170
Billion-Dollar AI Company Gives Up on AGI While Desperately Fighting to Stop Bleeding Money (futurism.com)
submitted 11 months ago by [M] to c/fuck_ai@lemmy.world
21 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 38 points 11 months ago (1 child)

Get fucked, so-called AI companies. Whatever gave these dumb asses any impression that LLMs will lead to AGI? I hope this bubble pops ASAP.

  • permalink
  • fedilink
  • source
  • hideshow 2 child comments
    •
  • [–] 0 points 11 months ago (5 children)

    My personal theory is that AGI is not possible with Binary architecture. I believe that it will take some form of analog computing, the simplest level of which is trinary. -1, 0, and 1.

    What is intelligence without the ability to be uncertain?

    Then there's full analog computing. It's not a robust field, but is deceptively powerful.

    There are these little solar powered robots that automatically move towards light, and it's done with a handful of transistors.

    To do the whole thing in binary would take a full processor and a few lines of code.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 10 child comments
    •
  • [–] 11 points 11 months ago* (1 child)

    Any number expressible in ternary, or a base 3 number system, is expressible in binary with a very simple formula to convert between the two. Binary just requires more digits. Fundamentally, a ternary computer is the same as a binary computer in terms of the problems that are decideable.

    Ternary computers have been around for a very long time. They are not new, and I fail to see how they are in any way relevant to AGI (which, to be clear, is something that exists purely in the realm of science fiction and not something we're actually going to be accomplishing any time soon). Ternary computing is certainly interesting and can offer potential performance improvements over binary computers in terms of speed/power efficiency for certain specialized applications, but it's not some magical new computing paradigm or something. Oh, and by the way, there's multiple ways of making ternary systems: -1, 0, and 1 is just one system (called balanced ternary).

    Also, fuzzy logic (or logic that accounts for uncertainty) has been around for a very long time, and in fact, is exactly what neural networks are using right now. It's encoded using floating point numbers between 0 and 1, which, in binary, are encoded using 32 or 64 bits (or more rarely, 16 bits). Again, not anything new.

    And I have no idea what you're talking about with analog.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 11 months ago (1 child)

    I mean, analog computing is a thing. The rest of it as you pointed out is mostly incoherent nonsense.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 11 months ago

    I think so too. Even how action potentials are received by a neuron, activate a neuron and how it is passed to another neuron is an amazingly complex process. However, these processes might not contribute to the intelligence. In current models they are abstracted away to direct mathematical transformations. But fundamentally, we simply don't know how or what gives rise to intelligence or even what is intelligence. So, I am fairly sure that AGI is very far away.

  • permalink
  • fedilink
  • source
  • parent