Learning in the computer world is simply increasing the number of databases or their connections.

The learning is simply increasing the number of skills. And every single skill that person or the machine has is stored in the database. In those databases are listed the actions what solution needs for working. And if we are thinking that way. The skill is only a series of actions stored in the memory of the computer.  

The machines are learning similar way as humans are learning. The skills of the machines are the actions stored in databases. And the learning means that the new databases are forming in the memory of the computer. The selected solution can be automatic or it can require human acceptance.  If we think that the learning process is automatic. 

There must be something that limits the data mass stored in the memory of the computer. If the computer stores everything in its memory the data mass is rising very high. And that causes that the memory is not enough. Even the largest computers have a limited number of mass memory. So the parameters must determine what kind of data is stored in the long-term memory of the computer. 

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

The idea of the learning machines is that they can collect data. And then process that data to the form that benefits the system. The purpose of the system is to bring positive things to the creators or owners of the system.

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

The computer must have certain parameters that describe what kind of solution is stored in the computer. That denies the fill of the hard disks. 

So some values are describing the best possible solution. And if those parameters are filled the computer stores the solution in its database.  In this kind of system, the computer just makes the database, where is stored every single movement that the AI makes. If we are thinking that the AI is controlling the physical robot. Those values that describe the best solution for storing in the memory of the computer can be the time and energy use that the solution requires. 

So the computer is calculating the most economic and fast way to act. And if we are thinking the things like robot cars as the actor in this process. We can return to the modular learning method. If the AI has an access to the other robot car's databases, it can ask for advice from them. That AI asks if there is a similar car that has driven to the same location at the same time. 

And then the system selects the route that is fastest and the most economical.  Or if the system has an access to the traffic-control computers it can see if there is a rush. And then select the alternative route. The data that the system can get is determining the effectiveness of the AI. The system requires the descriptions for the rush. 

And the second thing the system requires is the values of the speed of vehicles and the number of vehicles in both driving lanes what are requiring that the system selects the alternative route. In this version, the traffic control system shares real-time data for the vehicles. By using that data they can select the optimal route. 

(https://visionsoftheaiandfuture.blogspot.com/2021/08/learning-in-computer-world-is-simply.html)