The black box model in machine learning is the thing that should make robots learn things.

The black box model in programming means that the user doesn't see the code itself. The only needed thing is functionality. The functionality is made in artificial intelligence by using the modules. The modular program is made by using pre-programmed tables or elements that will be connected to the entirety.

The learning machines can connect the tables to their entirety. And the complicated learning machine can make large entireties and sub-entireties. Things like walking to a shop can be one entirety, and that thing means that the number of skills that artificial intelligence can do expands like a fractal. 

The thing is that a learning machine can make connections between those tables like the real nervous system makes new connections between neurons. But there is a problem with that model. Machine learning requires some values that describe the tables it must choose. There is the possibility that the programmer will drive the mission of the robot first time by using the remote control. 

Then that artificial intelligence selects tables that have the best match, with the remote-controlled operations. This is one way to make a machine learn things by using modules. Making those modules requires extremely good programmers because there is needed to describe all actions that robots are doing. 

A neural model for making human-looking robots that can make things like build houses for people is making programs for robots easier. In the neural network model, every joint has a microchip that describes what the joint must do. When the main program is ordering the robot to make something the system can call those other computers for assistance. So in this model, the programmer who works with the entirety must not describe separately what every joint of the robot must do. 

That person can operate with the final solution. Programming the joints separately requires lots of work. But in that model, the programmer who wants to program how the machine must act when it works with other humans has an easier to making commands to the fingers. There is the possibility that for every finger the processor tells sub-processors that the joint that those microchips control is needed. 

So in that case the microchips are forming a hierarchic network. And those other processors can also give calculation assistance to head processors when they have free time. In that case, the microchips are forming a hierarchic entirety that connects the main network and sub-networks to one entirety. 

Image:) https://scitechdaily.com/unpacking-black-box-models-exsum-mathematical-framework-to-evaluate-explanations-of-machine-learning-models/