How to create back propogation in a robot algorithm

[Jalopy]

Back propogation, according to Prometheus in another thread, is the human ability to learn. A feature that artificial intelligence has not been able to create in robots. 

Maybe the solution is to program robots to learn through value-computation. 

To teach a robot how to learn new data, it's a matter of programming the robot to process data in its envioronment according to a pyramid of value. 

For example, factor A = value 1 factor B = 2 points, factor C = 3 points

The robot would process each factor according to it's value, and assign priorities according to same. 

Thereby, the factors given most priority would be the most valuable ones, and the ones the robot would process as same. 

For example, teaching a robot to learn the alphabet;

The value of letter Z = 1 point, the value of letter Y is 2 points, etc.

The robot would automatically allocate the most value to letter A, 26 points. Then it would allocate the letter B the value of 25 points, the letter C 24 points, the letter D 23 points etc etc. 

If that process of allocation and computation were digitally displayed, it would be like the robot learning the alphabet. 

 

[Ghideon]

13 minutes ago, Jalopy said:

A feature that artificial intelligence has not been able to create in robots. 

Back propagation* is used in machine learning, for instance in neural networks. Are you claiming that none of these have been applied to robotics?

19 minutes ago, Jalopy said:

To teach a robot how to learn new data, it's a matter of programming the robot to process data in its envioronment according to a pyramid of value. 

For example, factor A = value 1 factor B = 2 points, factor C = 3 points

The proposed idea does not make much sense, sorry.

 

*) See for instance https://en.wikipedia.org/wiki/Backpropagation

 

 

[Prometheus]

17 hours ago, Jalopy said:

Back propogation, according to Prometheus in another thread, is the human ability to learn.

I said:

"Probably the biggest difference (between human and artificial intelligence) is that modern machine learning algorithms use back-propagation, whereas there is no such (known) mechanism in brains."

Hopefully what others have said here and there have cleared up that confusion for you.

17 hours ago, Jalopy said:

Maybe the solution is to program robots to learn through value-computation. 

Modern machine learning algorithms have a function which calculates some distance (in a mathematical sense, not physically) between a predicted outcome and an actual outcome. It's sometimes called a reward function. The aim of the algorithm is to minimise the distance between its predictions and the actual event - it is 'rewarded' for getting its predictions as close to the event as possible.

As far as i can tell from your example of the alphabet you are proposing some kind of weighted reward function, preferring earlier letters than later ones for some reason. In that case the AI would just learn to always predict As, or maybe Es, as these are very common letters in English and you've assigned them a high value.

You should learn a little about the machine learning field: there are some excellent tutorials out there, but most assume some level of mathematics, for which there are also excellent tutorials.