result would be, and now have to use that as a guideline to create the result.

      • XEAL@lemm.ee
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        IMO they’re not bad, but they require a lot of tweaking and trial and error.

        I’ve learnt some Python thanks to ChatGPT’s help. When I say “some” I mean that I was able to create a custom implementation that uses a web interface and custom tools. The more lI learnt, the less I needed ChatGPT, but I always require some more coding help.

        However, these LLMs are not sentient super smart AIs.

    • uralsolo [he/him]@hexbear.net
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      1
      ·
      edit-2
      1 year ago

      I like them as non-profit tools for personal use, but the hatred is justified IMO because we’re already seeing people with writing jobs lose that job and get replaced by an LLM and an “editor” who is paid less than the writer was.

      Also, for stuff like art competitions and magazines, there is a need to develop a rigorous method of verification of what is and isn’t AI-generated. I’ve been published in a magazine before, but if I were to submit a story now I’d be competing against a massive wave of generated stories.

      • XEAL@lemm.ee
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        I like them as non-profit tools for personal use, but the hatred is justified IMO because we’re already seeing people with writing jobs lose that job and get replaced by an LLM and an “editor” who is paid less than the writer was.

        That’s capitalism in all of its glory. People never mattered to the ones who want to make money; they just want want to as much profit as possible with the minimal investment. Someone at work created a tool that turns a work day of painstating tasks into a 5 minute wait? Fire the people, keep the tool. You may call LLMs or AIs enablers, but it’s like hating baseball bats because some use them to crack open skulls instead of hitting baseballs.

        Regarding the verification of AI-generated content, I just can say I agree, but it’s going to be hard to detect.