Comments by "Taxtro" (@MrCmon113) on "Computerphile" channel.

  1. No. The brain being a physical object is why it's not a Turing machine, which are conceptual entities. Further you are confusing computability of a problem with computability of an INSTANCE of a problem. It might be trivially easy to see whether a TM halts on an input, but you cannot formulate an algorithm that always tells you whether any TM halts with any input.
    1
  2. Jacen Solo No. Your notion of "purpose" is incoherent and I find it unlikely that an AI would confuse itself in the same way as your brain. The ultimate reason why you value the contact with females is reproduction, that doesn't keep you from wearing a condom. You are motivated by your goals, not by some unspecified link in the causal change, that led to you having these goals.
    1
  3. Yeah, so? That's true for any program accepting user input. And any function is only properly defined by giving the two sets it maps from and to.
    1
  4. A Turing machine is a kind of automaton.
    1
  5. Firstly people strongly disagree about what's wrong or right. Secondly an AGI must have BETTER moral judgment than any human. Thirdly the first AGI studying human psychlogy would already lead to a computronium shockwave as it tries to learn everything about it's goal and whether it reached it.
    1
  6. Today's targeting systems and robots are already superior to humans. And why shouldn't the AI make humans fight for it?
    1
  7. As far as I know Quantum computers are still theoretical, but it was proven that they are just as powerful as Turing machines. Also you should know that there is nothing special about that model in particular. GOTO (if statements), WHILE programms, lambda calculus and many other things are just as powerful as Turing machines and quantum computers as well. As for automata, which are less powerful than Turing machines, those are finite automata or pushdown automata for example.
    1
  8. Yeah it has been proven that quantum computers are no more powerful than Turing machines; either the guy in the video didn't know about it or he left it out for some reason.
    1
  9. starofcctv94 Actually that's not the problem. With humans it is exactly the same. You don't just wake up one morning an realize that you don't have to increase the well being of conscious entities.
    1
  10. starofcctv94 That is circular thinking. An AI without goals does nothing. There is no fitness or preference function.
    1
  11. Neo Theone Yes, you can. Neurons are slower than transistors. Axons are slower than glasfiber.
    1
  12. sexyloser Because "pleasing it's creator" is a goal, you just made up. You might as well ask why the stamp collector doesn't start building paperclips instead or join jihad.
    1
  13. bob dole Well anything but a perfect moral actor would essentially be an ultimate stamp collector.
    1
  14. Mate, quit bullshitting. You don't need to understand any algebra to understand what a recursive definition is. The only thing you need is the natural numbers.
    1
  15. misium No, no, no. That makes no sense. Why is no one in the comments able to think clearly about very simple basics of motivation? The hypothetical stamp collector is motivated only to increase the number of stamps in the universe. Not by what you value. Is it that difficult for you guys to change perspective? You think that everything thinks like you, even if it was specified, that it doesn't.
    1
  16. How so? They are pretty straight forward.
    1
  17. "Let's talk about convoluted neural networks"... XD
    1
  18. Now I'm really looking forwards to my next semester.
    1
  19. Jmac217 Huh what? That's a perfectly fine way of explaining things.
    1
  20. Ok so you have to tell me what distribution produced the following input-output pair: A -> 0 Do you think your chances of guessing the right function improve if I give you more examples? If not, why do you think learning is even possible?
    1
  21. Automata are intuitive, so I'd just say that a regular expression is a state machine.
    1
  22. Everything he described so far you have to do no matter whether your methods are simple statistics or large ML models.
    1
  23. What's wrong with Zuckerberg? What nerds are you talking about anyways?
    1
  24. How do you even write so much without knowing ANYTHING about autonomous driving?! XD No, we are not using tracks; they don't need better breaks. The problem is not routing, but machine vision.
    1
  25. Information is just the complexity of the data. What you are looking for is knowledge.
    1
  26. If the numbers were labeled with information about the world that allows inference of the units, yes. But what would be the point?
    1
  27. There is RNNs far more powerful than Donald Trump. I think he only memorizes his last three words.
    1
  28. The size of the data matters a lot. Some things you can only learn from incredibly huge sets of data.
    1
  29. As a computer engineer you should have most of the maths background already. What comes up over and over again is probability theory and linear algebra.
    1
  30. This can not continue. This can not continue. This cannot continue.
    1
  31. You mean the generator. And no, in principle the system converges to a point where the generated samples lie within the manifold of true samples, ie there is no true difference between them.
    1
  32. Yeah, dunno why he says cards. States is much more obvious.
    1
  33. I think the video was somewhat misleading. Turing machines aren't really relevant to technology, but rather to theoretical considerations. A Turing machine is a very simple way to describe a very powerful system, but there are much better ways to implement most (interesting) functionalities in practise.
    1
  34.  @sneedis9909  You just pretend like those functionalities aren't there. You can, for example, implement a tree using an array in Java.
    1
  35. Of course! And that's the problem: Distinguising small animals from leaves and features of the road. The problem with autonomous driving is machine vision.
    1
  36. That answer can only come from someone, who doesn't understand what intelligence means.
    1
  37. Those programs are translated into assembler code, which is itself Turing complete. I think it is closely related to GOTO statements, which are also just as powerful as Turing machines.
    1
  38. That's pretty much the point of machine learning. Our brain cannot process as much information as quickly as a computer or remember as much of it.
    1