They built a sandbox, as described in this blog and by the fintech industry.
That machine views the world as an auto-traded pit of move sequences which is a projection of a rule book in the smart contracts layer.
In other words, I have a rule set for chess,and I apply it to the queen side rook, knight move combinations. Under the projection of the rule book, one side wants to bottle those pieces up with a small ;move tree', and expand the other side with a more expanded move tree. That is, the bottler wants a great selection of moves than the bottled-up.
Comparison of tree structure is our generator calculus, for aloh zero they called it neural net weightings, but they may have gone n t generator theory.
In any event, aloh zero wants a probabilistic generator, one that is accurate for some long sequence that gets to the end game. It wants a probability distribution of move sequences for that corner of the board.
If in the limited move space, the generator for one side occupies more moves, then it generate tempos for other parts of the board. A tempo is like ash, it can be spent elsewhere on the board, in another trading pit, like the king side corner where the opponent has castled.
Smart contracts are the rules, including subsets of ruled designed for some pieces over sme subspaces of the game. The goal is to price the moves as it plays, it has a portfolio management problem.
No comments:
Post a Comment