3 ms·

> It might have been an interesting argument 20 years ago. It’s just silly now. Is it? These networks are capable of copying something, yes. Do we have a good

> It might have been an interesting argument 20 years ago. It’s just silly now. Is it? These networks are capable of copying something, yes. Do we have a good understanding of what that is? Not really, no. At least I don’t. I’m sure lots of people have a much better understanding than I do, but I think its hard to know exactly whats going on. People dismiss the stochastic parrot argument because of how impressive big neural nets are, but it doesn’t really invalidate that argument. Is a very, very, very good parrot that learns from everyone at once doing basically the same as what we do? I’d argue no, at least not fully. It’s absorbed aspects of us extremely well/is a very weird, sophisticated mirror, yes, and is copying something somehow, probably in a way reminiscent of how we copy. Is it basically the same as what we’re doing when we think? Partially? Fully? Not at all? A typical engineer would say “good enough”. That type of response is valuable in a lot of contexts, but I think the willingness to apply it to these models is pretty reckless, even if it’s impossible to easily “prove” why. To be clear on the exact statement you made, I think you’re right/it’s pretty clear neurons play some very important role/potentially capture a ton of what we consider intelligence, but I don’t think anyone really knows what exactly is being captured/what amount of thought and experience they’re responsible for.

Noble_Lie 3y ago
That person's argument is borderline insane to me - a severe lack of knowing what is unknown, a reverence of current best-models (regards modern science, including neurology - yet, open minded investigations beyond are also a requisite here.) And the pompousness is what truly boggles my mind ("Its silly to believe this, now.) A look in the mirror would suffice to say the least... Anyway, thank you for a great answer and conversation throughout this thread. Regards neural networks, parroting and the emulation of intelligence (or the difference between an emulation and the "real thing"): Well, somewhat like you say, we cannot propose a valid comparison from one to the other without an understanding of one (consciousness) or both. It's fascinating that there are some open, valid and pressing questions about what / how the output of this new wave of software is concretized (from foundational, semi-statistical algorithms in this case.) Yes, I do agree neurons have something to do with the "final output". But this is a subsection of the problem - organic neurons is-an/are order(s) of magnitude in complexity beyond what the tricky "parrot" is up to. Moreso, these components perform very different functionally - the known functions of the neuron compared to ANN, backprop etc. The entire stack.) P.S: One interesting theory I like to simulate and/or entertain is that every organic cell in the body has something to do with the final output of consciousness.
- didericis 3y ago
  Thank you, appreciate the compliment. And yeah, are definitely a lot of open questions related to all of this. Love how its brought so many deep questions into focus.
- gitfan86 3y ago
  Your theory makes sense in an evolutionary context. It is possible that all cells and organisms have some general intelligence. Humans do not have the ability to recognize this because evolutionarily it was only helpful to recognize intelligence when it could pose a threat to us. And the biggest threat to us in general was other humans as we are tribal animals. So we don't see it, we only see specialized intelligence that historically posed a threat to us. It would explain why most "experts" didn't see GTP-4's abilities coming. Many of them expected that it would take a major algorithm or technology improvement to do "real intelligent" things, because they fundamentally misunderstood intelligence.
- istjohn 3y ago
  Please read the comment I was responding to. I was addressing the suggestion that perhaps the brain is as relevant to intelligence as the eye. Cognitive Neuroscience has been a thriving field for nearly half a century now. I didn't say we have it all figured out, just that it's obvious neurons are a piece of the puzzle.

Noble_Lie 3y ago

That person's argument is borderline insane to me - a severe lack of knowing what is unknown, a reverence of current best-models (regards modern science, including neurology - yet, open minded investigations beyond are also a requisite here.) And the pompousness is what truly boggles my mind ("Its silly to believe this, now.) A look in the mirror would suffice to say the least... Anyway, thank you for a great answer and conversation throughout this thread. Regards neural networks, parroting and the emulation of intelligence (or the difference between an emulation and the "real thing"): Well, somewhat like you say, we cannot propose a valid comparison from one to the other without an understanding of one (consciousness) or both. It's fascinating that there are some open, valid and pressing questions about what / how the output of this new wave of software is concretized (from foundational, semi-statistical algorithms in this case.) Yes, I do agree neurons have something to do with the "final output". But this is a subsection of the problem - organic neurons is-an/are order(s) of magnitude in complexity beyond what the tricky "parrot" is up to. Moreso, these components perform very different functionally - the known functions of the neuron compared to ANN, backprop etc. The entire stack.) P.S: One interesting theory I like to simulate and/or entertain is that every organic cell in the body has something to do with the final output of consciousness.

didericis 3y ago
Thank you, appreciate the compliment. And yeah, are definitely a lot of open questions related to all of this. Love how its brought so many deep questions into focus.
gitfan86 3y ago
Your theory makes sense in an evolutionary context. It is possible that all cells and organisms have some general intelligence. Humans do not have the ability to recognize this because evolutionarily it was only helpful to recognize intelligence when it could pose a threat to us. And the biggest threat to us in general was other humans as we are tribal animals. So we don't see it, we only see specialized intelligence that historically posed a threat to us. It would explain why most "experts" didn't see GTP-4's abilities coming. Many of them expected that it would take a major algorithm or technology improvement to do "real intelligent" things, because they fundamentally misunderstood intelligence.
istjohn 3y ago
Please read the comment I was responding to. I was addressing the suggestion that perhaps the brain is as relevant to intelligence as the eye. Cognitive Neuroscience has been a thriving field for nearly half a century now. I didn't say we have it all figured out, just that it's obvious neurons are a piece of the puzzle.