3 ms·

> That would point to higher-level phenomena going on inside ChatGPT and its ilk, than merely statistics and predictions. No, it wouldn't, because nothing in "

> That would point to higher-level phenomena going on inside ChatGPT and its ilk, than merely statistics and predictions. No, it wouldn't, because nothing in "higher-level phenomena" precludes it being caused by statistics and predictions.

calf 4y ago
I don't understand that. In computer science everyone learned that computation is best described and explained at several levels of abstraction. E.g., HW/SW interface; machine code vs C++; RTL vs architecture, the list of levels of abstractions goes on and on. So what is the reason for not appropriately extending this idea to analyzing whatever a neural network is doing?
- unlikelymordant 4y ago
  I look at in the following way: understanding something by abstracting over lower level details doesn't mean the abstraction is how things actually work, the extra layer of abstraction may just be a nice way of thinking about something that makes thinking about it easier. But in the end the true mechanics are the sum of low level details. In general abstractions are not perfect, hence 'leaky abstractions'.
  - calf 4y ago
    That's a philosophical argument, whereas people are implicitly denying that ChatGPT can have higher level abstraction at all as if it were valid to claim that Stroustrup's contribution to C++ was ultimately just a way to write a lot of assembly code, so let's forever only think in assembly terms. (As a separate tangent, I don't accept the philosophy that abstractions are merely human niceties or conveniences. They are information theoretic models of reality and can be tested and validated, after all, even the bottom level of reality is an abstraction. The very argument used to deny the primacy of abstractions itself requires conceptual abstractions, leading to a circular logic. But then, I'm not a philosopher so what do I know.)
    - feanaro 4y ago
      (Poster from several comments upstream.) I'm not trying to deny the ChatGPT model has formed higher-level abstractions; in fact, I think it's fairly obvious you have to do this in some way in order to obtain such a level of breadth and accuracy of response. However those abstractions and their formation still boil down to statistics. This is similar to how e.g. mechanics of macroscopic bodies still boils down and reduces to quantum field theory and gravity, even though that's not the best way to explain or understand what's going on.
- zahrc 4y ago
  You’re definitely on to something. An example: when we interact with other human beings, we often really only care about the surface, don’t we? Mannerisms, looks, behaviour. Very rarely do we question those with “why?”. But who does? Psychologists. Same with any technology. Consumers don’t care about the why, they care about the result. Scientists and engineers care about the “why” and “how”. Now, is it important to understand “what’s behind the curtain”? Yes. But for who is it important?

calf 4y ago

I don't understand that. In computer science everyone learned that computation is best described and explained at several levels of abstraction. E.g., HW/SW interface; machine code vs C++; RTL vs architecture, the list of levels of abstractions goes on and on. So what is the reason for not appropriately extending this idea to analyzing whatever a neural network is doing?

unlikelymordant 4y ago
I look at in the following way: understanding something by abstracting over lower level details doesn't mean the abstraction is how things actually work, the extra layer of abstraction may just be a nice way of thinking about something that makes thinking about it easier. But in the end the true mechanics are the sum of low level details. In general abstractions are not perfect, hence 'leaky abstractions'.
- calf 4y ago
  That's a philosophical argument, whereas people are implicitly denying that ChatGPT can have higher level abstraction at all as if it were valid to claim that Stroustrup's contribution to C++ was ultimately just a way to write a lot of assembly code, so let's forever only think in assembly terms. (As a separate tangent, I don't accept the philosophy that abstractions are merely human niceties or conveniences. They are information theoretic models of reality and can be tested and validated, after all, even the bottom level of reality is an abstraction. The very argument used to deny the primacy of abstractions itself requires conceptual abstractions, leading to a circular logic. But then, I'm not a philosopher so what do I know.)
  - feanaro 4y ago
    (Poster from several comments upstream.) I'm not trying to deny the ChatGPT model has formed higher-level abstractions; in fact, I think it's fairly obvious you have to do this in some way in order to obtain such a level of breadth and accuracy of response. However those abstractions and their formation still boil down to statistics. This is similar to how e.g. mechanics of macroscopic bodies still boils down and reduces to quantum field theory and gravity, even though that's not the best way to explain or understand what's going on.
zahrc 4y ago
You’re definitely on to something. An example: when we interact with other human beings, we often really only care about the surface, don’t we? Mannerisms, looks, behaviour. Very rarely do we question those with “why?”. But who does? Psychologists. Same with any technology. Consumers don’t care about the why, they care about the result. Scientists and engineers care about the “why” and “how”. Now, is it important to understand “what’s behind the curtain”? Yes. But for who is it important?