5 ms·

So I’m fine with AI, because I don’t believe in it

18 points by joeyespo 12y ago

danieltillett 12y ago
Lets change AI to HIV - I'm fine with HIV because I don't believe in it.
- onion2k 12y ago
  This is a fun game. I'll play... I'm fine with unicorns because I don't believe in them. HIV and AI aren't analogous. HIV exists. We have lots of evidence for its existence. It's not speculative. AI (in the "sentient machine" singularity sense) doesn't exist, but it might in future. Consequently, right now, stating you don't believe AI will ever happen is fine. It's stupid, because pretty much every open-ended prediction is stupid, but it's not "wrong".
  - danieltillett 12y ago
    Not believing in something is not evidence for or against something. Saying you don't believe AI is possible for reasons X, Y or Z is of value, but saying you a fine with something just because you don't believe in it is meaningless.
    - drdeca 12y ago
      I'm not sure I follow. It may be useful in some cases to describe why you believe or disbelieve something (as you say). I do not see why it would never be useful to state that because you believe or disbelieve something, you have a certain emotional reaction to some thing. It seems like a potentially useful description of why one has or chooses a given emotional response. Even if it so happens that the belief leading to the emotional response is wrong, it could still be useful to describe the relationship between the belief and the emotions or feelings or opinions on policy. For example, suppose I believe that all insects have a significant chance of poisoning food they land on. In this case, might not it be useful to know that the reason I am not ok with flies is because of this belief? So even if the person is wrong about their beliefs about AI, I don't think their explanation of why they are OK with AI as a result is "meaningless".
      - danieltillett 12y ago
        The reason why is the same as I originally stated, reality does not care what you believe. Either AI is possible (or not) and it does not matter if you believe in it or not. To put it more bluntly it is meaningless what you believe, the only thing that matters is why you believe what you believe.
        drdeca 12y ago
        Maybe I do not understand what you mean by meaningless? Beliefs can have consequences that do not depend on the cause of the belief. The actions one takes depend on one's beliefs more directly than they do ones reasons for those beliefs, in many cases. (For the insect+food example, consider the options of taking it as axiomatic on a whim, or being made to believe it via propaganda+magic(magic here just for sake of argument). What causes the person to believe the thing does not effect whether the person intends to allow a fly to land on their food.) (If I believe my hand is on fire, whether I attempt to remedy this does not depend on whether I see it or feel it, nor on whether it is on fire. If I attempt to remedy it, while it is not on fire, there may be an unpleasant result. The truth does of course matter, but beliefs are a useful predictor of future actions.)
codr4life 12y ago
We're already seeing certain aspects of consciousness simulated using said algorithms. Nobody is claiming to be able to do the whole shebang as far as I know. It's black box reverse engineering at it's finest. I'm pretty sure that we'll never reach individualization though, as that implies something that not all agree has anything to do with the brain. And a computer is all about the mental level, logics, trying to model feelings will never work out. Close enough to the uncanny valley to make us uncomfortable, sure. Bot not the real thing as it has nothing to do with logics.
- codr4life 12y ago
  I sometimes feel like we're wasting time trying to create the next generation of Clippy, that the quest to reduce human consciousness to a series of yes- and no-questions is misguided. If there's anything we've been primed to for millions of years, it's complex social relations, to be able to tell someone who wants to kill you from someone who doesn't, someone who hates you from someone who loves you, telling the truth from lying, pure logics from a human response. Maybe we should focus more on the things that a computer excels at and accept that it will never be our secret friend but rather a tool? Get a dog for gods sake :-)
  - codr4life 12y ago
    I guess what I'm saying is the quest for a general purpose AI is probably a dead end. Specific AI's for specific tasks is by far the most successful approach so far.
    - donkeyd 12y ago
      If you start putting multiple 'Specific AI's' together, as is common in CS, you will eventually get a more general purpose AI. The only constrain is time, with enough time, things that we can't even imagine right now will be possible. With the Internet of Things I can imagine that distributed AI will become a thing. Interconnectivity will eventually connect all the 'specific purpose AI' to become a 'general purpose AI'. I could keep thinking about this for hours, but that won't be good for my productivity...
kolev 12y ago
Not believing in something does not disprove its existence.
PhasmaFelis 12y ago
Kim Stanley Robinson is, FYI, an amazing sci-fi author, one of the best we've got going right now, and if you haven't read his stuff already you should. But I can't agree with him on this particular thing. It seems like a rehashing of the same old religious argument, that there's some magical, ineffable, indefinable property of savannah-grown meatbrains that no manufactured device can ever, ever have. I can't buy that. Human brains are just an arrangement of matter, and any arrangement of matter can be replicated. Whether we will make conscious machines, I can't say, but there's no rational way to assert that it can't be done.
wyager 12y ago
What an idiotic sentiment. We have no reason to believe that the entire universe isn't computable on a (Quantum) Turing Machine, or that any finite subset of the universe isn't computable on a (Quantum) Finite State Machine. The most brute-force approach is simply to make a very detailed scan of a brain and run it in a chemical simulator. This is extremely difficult, but not computationally or physically intractable. A project to exercise this method on a simple creature (C. elegans) is already underway. Recommended reading: http://en.wikipedia.org/wiki/OpenWorm http://en.wikipedia.org/wiki/OpenWorm http://en.wikipedia.org/wiki/Turing_completeness http://en.wikipedia.org/wiki/Turing_completeness http://en.wikipedia.org/wiki/Church–Turing_thesis http://en.wikipedia.org/wiki/Church–Turing_thesis http://www.cs.berkeley.edu/~christos/classics/Deutsch_quantum_theory.pdf http://www.cs.berkeley.edu/~christos/classics/Deutsch_quantu...
BrainInAJar 12y ago
The problem here is we don't have a very fundamentally good theory of mind. AI will come from philosophy departments, not CS.

18 points by joeyespo 12y ago

danieltillett 12y ago

Lets change AI to HIV - I'm fine with HIV because I don't believe in it.

onion2k 12y ago
This is a fun game. I'll play... I'm fine with unicorns because I don't believe in them. HIV and AI aren't analogous. HIV exists. We have lots of evidence for its existence. It's not speculative. AI (in the "sentient machine" singularity sense) doesn't exist, but it might in future. Consequently, right now, stating you don't believe AI will ever happen is fine. It's stupid, because pretty much every open-ended prediction is stupid, but it's not "wrong".
- danieltillett 12y ago
  Not believing in something is not evidence for or against something. Saying you don't believe AI is possible for reasons X, Y or Z is of value, but saying you a fine with something just because you don't believe in it is meaningless.
  - drdeca 12y ago
    I'm not sure I follow. It may be useful in some cases to describe why you believe or disbelieve something (as you say). I do not see why it would never be useful to state that because you believe or disbelieve something, you have a certain emotional reaction to some thing. It seems like a potentially useful description of why one has or chooses a given emotional response. Even if it so happens that the belief leading to the emotional response is wrong, it could still be useful to describe the relationship between the belief and the emotions or feelings or opinions on policy. For example, suppose I believe that all insects have a significant chance of poisoning food they land on. In this case, might not it be useful to know that the reason I am not ok with flies is because of this belief? So even if the person is wrong about their beliefs about AI, I don't think their explanation of why they are OK with AI as a result is "meaningless".
    - danieltillett 12y ago
      The reason why is the same as I originally stated, reality does not care what you believe. Either AI is possible (or not) and it does not matter if you believe in it or not. To put it more bluntly it is meaningless what you believe, the only thing that matters is why you believe what you believe.
      - drdeca 12y ago
        Maybe I do not understand what you mean by meaningless? Beliefs can have consequences that do not depend on the cause of the belief. The actions one takes depend on one's beliefs more directly than they do ones reasons for those beliefs, in many cases. (For the insect+food example, consider the options of taking it as axiomatic on a whim, or being made to believe it via propaganda+magic(magic here just for sake of argument). What causes the person to believe the thing does not effect whether the person intends to allow a fly to land on their food.) (If I believe my hand is on fire, whether I attempt to remedy this does not depend on whether I see it or feel it, nor on whether it is on fire. If I attempt to remedy it, while it is not on fire, there may be an unpleasant result. The truth does of course matter, but beliefs are a useful predictor of future actions.)

codr4life 12y ago

We're already seeing certain aspects of consciousness simulated using said algorithms. Nobody is claiming to be able to do the whole shebang as far as I know. It's black box reverse engineering at it's finest. I'm pretty sure that we'll never reach individualization though, as that implies something that not all agree has anything to do with the brain. And a computer is all about the mental level, logics, trying to model feelings will never work out. Close enough to the uncanny valley to make us uncomfortable, sure. Bot not the real thing as it has nothing to do with logics.

codr4life 12y ago
I sometimes feel like we're wasting time trying to create the next generation of Clippy, that the quest to reduce human consciousness to a series of yes- and no-questions is misguided. If there's anything we've been primed to for millions of years, it's complex social relations, to be able to tell someone who wants to kill you from someone who doesn't, someone who hates you from someone who loves you, telling the truth from lying, pure logics from a human response. Maybe we should focus more on the things that a computer excels at and accept that it will never be our secret friend but rather a tool? Get a dog for gods sake :-)
- codr4life 12y ago
  I guess what I'm saying is the quest for a general purpose AI is probably a dead end. Specific AI's for specific tasks is by far the most successful approach so far.
  - donkeyd 12y ago
    If you start putting multiple 'Specific AI's' together, as is common in CS, you will eventually get a more general purpose AI. The only constrain is time, with enough time, things that we can't even imagine right now will be possible. With the Internet of Things I can imagine that distributed AI will become a thing. Interconnectivity will eventually connect all the 'specific purpose AI' to become a 'general purpose AI'. I could keep thinking about this for hours, but that won't be good for my productivity...

kolev 12y ago

Not believing in something does not disprove its existence.

PhasmaFelis 12y ago

Kim Stanley Robinson is, FYI, an amazing sci-fi author, one of the best we've got going right now, and if you haven't read his stuff already you should. But I can't agree with him on this particular thing. It seems like a rehashing of the same old religious argument, that there's some magical, ineffable, indefinable property of savannah-grown meatbrains that no manufactured device can ever, ever have. I can't buy that. Human brains are just an arrangement of matter, and any arrangement of matter can be replicated. Whether we will make conscious machines, I can't say, but there's no rational way to assert that it can't be done.

wyager 12y ago

What an idiotic sentiment. We have no reason to believe that the entire universe isn't computable on a (Quantum) Turing Machine, or that any finite subset of the universe isn't computable on a (Quantum) Finite State Machine. The most brute-force approach is simply to make a very detailed scan of a brain and run it in a chemical simulator. This is extremely difficult, but not computationally or physically intractable. A project to exercise this method on a simple creature (C. elegans) is already underway. Recommended reading: http://en.wikipedia.org/wiki/OpenWorm http://en.wikipedia.org/wiki/OpenWorm http://en.wikipedia.org/wiki/Turing_completeness http://en.wikipedia.org/wiki/Turing_completeness http://en.wikipedia.org/wiki/Church–Turing_thesis http://en.wikipedia.org/wiki/Church–Turing_thesis http://www.cs.berkeley.edu/~christos/classics/Deutsch_quantum_theory.pdf http://www.cs.berkeley.edu/~christos/classics/Deutsch_quantu...

BrainInAJar 12y ago

The problem here is we don't have a very fundamentally good theory of mind. AI will come from philosophy departments, not CS.