nasa and 15year old nerds are at the same level of building robots i haven;t seen anything different in robotics not from japan or a kid building his own personal robots its all algorithms,they should make an A.I that adapts to its surroundings though learning
Isn't it possible to evolve the software in a simulated 3D environment? And once the software is evolved right, input it back in the robot?
I am saying that because the software can evolve much faster in a simulated environment than in the real world. Of course i guess you would have to model the robot in the right proportions and all, but i don't see why it would be a problem.
With that, you could use distributed computing to evolve the software even faster.
@StormWolf01 That works for some tasks, however for many tasks writing an accurate simulator would be harder than actually learning in the real world. I think you are right that in this particular task the learning could be done in simulation.
@bigpauly2000 The simulator doesn't have to be perfectly (or even extremely) accurate - I saw a paper a while back where Q-learning was used first in a simplified model, then in a real environment, to jump start the learning process (which is usually very slow with Q-learning for complicated state spaces/tasks).
There was an analogy that even a totally novice driver rapidly learns that turning the wheel left makes the car turn left, even if the degree/speed of the turn are not known to them.
At least she didn't smudge her lipstick!
filmmaker91361 1 month ago
When will there be a pleasure model available to buy?
Finkie147 8 months ago
what people dont know is that the ball was hot and the poor robot was burning its hand! XD
cachinchilla3000 1 year ago 5
nasa and 15year old nerds are at the same level of building robots i haven;t seen anything different in robotics not from japan or a kid building his own personal robots its all algorithms,they should make an A.I that adapts to its surroundings though learning
fashanu1000 1 year ago
@fashanu1000 they have been trying that for 40 years...and getting no where.
astrialkil 7 months ago
it's like "ooooo, poke hehehehehehe, Poke! hehehehehe, Poke!"
DocUnsane 1 year ago
Isn't it possible to evolve the software in a simulated 3D environment? And once the software is evolved right, input it back in the robot?
I am saying that because the software can evolve much faster in a simulated environment than in the real world. Of course i guess you would have to model the robot in the right proportions and all, but i don't see why it would be a problem.
With that, you could use distributed computing to evolve the software even faster.
StormWolf01 1 year ago
@StormWolf01 That works for some tasks, however for many tasks writing an accurate simulator would be harder than actually learning in the real world. I think you are right that in this particular task the learning could be done in simulation.
bigpauly2000 1 year ago
@bigpauly2000 The simulator doesn't have to be perfectly (or even extremely) accurate - I saw a paper a while back where Q-learning was used first in a simplified model, then in a real environment, to jump start the learning process (which is usually very slow with Q-learning for complicated state spaces/tasks).
There was an analogy that even a totally novice driver rapidly learns that turning the wheel left makes the car turn left, even if the degree/speed of the turn are not known to them.
electroum 1 year ago
I think it looks awesome!
It looks like the arms our working.
McClover 3 years ago
Well it looks stupid with is arms like that and it couldnt pick it up but it was cool other than that
TheHaloholic 3 years ago
Cool hehe
DeppObsession1963 4 years ago 3