3 ms·

In 1949, some years after the invention of neural networks, Norbert Weiner, one of the leading minds of the time, was convinced that AI (AGI as you may call it)

by pron 7y ago

In 1949, some years after the invention of neural networks, Norbert Weiner, one of the leading minds of the time, was convinced that AI (AGI as you may call it) or a full understanding of the brain is no more than five years away. Alan Turing thought Weiner was delusional, and that it may take as much as fifty years. Seventy years later, we are nowhere near insect-level intelligence. I don't see any fundamental barrier preventing us from achieving AI, but if someone from the future came to me and said that AI will be achieved in 2130, I would find that quite reasonable. If they said it will be achieved in 2030 or 2230, I would find those equally reasonable. Our current scientific understanding is that we have no idea how far we are from AI, we don't know what the challenges are, and we don't even know what intelligence is. We certainly have no idea whether the approach we are now taking (statistical clustering, AKA deep learning) is a path that leads to AI or not. In the sixties, the leading minds of that time were also working hard on the problem and did not find it any further away from us as we do today. That some people are optimistic is irrelevant. The fact is that we just have no idea.

cr0sh 7y ago
> Seventy years later, we are nowhere near insect-level intelligence. That's arguable: For instance, we have the entire connectome of c. elegans mapped out; we can easily simulate it, and it seems to act the same as the actual nematode. So, in one sense, we are at that level. However, we still have no clue how such a simple system actually works to produce the level of "intelligence" it has. So in that sense, we're not at that level at all. > We certainly have no idea whether the approach we are now taking (statistical clustering, AKA deep learning) is a path that leads to AI or not. One clue we do have: We may not be on the right path with that method; it's something the "grandfather" (or whatever) of AI (Hinton) has mentioned, and which I have stated before about... That is, the fact that we currently have no understanding of the mechanism by which biological neural networks implement anything like "backpropagation". From what we currently understand, as I currently understand it, we have yet to find such a mechanism that would allow for it. It's also one of the leading reasons why our current artificial neural networks consume so much power, as compared to biological systems...
- pron 7y ago
  > For instance, we have the entire connectome of c. elegans mapped out... So, in one sense, we are at that level. Well, whatever "intelligence" C elegans has, I think everyone would agree that it's far from insect-level; it's microsocopic-nematode-level. But I am not sure a simulation of C elgans rises to the level of "artificial". As you note, we don't understand it yet. But we may have already built systems that are more "intelligent" (whatever that means) than C elegans, and we may have done that decades ago. > From what we currently understand, as I currently understand it, we have yet to find such a mechanism that would allow for it. True, but our path to artificial intelligence may not end up going through neural networks at all. We've not achieved flight by mimicking biological flight. I'm not saying it won't, either, but we cannot say for sure that it will. We really don't know.
- synthmeat 7y ago
  > We can easily simulate it. To my knowledge, this is nowhere near truth. They got it to wiggle by basically scripting muscle contractions, no neural networks are involved in the process. When you turn on the network, it does nothing. (And I love project, the idea and community behind it.)
carapace 7y ago
But Weiner's Cybernetics withered on the vine, or floated off into fluffy "Second Order" cybernetics. There was an experiment, I don't have the details to hand at the moment, I'm sorry, but Gordon Pask and someone else mage a cybernetics "machine" out of a dish of chemicals, and got it to grow an "ear" (filaments that were sensitive to certain sound vibrations, just like the hair cells in your inner ear)! If you really think about what they did (and you have to know some Cybernetics to understand it) then it's actually pretty scary. Like more-scary-than-atom-bomb scary. I'm only mentioning it here because we're about to need to grapple with this sort of thing in a minute or two... "Introduction to Cybernetics" W. Ross Ashby (1956) http://pespmc1.vub.ac.be/ASHBBOOK.html http://pespmc1.vub.ac.be/ASHBBOOK.html (PDF kindly made available from that page.)
- pron 7y ago
  My point wasn't about the particular form Weiner's cybernetics ended up taking; I was only commenting on his optimism regarding neural networks in the late forties and early fifties (you can say it's about his vision of Cybernetics rather than it's actual manifestation).
  - mbeex 7y ago
    Sorry for a German's nitpicking, but it seems to be not a typo: The man's name is Wiener.
    - carapace 7y ago
      Thanks!
  - carapace 7y ago
    Fair enough.
DonHopkins 7y ago
>Seventy years later, we are nowhere near insect-level intelligence. While the insects have been getting smarter! https://www.tedmed.com/talks/show?id=7286 https://www.tedmed.com/talks/show?id=7286 https://www.dailystar.co.uk/news/latest-news/403924/spiders-intelligent-larger-evolve-faster-than-humans https://www.dailystar.co.uk/news/latest-news/403924/spiders-...

by pron 7y ago

cr0sh 7y ago

> Seventy years later, we are nowhere near insect-level intelligence. That's arguable: For instance, we have the entire connectome of c. elegans mapped out; we can easily simulate it, and it seems to act the same as the actual nematode. So, in one sense, we are at that level. However, we still have no clue how such a simple system actually works to produce the level of "intelligence" it has. So in that sense, we're not at that level at all. > We certainly have no idea whether the approach we are now taking (statistical clustering, AKA deep learning) is a path that leads to AI or not. One clue we do have: We may not be on the right path with that method; it's something the "grandfather" (or whatever) of AI (Hinton) has mentioned, and which I have stated before about... That is, the fact that we currently have no understanding of the mechanism by which biological neural networks implement anything like "backpropagation". From what we currently understand, as I currently understand it, we have yet to find such a mechanism that would allow for it. It's also one of the leading reasons why our current artificial neural networks consume so much power, as compared to biological systems...

pron 7y ago
> For instance, we have the entire connectome of c. elegans mapped out... So, in one sense, we are at that level. Well, whatever "intelligence" C elegans has, I think everyone would agree that it's far from insect-level; it's microsocopic-nematode-level. But I am not sure a simulation of C elgans rises to the level of "artificial". As you note, we don't understand it yet. But we may have already built systems that are more "intelligent" (whatever that means) than C elegans, and we may have done that decades ago. > From what we currently understand, as I currently understand it, we have yet to find such a mechanism that would allow for it. True, but our path to artificial intelligence may not end up going through neural networks at all. We've not achieved flight by mimicking biological flight. I'm not saying it won't, either, but we cannot say for sure that it will. We really don't know.
synthmeat 7y ago
> We can easily simulate it. To my knowledge, this is nowhere near truth. They got it to wiggle by basically scripting muscle contractions, no neural networks are involved in the process. When you turn on the network, it does nothing. (And I love project, the idea and community behind it.)

carapace 7y ago

But Weiner's Cybernetics withered on the vine, or floated off into fluffy "Second Order" cybernetics. There was an experiment, I don't have the details to hand at the moment, I'm sorry, but Gordon Pask and someone else mage a cybernetics "machine" out of a dish of chemicals, and got it to grow an "ear" (filaments that were sensitive to certain sound vibrations, just like the hair cells in your inner ear)! If you really think about what they did (and you have to know some Cybernetics to understand it) then it's actually pretty scary. Like more-scary-than-atom-bomb scary. I'm only mentioning it here because we're about to need to grapple with this sort of thing in a minute or two... "Introduction to Cybernetics" W. Ross Ashby (1956) http://pespmc1.vub.ac.be/ASHBBOOK.html http://pespmc1.vub.ac.be/ASHBBOOK.html (PDF kindly made available from that page.)

pron 7y ago
My point wasn't about the particular form Weiner's cybernetics ended up taking; I was only commenting on his optimism regarding neural networks in the late forties and early fifties (you can say it's about his vision of Cybernetics rather than it's actual manifestation).
- mbeex 7y ago
  Sorry for a German's nitpicking, but it seems to be not a typo: The man's name is Wiener.
  - carapace 7y ago
    Thanks!
- carapace 7y ago
  Fair enough.

DonHopkins 7y ago

>Seventy years later, we are nowhere near insect-level intelligence. While the insects have been getting smarter! https://www.tedmed.com/talks/show?id=7286 https://www.tedmed.com/talks/show?id=7286 https://www.dailystar.co.uk/news/latest-news/403924/spiders-intelligent-larger-evolve-faster-than-humans https://www.dailystar.co.uk/news/latest-news/403924/spiders-...