PDA

View Full Version : Modelling concepts, ideas etc in the same way as objects(robotics)



WaxRubiks
2018-Aug-25, 11:19 AM
I wonder how far it would be possible to model abstract concepts, ideas, plans etc in the same way as physical objects could be modelled in a computer system, in order for robots to be able to do more sophisticated things.

can't thing of an example at the moment....*runs away*

Maybe 'actions' could be treated as objects eg a car journey becomes an object, with certain properties, and maybe those properties also are treated as objects...if that makes sense...if you could build a hierarchical system like this maybe only very simple actions need to be treated as actions/processes, which perhaps a computer system could cope with.

Ken G
2018-Aug-25, 12:45 PM
Given the prevalance of machine learning these days, my guess is the days of computer programming for complex AI processes are almost over. Instead of programming computers to think, we will simply teach them to, the way we teach human brains to do it. In that analogy, programming will be restricted to tasks we associate with instinct in animals, whereas higher analysis capabilities will simply be taught by "experience." If so, you won't have to program a computer to recognize objects or actions, you can just wait for them to figure it out on their own-- like people do.

WaxRubiks
2018-Aug-25, 01:10 PM
Given the prevalance of machine learning these days, my guess is the days of computer programming for complex AI processes are almost over. Instead of programming computers to think, we will simply teach them to, the way we teach human brains to do it. In that analogy, programming will be restricted to tasks we associate with instinct in animals, whereas higher analysis capabilities will simply be taught by "experience." If so, you won't have to program a computer to recognize objects or actions, you can just wait for them to figure it out on their own-- like people do.


my belief is that biological neural networks is only part of what is going on in the brain, so I think there may be limited success with robot neural networks, and that they might not be capable of learning very complex behaviour, but I could be wrong.

It might be easier to apply a brute force approach to robot thinking, using present digital processes. Might be on a wild goose chase trying to get neural networks in robots to work well enough to get them to be useful, plus they wouldn't need such a learning process, so easier to get them out of the factory and doing work.