Machine learning and traffic.
How to make computers come out of the box? The simplest way is to make new boxes for the robots and computers. In that model, every box is a new skill for the AI.
The thing in machine learning and learning algorithms is that they are not far different from us what we maybe want to believe. If we are thinking about the AI algorithm that is controlling things like traffic we might say that the AI is limited. The reason for that is that the AI controls only things that are connected to the traffic. But then we must realize that the traffic controlling algorithm will not simply know anything else. But it knows how traffic is working.
The thing is that when humans are learning new skills like driving. They go to driving school. That driving school gives us the skill to drive. So in that learning process, the driving school is making the marks in memory that involve all skills that we need for driving.
The driving school creates the module that involves all movements that we need for driving. Learning is simply recording the movement series. And then, those movement series are connected to certain actions or signals.
Then those movements must connect with the certain situation. So we can think that modular programming is like a group of boxes. In every single box is a series of actions.
And programmers can connect those boxes freely. So modular programming is like putting the domino buttons in a certain order. Each box or button involves certain series of movements.
The traffic control algorithm doesn't have physical actions. Except when the system controls active screens and traffic lights.
If all vehicles would be robots. The traffic control system could give the orders to driving computers. And the car's internal computer makes the decisions about driving those cars. The system is a two-layer neural network where the vehicles are communicating with each other. And the traffic control system operates above this base layer.
The system collects data from sensors about the driving speeds and positions of the vehicles on the road. The purpose of the algorithm is to make the traffic work flexible.
And the mass of data that the system has determines how effective it is. If the algorithm knows the goal of every vehicle it can make individual tracks for those vehicles and even fit their speed to each other. And that thing makes it possible to make flexible and fast traffic. The AI can handle large entireties.
But that algorithm is in the box. That means it can make things very fast. But those things must program into its memory. We can think of those boxes where the AI is as nodules. Each module is like a box that involves a new skill that the AI can use for expanding its skills. The thing that makes it possible to paint houses is that there is some kind of robot that makes it possible to make that thing.
There is the possibility that the robot cars have robot men. Which mission is to carry luggage and as an example clean the vehicles. So if the traffic control computer takes orders that it should paint a house it can ask some robot worker to make that thing. The thing is that the AI can scale the mission all over the entirety. And if it will not know what painting means, it can simply ask that thing to all robots.
But the only thing that the AI knows is how the traffic works. The system knows how to control traffic lights and adjust the speed limits. But if we want to give orders to that algorithm that it should paint a house the system will not know what to do, because it probably even knows the word "paint". That means that the AI is like in the box. And how the Ai can come out of the box? AI cannot yet learn like humans. But there is a way to expand the operational sector of AI simply by making the new box where is a new skill.
https://onlyimaginationlimitsinnovation.blogspot.com/
Comments
Post a Comment