it begins... (true AI finally coming soon)

0 favourites
From the Asset Store
Math questions appear quickly on the page, so you must respond quickly to get points!
  • AI, or artificial intelligence has come to mean something not like intelligent, but pretending to be intelligent, like npc's in games, the cpu players, and various other nonintelligent things

    artificial...it's intelligence, it's not real intelligence

    there is a new term, if you're into this kinda thing

    SI, or synthetic intelligence

    this is true intelligence, just made from nonnatural objects

    this doesn't exist yet. there's some really cool stuff out there and it's getting closer

    these two show we're ultra close, and honestly, I'm gonna nerd out for a second, but when it comes, it will change the world forever, in a matter of weeks

    something truly intelligent, like we are, not pretending to be, could just download knowledge without taking any time to learn, and remember without having to try, perform math at computer speeds (remember CPUs and GPUs are measured in TFlops nowadays I believe, that stands for trillions of floating point operations (decimal math problems) per second...that's f'in ridiculous), never fail to see a connection, never have something distract them, or forget to think of something. it will seriously invent anything we can think of, that's possible, in minutes, once it gets here....if anyone really cares, I've researched this alot, and we can talk further, but the way intelligence and brain cells work, if you understand it, once you make something that is truly as intelligent as a rat, it really would be no more effort than it takes you to tell construct to paste an array of 10,000,000 sprites (assuming a pc or construct wouldn't crash), to make something 10,000,000 times as intelligent as a rat. it would change the world overnight

    anyway, here's evidence we're not that far

    first, this is amazing:

    http://news.bbc.co.uk/2/hi/science/nature/7979113.stm

    second, this is also amazing, watch the whole video if you bother to watch the beginning, it only get's impressive later on, when it almost falls for various reasons, you can tell what it's doing is very much like what living things do:

    http://www.bostondynamics.com/content/s ... ion=BigDog

    your thoughts

  • Woohoo.. Terminator will be alive!

  • This is still just an artficial intelligence. You should get yourself interested in Neural Networks to find out that these tasks Adam's capable of, aren't really that innovative. Of course, costs of making such instrument are large, but that's because of big memories (well... maybe not so big, probably 0.000001% of brains memory capacity ^^) needed to handle simple tasks.

    Neural Network means a lot of math and it kinda looks like our neural system, but it's simplified. To very accurately simulate single real neuron's work, you'd have to evaluate thousands of differentional equations. Simulation on rat's brain is possible for supercomputers just for few seconds. We're nowhere near to making any real intelligence.

    If people were given such motoric abilities BigDog possesses, they'd also forget about falling on slippery places. But again, this stuff is "quite easy" (meaning there are at last 100 teams on Earth capable of creating it) to make an ideal model, but quite hard to get sponsors for it ^^.

  • This is still just an artficial intelligence.

    yeah

    SI, or synthetic intelligence

    this is true intelligence, just made from nonnatural objects

    this doesn't exist yet. there's some really cool stuff out there and it's getting closer

    but big dog is by far the most advanced walker I've seen,

    I think no one's hit the nail on the head

    but people are thinking in the right direction

    instead of making a robot programmed to deal with a variety of situations

    they make a robot that tries to find it's own solutions

    both of these cases are steps in the right direction

    but in all honesty

    I don't think we're slowly going to get there

    I think someones going to stumble upon exactly what makes intelligence work

    and then we'll be there

    I know about neural networks. not matter how large a traditional neural network is, it'll never do what we do, they're missing a crucial element

    I don't know what it is yet, or I'd have created an SI, but if you study enough it's clear neural networks aren't thinking (I know you weren't saying they were)

    Simulation on rat's brain is possible for supercomputers just for few seconds.

    what? if you mean they can simulate a rat's brain, I disagree, there actually is a project with supercomputers to simulate a single neuron, they're not even close to being able to simulate a rat's brain. especially since they don't know which processes matter

    If people were given such motoric abilities BigDog possesses, they'd also forget about falling on slippery places. But again, this stuff is "quite easy" (meaning there are at last 100 teams on Earth capable of creating it) to make an ideal model, but quite hard to get sponsors for it

    100 teams out of the population of the planet is far from easy

    but...yeah, I love this stuff

  • I know people having enough theorethical knowledge to do BigDog.

    BigDog is advanced walker, but don't bring any revolution to movement. It's a matter of time there's BigGepard invented hunting down "bad guys" and stuff.

    Adam did the work he was supposed to do. The only thing that could be considered innovative is that he probably had to guess by himself what kind of mathematical model (NN model) to use to solve problem. But from reading that short article I'm not sure if Adam is even able to think about using techniques to solve problem.

  • I still think human intelligent computers are waaaaaay off. Current processors barely dent the parallel computation power of the human brain, and even if you could match that, you need to understand how the neurons are connected, and then feed in the years of realistic experience that allow a person to mature in to an intelligent adult.

  • actually ash

    the thing is the brain uses electro-chemical impulses to transfer information

    I don't remember the exact figures, but it's at least several orders of magnitude slower than the electrical impulses in a microchip

    also the scale of a neuron is much larger than that of a microswitch, and the distance information has to travel before being processed again is another factor

    now I'm not saying SI will definitely emerge through software running on standard microprocessors, but the problem isn't speed, not even by a long shot

    it takes the brain about 6 to 8 steps to recognize a human face with near perfect accuracy

    an "AI" program that does the same with embarrassing inaccuracy takes hundreds of thousands of steps to do it badly

    there's something fundamentally different about what each is doing

    I think solving that problem is the obstacle. no matter how fast computers get, they won't think, unless we solve the puzzle

    but I think modern tech, is more than fast enough, if we could only find the damn algorithm

  • We still don't know why the brain is so good.

    And I love the Big Dog. I want one. With a silencer.

    Then I'll ride it into the sunset, hat spinning in hand. Yeehaw!

  • it takes the brain about 6 to 8 steps to recognize a human face with near perfect accuracy

    an "AI" program that does the same with embarrassing inaccuracy takes hundreds of thousands of steps to do it badly

    there's something fundamentally different about what each is doing

    I think solving that problem is the obstacle. no matter how fast computers get, they won't think, unless we solve the puzzle

    Perhaps the 6 to 8 steps you mention are just call-functions for "libraries", which have their own steps.

    As you maybe mean: more research is needed to understand how the brain fully works.

  • No, current tech isn't anywhere near fast enough. A neuron may be bigger and slower than a microchip, but there are millions of them communicating with each other all the time, simultaneously. For a chip which can only do one thing at a time, that's a tall order. Unless you have as many chips as there are neurons in the human brain, which is around 10 - 100 billion. Got that many cores on your Athlon?

  • Try Construct 3

    Develop games in your browser. Powerful, performant & highly capable.

    Try Now Construct 3 users don't see these ads
  • >

    > it takes the brain about 6 to 8 steps to recognize a human face with near perfect accuracy

    > an "AI" program that does the same with embarrassing inaccuracy takes hundreds of thousands of steps to do it badly

    > there's something fundamentally different about what each is doing

    > I think solving that problem is the obstacle. no matter how fast computers get, they won't think, unless we solve the puzzle

    >

    Perhaps the 6 to 8 steps you mention are just call-functions for "libraries", which have their own steps.

    As you maybe mean: more research is needed to understand how the brain fully works.

    In any case, the effort needed for a brain to match a face is clearly a ridiculous magnitude lower than that which is needed for a computer; I'm not sure realistically we can ever expect anything with the capabilities or realism of a human brain without recreating it entirely.

  • computers are way smarter,or way stupider than humans depending at how you look at it.

    a modern computer can do operations at a speed humans could never match, our brains dont perform calculations like that, they just cant.

    a computer is smart in that sense

    but a computer could never make abstract descisions or come up with ideas like a human mind can, it needs rules and order.

    emulating this type of intelligence isnt efficient, and doing it wouldnt be the right way to go about it. youd need to invent a new type of computer imo.

    comparing computers and brains is like trying to make a car walk, it cant really be done. its a task which needs a device designed to do it for it to be possible.

    a rat isnt extremely processy, running all the neurons and such through a computer is. a rat only thinks about food, water and safety, it will only need to interpret its vision, the only difficult task, which cant be done by computers right now since no easy way has been created.

    i dont know but AI is a long way off, autonomous robots arent ai, they just do predeterminded stuff based on a few situations that they know of. a real AI would need to form ideas for itself, which computers were never designed to do, all they can do is move numbers.

  • there was these techniques I read about with holograms (the old kind, where you needed a laser shooting through a piece of holographic film to view them), where you can take two holograms, and either by overlapping them, or bouncing lasers off them separately you could make a point of light appear that intensified depending on how similar the two images were, you could also make it project only the similarities, so if someone had grown a mustache in the second image, that section wouldn't appear, and you could also make it show only the dissimilarities. aside from setting everything up. that's a one step process, think of how many steps it would take a computer program to do something as simple as that).

    there are many other things the brain does that are similar to holograms, for instance. with the old holograms, you could cut them in half, and shoot the laser through either half of the film, and still get the whole image. the smaller the fragments get, the fuzzier the images get. this is similar to the brain, other than very rare types of brain damage and memory loss, if you injure your head, you don't forget the ages of 16 to 17, you either remember everything, or get a general fuzziness, or lose everything. you can also store ridiculous amounts of information in a small hologram and retrieve them by shifting the angle or position of the laser, or using different frequencies of lasers. the brain hold more information than it should given it's size. there are alot of other similarities. I didn't think of the brain as a hologram theory, there are many books about it, but it makes alot of sense to a degree.

    I don't think the brain is exactly like a hologram, but in the same way you can predict or simulate the path of a thrown object with a parabola much more easily than you can simulate the physics by trying to account for all of the forces of quantum physics, and all the particles involved from the ones in the ball and the air, to the ones in the planet creating most of the gravity pulling the ball, there are equations that describe holographic processes as well, namely the fourier transform, which also happens to describe or help describe alot of other cool scientific processes. so if we could find the right thing to emulate, I still computers are more than fast enough. whatever the brain is doing, it's slow (by computer processing standards) it's just running a much better written piece of software, I don't think emulating the right part of the hardware would be too slow even on a desktop pc, let alone a supercomputer.

    Anyway, I think figuring out what intelligence is is the main problem. if not, we're just going to be wasting processing power simulating things we don't need to , or making things that pretend to have intelligence better and better, but I think the more people make machines that solve their own problems, like the robots that formed theories they were not preprogrammed to form, and the big dog, which solves arbitrary balance problems as opposed to the vast majority of walking robots that have a preset way of walking that fails if the step is a few inches from where it was supposed to be. at least when engineers making a robot that tries to learn, they are bumping against the correct wall. I'd prefer to see a robot learn it's abc's from scratch, and fail at reaching a 2nd grade level of learning, than a robot that seems to carry on a real conversation, but is really just doing dictionary searches, sifting through grammar and logic trees, and later assembling response sentences from preconstructed parts. Even then I don't think anyone will stumble upon the right answer without first developing a working theory of what intelligence is, but as I said, at least they're trying to solve the right problem,

    how to BE intelligent

    not how to SEEM intelligent

    by the way, this project:

    http://news.bbc.co.uk/1/hi/programmes/c ... 165420.stm

    is simulating 8000 neurons on a supercomputer down to a molecular level, the brain has millions, so it isn't going to create intelligence, but in theory it would if it had millions of times more processing power

    their goal isn't to create intelligence, but to understand it. like I said, no one knows yet what intelligence is, there aren't even any generally accepted theories of it. we have no idea what's going on. don't even get me started on consciousness

  • en.wikipedia.org/wiki/Paragraph

    that's the meanest thing anyone's ever said to me.

    no, but seriously,

    I didn't realize how long that was until my supercomputer ran out of ram midtyping,

Jump to:
Active Users
There are 1 visitors browsing this topic (0 users and 1 guests)