A little bit more about the robot, what would be equipped with artificial brains

A little bit more about the robot, what would be equipped with artificial brains

https://controllingsociety.blogspot.com/

(Wikipedia)

Robots are not humans

When we are thinking about robots, what are equipped with artificial brains, we might think that this system operates like a human. If we would think the case, that the robot is equipped with the artificial neurons, we must remember that this kind of system is stable. It cannot create new connections between those silicone-carbon neurons, and that means that this system needs some kind of mass memory like flash memory for recording things and creating the databases, what it can benefit for creating the solutions for problems.

And the database is the thing, what makes the intelligence. The idea of intelligence is that thing would connect the stored data together, and then make the new models for the necessary actions. And in the case that the robot would collect data spontaneously by using sensors, that thing might be very dangerous, because that kind of things would cause that the robot would become the thing, what recognizes itself. The thing is that, when we are creating a massive database, we would offer change to create own connections between memory blocks. That process could be similar, is the learning thing robot or human.

Teaching the robot is similar process than teaching child

When we are thinking about the situation, where we are teaching robot for walking, we are facing the thing, that when the robot is recording things, the actions would become more independent in each time, when we are taking this robot out from laboratory for walking. In this kind of scenarios, the artificial intelligence would record the actions, what the controller would do, and the recording would happen in the natural conditions by walking with a robot at the streets.

But that process would also be supported by using computer game type workspace, where the programmer would fix the errors of the actions of the robot. That kind of thing is called hybrid learning. There the simulation supports the learning in the natural environment. This kind of things would be a very effective tool when the robots are learning actions, what are needed in complicated environments, like in cities. That means that when the robot is learning like a child, everything would be recorded, and the programmer must not write every movement to the console.

Why do robots learn chess easier than making sandwiches?

The thing what differences the chess-game from the sandwich is that there are strict regulars in the chess board. The computer must separate only two colors and the forms of the chess buttons. If we will want to create the chess-robot, what needs to operate by using the specific buttons, and that means it will need only the hand, that is equipped with RFID (Radio Frequency ID) sensors. Robot hand and other physical systems can be connected with every chess program, and the system can get the movements of the buttons from the database.

Then it can analyze the movements, and increase its own tactics by comparing the movements of the opponents and making the game simulations against those humans. and the thing is that the bigger database would make that program more intelligent and its knowledge of chess is increasing, which means that it can know more sophisticated tactics against human players. And the growing of databases is calling as a learning process. The big database would involve more game styles and movements, and the computer will connect them together for creating more and more new ways to play against other players.

The key is how to recognize things? 

The RFID system would tell the robot hand, where each button locates, and the thing, what the system must only know is the location of the hand, that it would not move button outside of the chessboard. And that thing can be made by teaching the movement of the hand, and there could also be small transmitters, which will locate the manipulator in the right place. The system can teach by putting the button in the hand, and then the operator would write the name of each button, and the system records the RFID signal, what it compares with the database, where the roles of the buttons have been stored.

But when we are starting to make sandwiches, that robot needs more information from the sensors. In the food cannot install the RFID sensors, and the robot must recognize the materials by using some other sensors. And the problem is that robot must know every single component like mayonnaise, sausages butter and other things like bread. Then it must take the right amount of the things and cut as an example of the sausages for the right size. And every movement and move track must be programmed to the system separately. Of course tools like knives can be RFID-marked, but the food can be recognized with image-id systems. And this makes this kind of thing very difficult for the computers, what are in the best with the limited logical problems like calculations.