Why prediction solves the AI problem
Finding the rules of dependencies and their application to the calculation of new unknown situations it is a prediction.
All information in the computer is encoded as a sequence of bytes. That is, everything is projected into a computer in a byte sequence. And dependencies are also projected onto such byte sequence.
It is necessary and sufficient to make byte-stream prediction in order to obtain the target result of the project - the ability to universally solve arbitrary tasks and issues that can be implemented.
First, we put a lot of text in the stream. It is assumed that the rules are considered for everything that falls into this stream.
After, we feed the text with additional markup into this stream:
You can easily say the continuation text, which may be in place of the ellipsis, although before this rule has not been seen within such tags.
It is your brain that will apply the principle of function selection in the same way as in the previous examples. That is, your brain made a prediction.
And if instead of simple patterns in tags there will be question-answer or task-solution pairs, then such tags will give answers.
<Question>What is the color of a grass?<Answer>Green<EndQuestion>
<Question>How much will be two plus two?<Answer>Four<EndQuestion>
<Question>How many days a year?<Answer>...
Of course, if the program can find the necessary rules as a mathematical prediction. To do this, she must be able to find types, simple structures, complex structures, and much more.
Here I have shown not all the details that need to be done when using prediction for the solving task, and not all ways to use such prediction. But this is enough to understand that I mean that it is necessary and sufficient to do byte-stream prediction.