The evolution of robotics

I have a dream. A dream that a robot will no longer be programed in general terms, but instead, it will be taught how to learn from another source. This source may be anything, the internet, other robots, humans, but the main factor being the robot is only firstly programmed to understand these sources by a human, and is then left on its own. My application of this would be to a game of soccer which I find is most easily demonstrated.

Currently, robots do not really have a concept of what a ball is, what a goal is, how to score a goal, what is a robot, what is soccer even. They are programmed in flat code to make desisions based on the inputs they recieve from their sensors, and although this may be good...it has flaws. The program is not adaptable during games, and it does not truly evolve, cause the robot does not learn from experiences. It may setup a really good play by chance, but never repeat the actions it took to get to that position as it does not remember.

What if a robot was taught how to interpret a text, like a webpage, much like a human does? This would open an entirely new avenue. A robot would learn to understand what a ball is, how to score a goal, different stratigems to make its way past competitors, without a user having to tell it anything at all. As it plays more and more games, it will become accostmed to what works in a game and what doesnt, taking the human error factor out of programming as it is able to find out for itself.

Is this dream a possible reality? Or is it just pipe?

Reply to
fr335tyl3r
Loading thread data ...

You invented a human !

fr335tyl3r

Reply to
Martial Chateauvieux

You and me both, I think about this all the time. Doesn't this pretty much fall under the realm of AI? People are working on this stuff, but I really don't know how close they're getting. And the definitions of what is AI become very subjective and opinionated. I'm pretty sure people have already created robots that build tables of data as they roam about. Future decisions the robot makes are then sequential in nature, not only a function of the present inputs but past history as well. Is this memory? Does it qualify as learning from past experiences? I'm not sure, but I think it's pretty cool and admittedly none of my robots are this sophisticated but I'll get there.

When robots are able to learn from past history, and alter their own programming to improve upon themselves formulating their own new ideas and solutions to problems then I'll be more likely to think they're more than just a collection of sensors, electronics and micros doing what they've only been programmed to do, sadly that describes my creations to a tee :-(.

-Dave

"fr335tyl3r optusnet dot au dot com>" I have a dream. A dream that a robot will no longer be programed in general

Reply to
Dave

Freestyler: The scenario you describe of a computer learning from the traditional sources we are familiar with, e.g. books, videos etc. , is probably still pretty far off. (Put's on asbestos suit. ) But learning AI's are in use, develpoment, and study all over the place. I don't know if anyone has tried to apply these techniques to robo-soccer, but it seems to me that at least a couple of the current techniques for computer learning might be fun to try on this game. I think pehaps the major obstacle to this approach to robo-soccer might be the cost. In order to train a neural net, or to evolve a genetic algorithm to play robo-soccer would require a workspace with a table or field to play on, along with enough competent robo-soccer playing bots to play a regulation game over and over and over and....... Then you have to consider the unexpected, the incompent robot that for some reason seems to have your evolved or trained alogorithm's number. It can't take into account any scenario that it hasn't seen before. It can train against the best from last year, but maybe this year there is that one bot that does something totally illogical or unforseen by your bots training/evolving , what does your bot do then? IMHO this point is where the real crux of the problem lies. By solving the original problem of creating a bot that learns from it's past experiences, you lose some of it's ability to deal with the unexpected. The flat progammmed bot you mentioned is not bothered by this, it plays the game regardless, it doesn't have any "expectations".

"Dave" wrote in message news: snipped-for-privacy@buckeye-express.com...

Reply to
Edward Dunaway

"fr335tyl3r optusnet dot au dot com>" I have a dream. A dream that a robot will no longer be programed in general

[Zagan] I don't think this is a pipe dream, but we have to remember that we don't really know in sufficient detail how the human brain works. It seems that a lot of effort is ongoing to create artificial intelligence based on the human brain. I can't argue that this effort is pointless, but I do have to wonder why human based intelligence is required. I don't know when it will happen or even how it will happen, but I do expect that machine intelligence will emerge sometime in the future, and I don't expect that this intelligence will be based on human intelligence.

It is my belief that machine intelligence will evolve on it's own terms and that human intelligence will not be required. It took nature billions of years for nature to create human intelligence. Can we really expect to reverse engineer this process? It is my expectation that machine intelligence will manifest in the near future, but it will (in a sense) create itself. It will be by our own efforts initially, but we will soon be removed from the equation and machine intelligence will operate by its own rules. When this happens we will not understand this intelligence any more than we understand our own.

My 2 cents.

// Jim

Reply to
Zagan

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.