Like Tree3Likes
  • 1 Post By Matticus
  • 2 Post By cyberfish

AI

This is a discussion on AI within the Tech Board forums, part of the Community Boards category; Tonight while I was eating my roasted capsicum I began to ponder the nature of Artificial Intelligence and the significant ...

  1. #1
    11DE784A SirPrattlepod's Avatar
    Join Date
    Aug 2013
    Posts
    485

    AI

    Tonight while I was eating my roasted capsicum I began to ponder the nature of Artificial Intelligence and the significant challenges that remain. I think that the first problem is that we don't really even have a definition of "intelligence". Let's look at the definition from intelligence: definition of intelligence in Oxford dictionary (British & World English)

    noun
    the ability to acquire and apply knowledge and skills:
    an eminent man of great intelligence
    [count noun] a person or being with the ability to acquire and apply knowledge and skills:
    extraterrestrial intelligences
    I have no idea why "extraterrestrial intelligences" are included in this definition, but it's suffice to say that that subject is far beyond the scope if this post (or, perhaps not).

    I think that everyone would agree that the definition is pretty broad. It's too broad.

    Creation of AI has been pretty much at a standstill for the last 40 or more years in my opinion. Whilst eating my capsicum (these are red bell peppers) I realised that maybe the most direct approach to creating AI would be not to concentrate on AI at all! Maybe we have to concentrate on implementing a complete simulation of the universe and then AI would, after a while, "emerge". I really think this might work. How can it not work when this is how the real universe ended up with intelligence?

    So, instead of creating AI we could achieve the same thing by ignoring the hard problem of AI and instead create a deatiled model of the universe and the AI problem will solve itself!

    Has anyone else been working along these lines?

  2. #2
    Registered User
    Join Date
    Jun 2011
    Posts
    2,568
    Interesting idea. Creating a complex simulation where intelligence became an emergent result, however, sounds a lot more difficult that simply creating AI, in my opinion. There are just too many variables in the equation that we cannot understand, much less perceive.

    Also, it would seem that the computational power required to achieve such a simulation is well beyond our abilities thus far, and much greater than required by AI itself.

    I'm certainly no expert, nor even a casual follower, of AI development, so I can only offer opinions that are no doubt draped in ignorance. But the way I figure it, the problem with either approach is that we are bound within our minds, and therefore can't examine them "from the outside" to get a true understanding of how they work.

    Good food for thought, though.
    SirPrattlepod likes this.

  3. #3
    Registered User
    Join Date
    Dec 2006
    Location
    Canada
    Posts
    3,183
    Modern AI is practically a loosely defined group of algorithms (searches, heuristics, machine learning, natural language processing, etc) that are useful in many applications that require some kind of decision making. AI applications are created by tailoring those algorithms to the specific needs of each application.

    I definitely wouldn't say AI has been at a standstill. It's probably one of the most exciting branches of computer science where important discoveries and inventions are being made weekly. Just go youtube "AI".

    There is still no practical general purpose AI, yet. Most modern developments are still very application-specific.

    Simulating the universe is a problem orders of magnitudes harder than creating a general purpose AI.

  4. #4
    Registered User MutantJohn's Avatar
    Join Date
    Feb 2013
    Posts
    1,191
    Simulating the universe is actually relatively easy. We have all the equations down on paper, including quantum mechanical ones too. However, actually running those simulation is impossible. Namely, I don't think there's even a simulation that could handle a mole of molecules (10^23 particles) and I think the most we've ever simulated was in the low trillions. The problem is time and RAM. I doubt there's a cluster large enough to could hold all the data and the time it would take to run is pretty astronomical.

    It's okay, quantum computing will fix it, probably, assuming you can use just one atom to store data or run a thread.

  5. #5
    Registered User
    Join Date
    Dec 2006
    Location
    Canada
    Posts
    3,183
    It's theoretically impossible. How do you simulate an universe that includes a computer that is simulating the universe?

  6. #6
    Registered User MutantJohn's Avatar
    Join Date
    Feb 2013
    Posts
    1,191
    Dude, that is soooooooooooooo meta.

    Simulations of physics are simple in the sense that all you're really doing is evaluating equations.

    Let's just consider classical physics for the time being.

    To simulate a collection of atoms all you really have to do is calculate the electric interactions between them and apply that to their motions as that's what creates forces between them. So, you'd have the classical view of the atom which is the circular electron clouds with hard sphere electrons. There's voltage and field equations so you can apply that calculating forces and then velocities, etc.

    If you want, you can even include chemistry and have atoms form molecules when certain conditions are met.

    Considering we have a good chunk of science completed, it is entirely possible to simulate the equatons, using the proper structures.

    Like, an electron for example would be like this :
    Code:
    struct electron {
    
       unsigned mass;
    
       float x, y, z; // position
    
       float vx, vy, vz; // velocity
    
       float ax, ay, az; // acceleration
    
       float charge;
    
       float kinetic_energy, potential_energy;
    };
    and I think that encompasses most of a classical electron.

    Using structures to represent particles should allow you simulate physics. The real problem is just that it's not feasible considering current technology.

    Edit : Plus, aren't you like an EE/CE kind of guy? You should know how computers are actually built. Just write code that would simulate that. So it'd be current flows and all that good stuff.

  7. #7
    Registered User manasij7479's Avatar
    Join Date
    Feb 2011
    Location
    Kolkata@India
    Posts
    2,498
    Well, to be fair, we don't know what to simulate either.
    Simulating classical mechanics would hardly be something new.

    And once you get to a lower level of abstraction, simulation does not make sense because of natural limits.
    For example, you could never simulate an electron correctly....(I think).
    Manasij Mukherjee | gcc-4.8.2 @Arch Linux
    Slow and Steady wins the race... if and only if :
    1.None of the other participants are fast and steady.
    2.The fast and unsteady suddenly falls asleep while running !



  8. #8
    Registered User
    Join Date
    Dec 2006
    Location
    Canada
    Posts
    3,183
    Edit : Plus, aren't you like an EE/CE kind of guy? You should know how computers are actually built. Just write code that would simulate that. So it'd be current flows and all that good stuff.
    That's how they verify CPU and GPU designs before committing to silicon.

    NVIDIA GTX 680, when simulated by a supercomputer, produces about 1 frame per day.

  9. #9
    Registered User MutantJohn's Avatar
    Join Date
    Feb 2013
    Posts
    1,191
    Quote Originally Posted by manasij7479 View Post
    Well, to be fair, we don't know what to simulate either.
    Simulating classical mechanics would hardly be something new.

    And once you get to a lower level of abstraction, simulation does not make sense because of natural limits.
    For example, you could never simulate an electron correctly....(I think).
    Well, you're not exactly simulating an electron. You're just solving the equations that govern them. All you would be doing is math, essentially, but that's assuming the universe is fundamentally representable with math which is a fair assumption, considering physics as it is now.

    Also, cyber, that's really cool. Man, I wish I had a 680. Stupid 460...

    Jk, my gpu is the best part of my computer. Stupid AMD processor, no one likes you! (I secretly do though)

  10. #10
    Registered User
    Join Date
    Dec 2006
    Location
    Canada
    Posts
    3,183
    I don't have a 680. I do have a 670, though.

    Don't really need it since I don't play games that much anymore. Just got it for sentimental value (I worked for NVIDIA and was involved in launching that chip, GK104).

Popular pages Recent additions subscribe to a feed

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21