8 ms·

People reading this on Hackers News will of course agree with his main point - Our computers can't 'Think' as such and when we talk about them as if they can, w

People reading this on Hackers News will of course agree with his main point - Our computers can't 'Think' as such and when we talk about them as if they can, we're setting up for misunderstandings. But he seems to also be saying that we'll never get there because computers just run algorithms. Arguments like that always remind me of this bit from Iain Bank's 'A Few Notes On The Culture' Certainly there are arguments against the possibility of [strong] Artificial Intelligence, but they tend to boil down to one of three assertions: one, that there is some vital field or other presently intangible influence exclusive to biological life - perhaps even carbon-based biological life - which may eventually fall within the remit of scientific understanding but which cannot be emulated in any other form (all of which is neither impossible nor likely); two, that self-awareness resides in a supernatural soul - presumably linked to a broad-based occult system involving gods or a god, reincarnation or whatever - and which one assumes can never be understood scientifically (equally improbable, though I do write as an atheist); and, three, that matter cannot become self-aware (or more precisely that it cannot support any informational formulation which might be said to be self-aware or taken together with its material substrate exhibit the signs of self-awareness). ...I leave all the more than nominally self-aware readers to spot the logical problem with that argument.

RashHeuristics 7y ago
>"that there is some vital field or other presently intangible influence exclusive to biological life" There is. The ability to instantiate one specific abstract into a physical item. An abstract can for example be the idea of triangularity and the physical substrate will be the brain. No AI running on a typical Turing machine can do that. For any idea you believe you have instantiated into the machine you can come up with an alternative interpretation of what you actually have instantiated. This means that any physical system alone underdetermines the abstract thing it is supposed to be an instantiation of. You can for example likely find a pattern within the atoms of the wall behind you that could be said to be an instantiation of the quicksort algorithm. Or it's just a wall. The physical item itself is silent on the matter.
- codeulike 7y ago
  You're talking about an ability, Iain Banks was talking about some unknown bit of physics, a field or somesuch. From what do you posit this ability to "instantiate one specific abstract into a physical item" comes from, and why do you think only wetware can do it? Reading it again, I think you might be talking about Qualia.
  - RashHeuristics 7y ago
    The introspective ability to hold abstract concepts in the mind is the source of the statement that humans can do so. It is what Aristotle called the intellectual soul, not qualia. Qualia is an experience and has no relation to holding abstract concepts. As I already wrote, Turing machines cannot do it for they suffer from multiple realizability, any abstract concept can be implemented in multiple physical constructs and any physical construct can be interpreted to be any number of different abstract items. I have not written that only "wetware" can do it, I have written that turing machines cannot do it and that no purely physical construct can do it (including purely newtonian/quantum wetware).
    - Joker_vD 7y ago
      Uh... but the neural networks (and Bayesian networks) do hold abstract concepts, and are implementable by Turing machines. So it appears that your assertion is wrong.
      - RashHeuristics 7y ago
        They do not hold concepts, they are simply physical items behaving in accord with physical principles. You as a human interpret them (wrongly) to hold concepts. You can as well interpret them to be something else. None of your interpretations change the physics and the physics have no mention of abstract concepts and can be interpreted to be any number of things. None of the interpretations make that particular interpretation a concrete implementation of it.
    - codeulike 7y ago
diablo1 7y ago
"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim" - Edsger Dijkstra
xamuel 7y ago
People in this debate miss a subtle issue, and it causes no end of confusion. The subtle issue is this. There is a difference between (1) accidentally building a computer which happens to think, and (2) intentionally building a computer which thinks, and knowing that it thinks. Obviously (1) is possible, because for any particular configuration of atoms, it is theoretically possible for someone to assemble atoms together in that configuration. I could randomly put atoms together and, by sheer luck, assemble a living, breathing Abraham Lincoln. But since it happened by random chance, I wouldn't know for sure that this was really a living, breathing Abraham Lincoln. I might observe it for awhile and notice that, golly, it sure does seem to be a living, breathing Abraham Lincoln. But I could never rule out the possibility that it only looks like that initially and that it'll eventually collapse or fall into some infinite loop or something. Intentionally building a machine and KNOWING that it has such-and-such properties of a thinking machine, is much more interesting. Such KNOWING is, itself, an act of thinking, and so there arises an interplay between the thinking of the created machine, and the thinking of the creator. Then you can start asking questions about whether the created machine, being a thinker, could also itself intentionally create another thinking machine, and so on. If you're tired of the endless non-productive debate about whether thinking machines are possible, and you'd like to go to the next level, and take steps toward actually quantifying things, I'd invite you to read some papers I've written on the subject. To lessen the learning curve, here are some very approachable slides, and if you like those slides, you can click through to one of the papers. https://semitrivial.github.io/MeasuringIntelligence2019.pdf https://semitrivial.github.io/MeasuringIntelligence2019.pdf
- atomack 7y ago
  I'm not sure (1) is 'obviously possible'. Supposing there's a finite number of atomic configurations that produce Abraham Lincoln, there's still an infinite number of atomic configurations so you're probability is strictly zero, no? It's not even an 'age of the universe to achieve' type event. And that assumes that your random throwing together of atoms includes the physical preconditions to create an Abraham Lincoln. For instance, if you were sitting in the middle of a black hole randomly assembling atoms I'm pretty sure it's physically impossible to create an Abraham Lincoln.
- 7y ago
_bxg1 7y ago
Put more simply: if we assume the human mind to be material, and deterministic, then by definition it is possible to replicate it in a computer (at least one with infinite resources). At the same time, contrary to the popular media narrative, today's "AI" is still lightyears away from that.
MaxBarraclough 7y ago
Strongly agree. Banks expresses those points well. The piece fails to account for that we already know that it's possible for physical entities to think, and to make independent decision. We know this because human beings exist. No good reason is offered to justify the idea that the difference in intelligence between humans and computers is a difference of category, rather than merely a difference of (ever shrinking) degree. From the article: > The intellectual incoherence of believing that computers can think or make independent decisions is self-evident Nonsense. This is a very strong claim about the universe, or rather, one of two equally silly claims: * That general intelligence relies upon the process of evolution by natural selection, and that it can never even in principle arise from a process of deliberate design * That general intelligence relies upon a particular substrate, specifically neurons, and that it cannot be achieved with other substrates, such as transistors Such a claim cannot be justified with a hand-wavey appeal to intuition, or with arguments about moral accountability, which are all the author has to offer. Also, this next quote strikes me as misleading. Perhaps someone more knowledgeable than me can confirm/disconfirm my take: > chess prowess depends on players’ ability to recognize general situations that are in some sense “like” ones they’ve seen before, either over the board or in books. Again, the computer largely sidestepped what is most significant for humans, and hence for the game, by analyzing every position from scratch, and relying on speed to make up for its weakness at gestalt pattern recognition. That may be true of chess, but what about Go? As I understand it, the reason computers can now play Go effectively (i.e. Alpha Go), is that a breakthrough was made in discarding unpromising subtrees of the game's possible progression, allowing effective moves to be selected relatively efficiently. That's not brute-force. That's precisely the kind of clever trick we'd expect from a skilled human player. The whole piece strikes me as poorly considered. The thoughts of Dan Dennett are far more worthwhile. If you've read this far, you may enjoy a quick read of the Wikipedia article on Dennett's intentional stance. [0] Dennett also makes the relevant observation that mankind is constantly creating machines with general intelligence and moral accountability: other human beings. This ties back to the article's points about moral accountability. [0] https://en.wikipedia.org/wiki/Intentional_stance https://en.wikipedia.org/wiki/Intentional_stance

RashHeuristics 7y ago

>"that there is some vital field or other presently intangible influence exclusive to biological life" There is. The ability to instantiate one specific abstract into a physical item. An abstract can for example be the idea of triangularity and the physical substrate will be the brain. No AI running on a typical Turing machine can do that. For any idea you believe you have instantiated into the machine you can come up with an alternative interpretation of what you actually have instantiated. This means that any physical system alone underdetermines the abstract thing it is supposed to be an instantiation of. You can for example likely find a pattern within the atoms of the wall behind you that could be said to be an instantiation of the quicksort algorithm. Or it's just a wall. The physical item itself is silent on the matter.

codeulike 7y ago
You're talking about an ability, Iain Banks was talking about some unknown bit of physics, a field or somesuch. From what do you posit this ability to "instantiate one specific abstract into a physical item" comes from, and why do you think only wetware can do it? Reading it again, I think you might be talking about Qualia.
- RashHeuristics 7y ago
  The introspective ability to hold abstract concepts in the mind is the source of the statement that humans can do so. It is what Aristotle called the intellectual soul, not qualia. Qualia is an experience and has no relation to holding abstract concepts. As I already wrote, Turing machines cannot do it for they suffer from multiple realizability, any abstract concept can be implemented in multiple physical constructs and any physical construct can be interpreted to be any number of different abstract items. I have not written that only "wetware" can do it, I have written that turing machines cannot do it and that no purely physical construct can do it (including purely newtonian/quantum wetware).
  - Joker_vD 7y ago
    Uh... but the neural networks (and Bayesian networks) do hold abstract concepts, and are implementable by Turing machines. So it appears that your assertion is wrong.
    - RashHeuristics 7y ago
      They do not hold concepts, they are simply physical items behaving in accord with physical principles. You as a human interpret them (wrongly) to hold concepts. You can as well interpret them to be something else. None of your interpretations change the physics and the physics have no mention of abstract concepts and can be interpreted to be any number of things. None of the interpretations make that particular interpretation a concrete implementation of it.
  - codeulike 7y ago

diablo1 7y ago

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim" - Edsger Dijkstra

xamuel 7y ago

People in this debate miss a subtle issue, and it causes no end of confusion. The subtle issue is this. There is a difference between (1) accidentally building a computer which happens to think, and (2) intentionally building a computer which thinks, and knowing that it thinks. Obviously (1) is possible, because for any particular configuration of atoms, it is theoretically possible for someone to assemble atoms together in that configuration. I could randomly put atoms together and, by sheer luck, assemble a living, breathing Abraham Lincoln. But since it happened by random chance, I wouldn't know for sure that this was really a living, breathing Abraham Lincoln. I might observe it for awhile and notice that, golly, it sure does seem to be a living, breathing Abraham Lincoln. But I could never rule out the possibility that it only looks like that initially and that it'll eventually collapse or fall into some infinite loop or something. Intentionally building a machine and KNOWING that it has such-and-such properties of a thinking machine, is much more interesting. Such KNOWING is, itself, an act of thinking, and so there arises an interplay between the thinking of the created machine, and the thinking of the creator. Then you can start asking questions about whether the created machine, being a thinker, could also itself intentionally create another thinking machine, and so on. If you're tired of the endless non-productive debate about whether thinking machines are possible, and you'd like to go to the next level, and take steps toward actually quantifying things, I'd invite you to read some papers I've written on the subject. To lessen the learning curve, here are some very approachable slides, and if you like those slides, you can click through to one of the papers. https://semitrivial.github.io/MeasuringIntelligence2019.pdf https://semitrivial.github.io/MeasuringIntelligence2019.pdf

atomack 7y ago
I'm not sure (1) is 'obviously possible'. Supposing there's a finite number of atomic configurations that produce Abraham Lincoln, there's still an infinite number of atomic configurations so you're probability is strictly zero, no? It's not even an 'age of the universe to achieve' type event. And that assumes that your random throwing together of atoms includes the physical preconditions to create an Abraham Lincoln. For instance, if you were sitting in the middle of a black hole randomly assembling atoms I'm pretty sure it's physically impossible to create an Abraham Lincoln.
7y ago

_bxg1 7y ago

Put more simply: if we assume the human mind to be material, and deterministic, then by definition it is possible to replicate it in a computer (at least one with infinite resources). At the same time, contrary to the popular media narrative, today's "AI" is still lightyears away from that.

MaxBarraclough 7y ago

Strongly agree. Banks expresses those points well. The piece fails to account for that we already know that it's possible for physical entities to think, and to make independent decision. We know this because human beings exist. No good reason is offered to justify the idea that the difference in intelligence between humans and computers is a difference of category, rather than merely a difference of (ever shrinking) degree. From the article: > The intellectual incoherence of believing that computers can think or make independent decisions is self-evident Nonsense. This is a very strong claim about the universe, or rather, one of two equally silly claims: * That general intelligence relies upon the process of evolution by natural selection, and that it can never even in principle arise from a process of deliberate design * That general intelligence relies upon a particular substrate, specifically neurons, and that it cannot be achieved with other substrates, such as transistors Such a claim cannot be justified with a hand-wavey appeal to intuition, or with arguments about moral accountability, which are all the author has to offer. Also, this next quote strikes me as misleading. Perhaps someone more knowledgeable than me can confirm/disconfirm my take: > chess prowess depends on players’ ability to recognize general situations that are in some sense “like” ones they’ve seen before, either over the board or in books. Again, the computer largely sidestepped what is most significant for humans, and hence for the game, by analyzing every position from scratch, and relying on speed to make up for its weakness at gestalt pattern recognition. That may be true of chess, but what about Go? As I understand it, the reason computers can now play Go effectively (i.e. Alpha Go), is that a breakthrough was made in discarding unpromising subtrees of the game's possible progression, allowing effective moves to be selected relatively efficiently. That's not brute-force. That's precisely the kind of clever trick we'd expect from a skilled human player. The whole piece strikes me as poorly considered. The thoughts of Dan Dennett are far more worthwhile. If you've read this far, you may enjoy a quick read of the Wikipedia article on Dennett's intentional stance. [0] Dennett also makes the relevant observation that mankind is constantly creating machines with general intelligence and moral accountability: other human beings. This ties back to the article's points about moral accountability. [0] https://en.wikipedia.org/wiki/Intentional_stance https://en.wikipedia.org/wiki/Intentional_stance