272
I think the AI pause is just another grift (lemmy.world)
submitted 1 week ago by to c/fuck_ai@lemmy.world
50 comments fedilink hide all child comments
 

My pet theory about why Anthropic and Open AI are pushing for a “pause” is because they are trying to angle for a government bailout before their bubbles pop.

you are viewing a single comment's thread
view the rest of the comments
[–] 29 points 1 week ago* (3 children)

LLM technology has hit a logarithmic wall, and only marginally improves from more compute and bigger training sets. Such improvement as we see currently is from improved harness / framework, which can only go so far. It's not getting to self improvement or AGI, and certainly not getting to wholesale replacement of (non bullshit) jobs, which is what they bet the farm on. If they get the pause this will be less egregiously obvious and they'll have a chance to IPO and raid everyone's retirement funds. Otherwise, they'll crash and burn before IPO (SpaceX had actually valuable tech and still got dragged down by xAI and X) likely taking out Oracle and Softbank along with them (sob :). Either way China will be supplying the majority of the world market for LLMs much cheaper and on renewables in the next year or two. Soon after it'll be a mostly local technology after all the moat building market cornering bullshit with (V)RAM goes away.

  • permalink
  • fedilink
  • source
  • hideshow 3 child comments
    •
  • [–] 20 points 1 week ago

    What I also don't get is where they will get ever larger data sets that will make a significantly better model. Even if the benefits were linear+, you're running out of new fresh, quality resources to tap.

    You add in another 50,000,000 corporate emails you harvested, and stastically, they're a whole lot like the 100,000,000,000 you already saw. Conversely, you destroyed the last copy of some 1929 treatise to scan it in, just to dilute it in a hundred billion other pages.

    Moreover, the further into the "long-tail" of data sets you go, the less value they provide for mainstream audiences. If you tap an archive and Claude Code gets excellent at Sinclair ZX80 BASIC programming, that's not really what a bunch of React jockeys need.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 5 points 1 week ago (1 child)

    While I do agree with this, you also don't need AGI to cause severe damage.

    And of course the Pentagon has been the most reckless of all, which doesn't help.

    I think it's both. They're trying to figure out how to keep all the money while also, you know, not killing us all. I don't even want to describe how it can be done for it to be fed into training data, but it doesn't take much, and it doesn't have to be intentional.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 week ago*

    you also don’t need AGI to cause severe damage.

    Agreed, but people can be stupid (or dodge responsibility) with lots of technologies or other things. TBH I don't think it's particularly dangerous in and of itself at this stage, that's more marketing (ooh it's scary and dangerous, must be powerful). It can be a force multiplier for well trained people doing dangerous things, black hat hackers for example. It will not be killing us all, it might contribute to some (orange) idiot releasing the nukes, but that's on the idiot, if it wasn't LLMs it'd be something else, not that we'd care.

  • permalink
  • fedilink
  • source
  • parent