Inside the Chaos at OpenAI
www.theatlantic.com
Sam Altman’s weekend of shock and drama began a year ago, with the release of ChatGPT.

Article from The Atlantic, archive link: https://archive.ph/Vqjpr

Some important quotes:

The tensions boiled over at the top. As Altman and OpenAI President Greg Brockman encouraged more commercialization, the company’s chief scientist, Ilya Sutskever, grew more concerned about whether OpenAI was upholding the governing nonprofit’s mission to create beneficial AGI.

The release of GPT-4 also frustrated the alignment team, which was focused on further-upstream AI-safety challenges, such as developing various techniques to get the model to follow user instructions and prevent it from spewing toxic speech or “hallucinating”—confidently presenting misinformation as fact. Many members of the team, including a growing contingent fearful of the existential risk of more-advanced AI models, felt uncomfortable with how quickly GPT-4 had been launched and integrated widely into other products. They believed that the AI safety work they had done was insufficient.

Employees from an already small trust-and-safety staff were reassigned from other abuse areas to focus on this issue. Under the increasing strain, some employees struggled with mental-health issues. Communication was poor. Co-workers would find out that colleagues had been fired only after noticing them disappear on Slack.

Summary: Tech bros want money, tech bros want speed, tech bros want products.

Scientists want safety, researchers want to research…

  • cwagner@beehaw.org
    14·
    7 months ago

    They believed that the AI safety work they had done was insufficient.

    Considering that every new model seems to be getting worse for anything but highly sanitized corporate usage, I’m not sure that I want more AI safety …

    For my usage, I use Chat GPT 3.5 turbo with the march checkpoint because I can’t get the current one to stop moralizing about bullshit instead of doing what it’s supposed to (I run two twitch bots with it). GPT4 used to be okay there, but the new preview is now starting to have the same issue with more frequent “I can’t do that Dave”-style answers, though it’s still mostly circumventable with enough prompt massaging, but it is getting harder.

    In a year, I don’t see anything but self-hosted models usable for anything not corporate glitz if trajectories hold, so fuck all that AI safety.

    • CosmoNova@feddit.deDeutsch
      5·
      7 months ago

      On top of all of this, those efforts to tame and control outputs from the developer side could be abused to simply appease investors or totalitarian markets. So we might see a Disneyfication like we‘re seeing on other platforms like Youtube with their horrendous filters, spawning ridiculous terms like „unlifed“. And just imagine the level of censorship we‘d see if they ever try to get into the Chinese market because clearly, the ‚non‘ in non-profit is becoming more and more silent.

    • RandoCalrandian@kbin.social
      4·
      7 months ago

      Pulled up a self hosted option last week to try it out. It’s not gpt4 level, but it’s damn close and I don’t worry giving access to my local documents

      PrivateGPT for anyone interested

        • RandoCalrandian@kbin.social
          3·
          7 months ago

          Mistral-7B-Instruct-v0.1 is the default, i’m downloading the Llama2 model to test it with now, but many models on HuggingFace should still work

          • cwagner@beehaw.org
            1·
            7 months ago

            I do not believe any 7B model comes even close to 3.5 in quality. I used LLama V1 64B, and it was horrible in comparison. Are you really telling me that this tiny model gives better general answers? Or am I just misunderstanding what you are saying?

            • RandoCalrandian@kbin.social
              1·
              7 months ago

              I didn’t say better, I said comparable
              And faster, without handing over my data and conversations for monetization

              Given the locally hosted benefits, and the ability to go to chatgpt for any answer minstrel gives that doesn’t satisfy you, makes it strong competition to chatgpt as the default tool

              Hosting it yourself also means you can swap llm’s out based on context and what they’re trained on. Highly tuned models perform better than chatgpt at the things they are meant to excel in.

              • cwagner@beehaw.org
                1·
                7 months ago

                Prompt:

                I’m currently trying to show on the Website Beehaw, that certain LLMs are far superior in writing than others. Examples of what bigger models do better than smaller ones: *

                Mistral-7B-Instruct-v0.1

                • ntire articles* vs Headlines Descriptions vs Product titles *Bul

                GPT 3.5-Turbo doesn’t support completion as it’s for chat, so I used an even worse one, text-davinci-003 which is far behind state of the art.

                Bigger models are able to handle more complex and detailed tasks with ease

                • Bigger models are better suited for natural language understanding and text processing
                • Bigger models are able to learn more accurate representations of context, thus improving the precision of the output
                • Bigger models can process data more quickly and efficiently, saving time and processing power when large volumes of data are used
                • Bigger models can better recognize more subtle nuances in language, which allows them to produce more accurate results
                • Bigger models are able to use more sophisticated algorithms, resulting in a more comprehensive and deeper understanding of the data being used

                Mistral 7B might be okay for some very specific cases, but it’s not comparable to proper models at all.

                edit: gave it a second chance, it’s a bit better (at least no complete nonsense anymore), but still terrible writing and doesn’t make much sense

                Paraphrasing The ability of a language model to generate text that has a similar meaning to the original text is called paraphrasing. This is a very common problem in natural language processing, and many LLMs are designed to be able to paraphrase text. However, there are some LLMs that are particularly good at paraphrasing, and these models are often preferred over smaller models because of their ability to generate more varied and unique text. Examples of LLMs that are known for their paraphrasing abilities include GPT-2 and transformers. These models

    • smoothbrain coldtakes@lemmy.caEnglish
      4·
      7 months ago

      It’s already easy to self host and we’ve optimized LLMs to run locally on not much serious hardware after we’ve trained them; I have GPT4ALL set up on my machine and it runs everything locally with my processor, no GPU or anything. Some of those datasets are uncensored, and I’ve seen what Stable Diffusion can do for image generation.

      I tend to use the GPT-4 built into Edge with my O365 corporate plan, because it suits my needs better for day-to-day challenges. It can still audit code and summarize things, which is all I really need it to do here and there.

      • cwagner@beehaw.org
        5·
        7 months ago

        Nothing that runs on my GPU / CPU comes even close to GPT 3.5, GPT4 is not even in the same universe, and that’s with them running far more slowly.