Page 6 of 6 FirstFirst ... 456
Results 101 to 118 of 118
  1. #101
    A reward. It uses displacement score. This ranks actions.



  2. #102
    Originally Posted by ADVANCESSSS View Post
    A reward. It uses displacement score. This ranks actions.
    The rewards part is pretty obvious for such a scenario. My question is how it learns and how it explores its action space.



  3. #103
    It generates random actions for all 4 motors. When it selects a saved set for the 4 motors and does learned action however, then it will tweak them instead of a full generation.



  4. #104
    Originally Posted by ADVANCESSSS View Post
    It generates random actions for all 4 motors. When it selects a saved set for the 4 motors and does learned action however, then it will tweak them instead of a full generation.
    The whole thing together sounds like reinforcement learning to me.



  5. #105
    do you want it to learn by imitating people?



  6. #106
    a ai baby could be programmed to know how many motors and limbs in order to find out if it could perform the movement.And compare it to the video of a person moving.If it could perform the movement then it would.



  7. #107
    Originally Posted by Lostscience View Post
    a ai baby could be programmed to know how many motors and limbs in order to find out if it could perform the movement.And compare it to the video of a person moving.If it could perform the movement then it would.
    It would also not need an explicit goal to be learned. It might have learned that something to eat is at a certain location and would learn to move because of it. But it would also sense the environment completely differently, e.g. through seeing, touching, hearing, smelling.
    What is demonstrated hear is a simple reinforcement learning approach, which is far simpler than the initial idea as I understood it. Driving a few motors through reinforcement learning is a well understood and solved problem.



  8. #108
    what would make it hard is that it would have to get up from laying down and do the action.but that could be solved by it knowing what the top and bottom are of a human being are.And then trying to get into that orientation.



  9. #109
    I wonder what Google Brain is up to next. They got their AIs to run to destinations and overcoming obstacles. But then what lol? When is it gonna go to food and put it in its mouth? Or be taught what science is instead of just doing animal things (food/sex)? Human activities passed food/sex primitives are gonna be more revolutionary.



  10. #110
    Member BluePrintRandom's Avatar
    Join Date
    Jul 2008
    Location
    NoCal Usa
    Posts
    18,786
    Break it and remake it - Wrectified
    If you cut off a head, the hydra grows back two.
    "headless upbge"



  11. #111
    If you think that's creepy you haven't seen nothin.



  12. #112
    Originally Posted by ADVANCESSSS View Post
    I wonder what Google Brain is up to next. They got their AIs to run to destinations and overcoming obstacles. But then what lol? When is it gonna go to food and put it in its mouth? Or be taught what science is instead of just doing animal things (food/sex)? Human activities passed food/sex primitives are gonna be more revolutionary.
    One of DeepMind's goals is to learn traits that are typical for mice.



  13. #113
    Oh for the love of...
    Google Brain is a neural network. So are you or aren't you making a neural network? You are comparing it to an NN...



  14. #114
    Lol. I'm trying to avoid NNs for now essentially. Google uses NNs for all their AIs, I was just curious about what goals they're onto next.



  15. #115
    Member SterlingRoth's Avatar
    Join Date
    Mar 2006
    Location
    Portland, OR
    Posts
    2,155
    Why are you trying to avoid NNs for now? It seems like if you want to produce groundbreaking research, you should be using cutting edge technology.



  16. #116
    I believe he should do it the way he is doing it for now.i would like to see what he comes up with.



  17. #117
    Originally Posted by Lostscience View Post
    I believe he should do it the way he is doing it for now.i would like to see what he comes up with.
    That's the thing; he's not coming up with it. He's paying someone else to make the ai scripts and from what I can read in the previous comments, he doesn't seem to have an understanding of how the scripts work.
    Besides "be groundbreaking" he doesn't seem to have any real goals either...
    Altogether this whole thing is... Weird. Really really weird.



  18. #118
    I hired a programmer, and told him instructions. I...know how the code works exactly. Though I don't know programming languages.



Page 6 of 6 FirstFirst ... 456

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •