Real Intelligence

Estimation and Alpha Zero

Arithmetic was the basis of higher math and science. We automated arithmetic in the 20th Century and by the end of that decade a computer beat a human grandmaster in chess. And now we have chess engines like Stockfish which see many moves ahead it cannot be beaten in what it does. 

But a new kind of computer program that is based on human intelligence, Alpha Zero, bounces the engines out of the field. Alpha Zero uses less time and iterations than the engines. They are emulating human brains.

The computer programs carry their superiority to many games: checkers, go, and others, but the computer cannot beat humans in computer the strategy games. The make the game harder to win by giving the opponent more resources, not by making them “smarter”.

Arithmetic is a tedious, unforgiving process and all is black and white. We have automated the process, but that is like putting a band aid on hand cuffs. We could not beat the chess engines. 

AZ is based on programming that mimics human thinking. The computer plays millions of games and learns how to win. If AZ is an imitation on human thinking, what can human thinking do?

One way to start is to teach estimation and explore its limits. If estimation could include word problems, it has an important role where close is good enough. If you need exact, use a computer.

Subitizing is a skill that disappears when the child learns to count. But does it? A painter looks at a room and says it needs 3 cans of paint. Square feet per can does not enter his head. How many loads of stone would this wall take? The architect must subitize.  And all this develops on its own. Can we help develop it?

Alpha Zero uses artificial intelligence patterned on human brain systems to play chess. What more we can do with Real Intelligence?