• schnapsidee@feddit.de
    link
    fedilink
    arrow-up
    15
    ·
    1 year ago

    Decisions like this just prove how massive the market for a self-hostable alternative is. They’re not banning it because it’s a bad tool, they’re banning it because they’re concerned about what happens to the source code their engineers paste into it.

    There are already a bunch of OSS attempts, and it likely won’t take long until we have something of comparable quality to ChatGPT is available for companies to host on their own hardware.

    • eight_byte@feddit.de
      link
      fedilink
      arrow-up
      4
      ·
      1 year ago

      Companies are also banning ChatGPT because its unclear from where the code it produces was stolen and how it’s licensed. Copy and pasting code from AI tools is an enormous risk for a software company.

      • schnapsidee@feddit.de
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        As I said, there are some self-hostable alternatives, but nothing even remotely enterprise ready yet. I’m keeping a pretty close eye on this because my boss wants to train a support chatbot on company data and run it on our own hardware. (And an alternative to copilot would be great too, as that’s banned for internal use.) There are some great tools to tinker around with, but I haven’t found anything that I would call production ready.

    • Mon0@kbin.social
      link
      fedilink
      arrow-up
      2
      ·
      1 year ago

      No, this just proves what everybody knows that has worked with ChatGPT. It is a nice tool if you want to write a story but everything else is just a time waste. Contrary to the media belief 99% of ChatGPTs answers to business related questions (including coding) produce a partially wrong or completely wrong answer.
      You rly can‘t trust the answers ChatGPT gives you at all.
      And coding … Copilot is already not good (in coding but very useful for auto completion) but ChatGPT is actually worse. ChatGPT fails even on easy coding tasks in most languages and even the JS solutions are mostly horrible.

      Sure the code is also a problem, but in the here and now the biggest problem are devs that just believe whatever ChatGPT prints out and in the end you have a PR full of code (including deprecated extensions and packages) from yesteryear.

      But self hosted models would be awesome nonetheless.

      • tinselpar@feddit.nl
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        You rly can‘t trust the answers ChatGPT gives you at all.

        I asked the local AI, and it agrees with you:

        The trustworthiness of ChatGPT as an educational resource is a topic of discussion. Here are some points from the search results:
        
            ChatGPT provides fast answers to inputs, but they're not necessarily trustworthy. It can give wholly or partially false information that seems very believable. People in the artificial intelligence research world deem this problem “hallucinations” 1
            .
            ChatGPT's responses are based on patterns in the text it was trained on, not on external facts and data. While the tool tries to provide correct information, its responses are not always trustworthy2
            .
            ChatGPT can sometimes provide incorrect or inaccurate answers. The frequency of incorrect answers can vary depending on the specific question, the available training data, and the complexity of the topic. While ChatGPT has been trained on a wide range of internet text to generate responses, it does not have real-time information or knowledge beyond its training data3
            .
            ChatGPT may make mistakes such as providing irrelevant or incorrect responses, repeating itself, or producing responses that are inconsistent with the context of the conversation. These mistakes can occur because ChatGPT is trained on vast amounts of text data, including unverified and potentially biased information, which can lead to incorrect or outdated information4
            .
            ChatGPT's credibility is influenced by its training data, which can introduce biases or inaccuracies, and the fine-tuning process, which involves human reviewers. It's important to approach the model's responses critically and verify information from reliable sources when needed3
            .
        
        **In summary, while ChatGPT can provide helpful responses, it is not always trustworthy and its responses should be verified against other sources.**
        

        It wouldn’t make that shit up would it? 🤣😂

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

    Well of course, ChatGPT has already leaked Samsung Semiconductor’s internal information earlier, and Apple is infamous for being secretive about their design.

  • MentalEdge
    link
    fedilink
    arrow-up
    9
    arrow-down
    10
    ·
    edit-2
    1 year ago

    How to neuter your own ability to compete: ban your workers from using the latest tool for boosting employee performance.

      • MentalEdge
        link
        fedilink
        arrow-up
        16
        arrow-down
        2
        ·
        edit-2
        1 year ago

        Then you don’t have much faith for your co-workers competence in wielding any given tool to its greatest utility. Using an LLM like ChatGPT to access data hardly automatically means you’re also a brain-dead search result copy-paster.

        Yes, its a new interface for existing data, the same way digital files are to data on paper. Only ever using the latter is really inefficient, and stupid in a world where the digital files exist. Not that the hardcopies cant be to their own utility, or be used as corroborating data.

        It’s a really good interface, if you know how to use it. This is like banning search engines because you expect your workers to be expert at everything, so they shouldn’t need support tools to sleuth for data.

      • PerogiBoi@lemmy.ca
        link
        fedilink
        English
        arrow-up
        4
        ·
        1 year ago

        I’m gonna vehemently disagree with you. As a knowledge worker, ChatGPT allows me to offload low level thinking and writing tasks so I can focus on bigger picture creative aspects.

        GPT speeds up my quality work output by around half. Those who refuse to incorporate it into their work flow will find they fall behind compared to those who have successfully integrated it.

      • quirzle@kbin.social
        link
        fedilink
        arrow-up
        3
        arrow-down
        1
        ·
        1 year ago

        Frankly, if ChatGPT isn’t increasing your performance significantly, you’re already falling behind the curve unless you’re doing manual labor.

        • MentalEdge
          link
          fedilink
          arrow-up
          6
          ·
          1 year ago

          Exactly. Used correctly, the amount of man-hours ChatGPT is able to save, is truly ludicrous.

      • animist@lemmy.one
        link
        fedilink
        arrow-up
        2
        arrow-down
        1
        ·
        1 year ago

        Better stop using xerox machines to make copies and write everything out by hand

    • Ulu-Mulu-no-die@lemmy.world
      link
      fedilink
      arrow-up
      5
      ·
      1 year ago

      Leaking industry secrets is a much bigger concern that boosting productivity a little bit.

      We’re talking about very specialized engineering work, it’s not something you can totally rely on a bot to do, though it might help sometimes, it’s fully understandable for specialized companies to want to ban GPT internally, until there’s a way for them to host a totally internal one.

      • MentalEdge
        link
        fedilink
        arrow-up
        4
        ·
        1 year ago

        On this I agree entirely. The potential for corporate espionage because of unwitting employees using an LLM through unofficial means is huge.

        At the very least, the corporation itself would have to be the customer, so that watertight terms might be negotiated, not the employee.

        • Ulu-Mulu-no-die@lemmy.world
          link
          fedilink
          arrow-up
          2
          ·
          edit-2
          1 year ago

          I don’t think being a customer would work either, language models are still on the training, noone knows exactly how users queries are used, that’s a big no no for every company having to protect their secrets.

          A self-hosted instance is a much better solution, if not the only “safe” one from that point of view, we’ll get there.

          • MentalEdge
            link
            fedilink
            arrow-up
            3
            ·
            edit-2
            1 year ago

            Interaction data does not become training data, unless you want it to.

            I know that how a piece of software created using machine learning works, is an unknowable, but training data and interaction data are not the same thing. ChatGPT in particular is designed to be restored to a known good start state, only using query data for context awareness within a given sessions. Not to train itself.

            Each query simply includes all previous queries, for context. That’s part of why it becomes increasingly erratic the longer a session goes on.

            And unless you do train with a given piece of data, that data is not entered into the LLM in any way. Not even the undefined unknowable way.

      • quirzle@kbin.social
        link
        fedilink
        arrow-up
        2
        ·
        1 year ago

        We’re talking about very specialized engineering work,

        We’re not though. This isn’t a policy preventing them from disclosing them from talking about specific company IP (which is almost certainly covered by existing NDAs already). This prevents them from using it internally at all.

        I use ChatGPT at work all the time, usually for getting very specific information about products I have to integrate with, quickly parsing new API documentation, and learning about unfamiliar processes at a conceptual level before I have to dive deeper for a project. It’s more the context around which I’ll be building the specialized IP. It’s the sort of stuff I can learn via Googling (or sometimes Stack Exchange), but can learn it faster in a more targeted manner by asking detailed questions to the chatbot.

    • RupeThereItIs@kbin.social
      link
      fedilink
      arrow-up
      4
      ·
      1 year ago

      It’s a MASSIVE security risk. What you tell ChatGPT is not private, if you knowingly or unknowingly tell ChatGPT secret information you have no control over where that information may go. Especially for a company for Apple that lives & breaths on surprise product releases.

      • MentalEdge
        link
        fedilink
        arrow-up
        3
        ·
        edit-2
        1 year ago

        This is true, but if you understand that queries don’t necessarily need to also become training data, what you tell it could absolutely be kept secret, provided the necessary agreements and changes were to be made. Nothing about an LLM means you can’t make it forget things you’ve told it. What you can’t make it forget, without re-training it from the ground up with that piece of information omitted, is what you told it in the training data.

        But queries, do not suffer this limitation.

    • itchy_lizard@feddit.it
      link
      fedilink
      arrow-up
      1
      ·
      edit-2
      1 year ago

      I agree with your sentiment if the tech were self-hosted, but there are huge security risks to pasting sensitive internal content into a third party took

      We’ve had similar issues with Grammerly. That shit is basically a cloud-based keylogger. My god it’s horrible that anybody installs such a thing.

      • MentalEdge
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        That depends on what kind of agreement exists between your company, and that of the third party tool. Yes, in the worst case, the answer is “none”.

        But most workflows involve quite a lot of third party tools, only they are all licensed, and with clear details worked out for what data can go where.

        That employees are using such tools without there being a proper deal… Is a temporary problem to which a ban is not the solution.