4 ms·

The real question is "How can you prove that somebody/something has the ability to think ?". If LLM can reason and answer satisfactorily based on context I woul

by mlboss 3y ago

The real question is "How can you prove that somebody/something has the ability to think ?". If LLM can reason and answer satisfactorily based on context I would deduce that it can think. What is the argument that human does not have transformer architecture. We are also predicting one word after next in our brain.

hammyhavoc 3y ago
Human brains are objectively and quantifiably vastly more complex, including context and planning, they don't sequentially predict the next word. Prime example of a human brain being architecturally different to an LLM: LLMs never interrupt themselves. They never get mid-sentence then say "actually, scratch that, I was wrong" or "now that I've thought about it, this is actually a better method to convey what I was trying to tell you". There's a lot of material on this exact topic highlighting how the human brain does not have a transformer architecture. Anecdata: I don't think in words. I think in abstract non-concrete thoughts. Language is what I use to convey my thoughts to others, program, or write notes to recall those abstract thoughts. However, I am also autistic, so take that with a large pinch of salt.
- tspike 3y ago
  Is it fair to say that you define thinking as operating in a nearly identical manner as the human brain?
  - hammyhavoc 3y ago
    Well, herein lies the interesting thing: my experience of "normal" thinking won't be like the average human being due to autism. Thus I acknowledge variance of how brains function even amidst human beings. It isn't an impossibility to consider that this could extend to artificial intelligence, but it's an LLM. It's intelligence-like (artificial), but it isn't intelligence by the traditional measure.
    - carlmr 3y ago
      >It's intelligence-like (artificial), but it isn't intelligence by the traditional measure. If it being artificial makes it intelligence like, isn't that putting the conclusion first?

by mlboss 3y ago

hammyhavoc 3y ago

Human brains are objectively and quantifiably vastly more complex, including context and planning, they don't sequentially predict the next word. Prime example of a human brain being architecturally different to an LLM: LLMs never interrupt themselves. They never get mid-sentence then say "actually, scratch that, I was wrong" or "now that I've thought about it, this is actually a better method to convey what I was trying to tell you". There's a lot of material on this exact topic highlighting how the human brain does not have a transformer architecture. Anecdata: I don't think in words. I think in abstract non-concrete thoughts. Language is what I use to convey my thoughts to others, program, or write notes to recall those abstract thoughts. However, I am also autistic, so take that with a large pinch of salt.

tspike 3y ago
Is it fair to say that you define thinking as operating in a nearly identical manner as the human brain?
- hammyhavoc 3y ago
  Well, herein lies the interesting thing: my experience of "normal" thinking won't be like the average human being due to autism. Thus I acknowledge variance of how brains function even amidst human beings. It isn't an impossibility to consider that this could extend to artificial intelligence, but it's an LLM. It's intelligence-like (artificial), but it isn't intelligence by the traditional measure.
  - carlmr 3y ago
    >It's intelligence-like (artificial), but it isn't intelligence by the traditional measure. If it being artificial makes it intelligence like, isn't that putting the conclusion first?