We’re developing a blueprint for evaluating the risk that a large language model (LLM) could aid someone in creating a biological threat. In an evaluation involving both biology experts and students, we found that GPT-4 provides at most a mild uplift in biological threat creation accuracy. While this uplift is not large enough to be conclusive, our finding is a starting point for continued research and community deliberation.
  • @selfMAEnglish
    19•5 months ago

    the orange site is fucking dense with awful takes today:

    … I’m not trying to be rude, but do you think maybe you have bought into the purposely exaggerated marketing?

    That’s not how people who actually build things do things. They don’t buy into any marketing. They sign up for the service and play around with it and see what it can do.

    this self-help book I bought at the airport assured me I’m completely immune to both marketing and propaganda, because I build things (which entails signing up for a service that someone else built)

    with that said, there’s a fairly satisfying volume of folks correctly sneering at OpenAI in that thread too. some of them even avoided getting mass downvoted by all the folks regurgitating stupid AI talking points!

    • @froztbyteEnglish
      11•5 months ago

      because I build things (which entails signing up for a service that someone else built)

      fucking THIS

      I am so immensely fucking tired of seeing “I built an AI to do $x” posts that all fucking reduce to 1) “I strapped a custom input to the openai api (whose inputs and execution I can’t control nor reproduce reliably. I am very smart.)”, 2) a bad low-scope shitty-amounts-of-training hyperspecific toy model that solves only their exact 5 requirements (and basically nothing else, so if you even squint at it it’ll fall apart)

      basilisk save us from the moronicity

      • @selfMAEnglish
        11•5 months ago

        this is the damage done by decades of our industry clapping at brainless “I built this on cloud X and saved so much time” blog posts that have like 20 lines of code to do some shit like a lazy hacker news clone, barely changed from the example code the cloud provider publishes, and the rest is just marketing and “here’s how you use npm to pull the project template” shit for the post’s target market of mediocre VPs trying to prove their company’s spending too much on engineering and sub-mediocre engineers trying to be mediocre VPs

        like oh you don’t say, you had an easy time “building” an app when you wired together bespoke pieces of someone else’s API that were designed to implement that specific kind of app and don’t scale at all past example code? fucking Turing award material right here

        • @froztbyteEnglish
          9•5 months ago

          by decades of our industry clapping at brainless

          secondarily, the remarkable thing here is just how tiny a slice of industry this actually is (and yet also how profoundly impactful that vocal little segment can be)

          e.g. this shit wouldn’t fly in a bank (or at least, previously have flown), or somewhere that writes stuff that runs ports or planes or whatever.

          but a couple of decades of being worn down by excitable hyperproductive feature factory fuckwads who are only to happy to shit out Yet Another Line Of Code… it’s even impacting those areas at times

          some days I hate my industry so fucking much

          • @froztbyteEnglish
            5•5 months ago

            reflection thought: tonight (in the impending load shedding time) is a good time to reread Mickens

        • Brian David
          5•5 months ago

          @self @froztbyte Another big part of it is the obsession with the “young genius disruptor coder”. Which has resulted in management buying into endless fads foisted on us by twenty-somethings, and then inevitably having to undo half the things they implemented 5 years later. Well, except for React, which apparently we can’t get rid of but must forever keep reimplementing with whatever new new pattern will actually make it scale for real this time.

        • @froztbyteEnglish
          5•5 months ago

          don’t forget the 5 blog posts you can milk out of a single example, and your Learnings (obvious fucking realisations) 3 months (one even slightly minor application/API/… revision) later

    • Brian David
      9•5 months ago

      @self Oh no, it’s the “build” language. This really is web3 all over again, isn’t it?

      • @selfMAEnglish
        14•5 months ago

        it never stopped. it is a single unbroken stream of the worst people you’ve ever met trying to monetize you