Dear Dad — What This AI Business Is All About • 1997

The Basics

Artificial Intelligence just means a machine that can do things we usually think only people can do — like recognize your voice on the phone or beat you at chess.

Right now there are two main kinds:

Rule-based systems — like the tax software you used last April. It follows a giant list of if-then rules written by accountants.
Learning systems — these look at lots of examples and figure out patterns on their own. Think of a VCR that eventually learns you always tape the 10 o'clock news.

The second kind is what people mean when they say "AI" in 1997.

Deep Blue — The Chess Machine

This spring IBM's Deep Blue beat Garry Kasparov, the world champion. It wasn't magic. It looked at 200 million positions every second using 32 processors working together.

Kasparov thinks in patterns and experience. Deep Blue thinks in speed and brute force. Both ways can win at chess. That's the interesting part.

Your old chess computer from Radio Shack was doing the same thing, just slower and with fewer rules.

Try a Tiny Version

Here's a simple learning machine. Type a number between 1 and 10. It will guess whether you're thinking of something "high" or "low" based on what people usually pick. It gets better the more you use it.

What It Still Can't Do

These systems are very narrow. Deep Blue can't drive a car or understand why you like certain movies. It doesn't have any idea what "winning" feels like.

Most experts in 1997 think true general intelligence is still decades away. The machines are getting better at specific jobs, not at being people.

So when you hear someone on the news say "computers are taking over," remember they're usually talking about a very good spreadsheet that learned a few new tricks.