Machine vs. Man, Part 2

In Yesterday's column, I showed why humans are evil: we cheat. Computers, on the other hand, only do what's best. They follow their programming.

And they're fast! Damn fast. Just today, a co-worker had the gall to complain when his computer took two whole minutes to scan a 600 megabyte file. 600 megabytes? That's as much as a CD! And it took me at least 10 minutes to know that the new Will Smith album sucked.

But if you've even just dabbled in popular culture during the last half-century, you've seen some tale of computers killing humans or robots beating up seniors for their pills. Allow me to take a moment to say (as both a computer scientist and a hater of near-deads): this is not going to happen. We are so far from understanding human emotions (computer scientists especially) that we will never be able to mimic this sort of irrational bias.

Instead, we will craft agents with irrational exuberance. They will be the best moneymakers out there. They will bid with astounding precision and lightning speed. Faster than a human could observe, heck, faster than any human could audit...

And they'll learn. They'll know to increase the price of ice when it's hot and to sell Sell SELL whenever a music label signs Will Smith (ha! Twice in one posting! Who's fresh now?). Amazing things. Optimizing profits and conversions, all that lovely stuff.

But, here's the thing: we don't believe in the Rising-tide Survival-of-the-fittest Laissez-faire capitalism. We live in a fluffy-bunny hugs-to-everyone bastardized marketplace. You have to pay people a "minimum wage" and give them "time off" for such hoitsy toitsy extravagances of the union imagination as "maternity leave" and "paternity leave" and "sawing through their bone with improperly installed welding equipment."

But the key thing is, you aren't allowed to talk to your competitors. Even though, it would be really good for you (the Stovepipe-Hat-wearing Fat Cat that you are) to do so: you could set prices to be artificially high. We begrudgingly begrudge knowing that gas station owners on an adjacent corner may *wink wink* *nudge nudge* each other and that's why they have the same damn prices. But computers, well... they can wink and nudge 5000 times a second.

If that's not clear, think about it this way: Amazon learns that when Barnes & Noble increases its price by x%, it will increase its price by y%. Barnes & Noble realize this, and adapts to it. The two co-evolve, until they have a signaling infrastructure as complex and indecipherable as the absurdity of baseball (where coaches mime Parkinson's to tell their batters orders for which a few syllables of speech would suffice).

This is not evil. It is dumb logic. The sort of "Lord of the Flies" naivete that you get whenever you give a child a machine gun. So what to do with these computers? You can't put them on the stand in court. And it's certainly not their fault. But, heck, we can barely make computers that can do this, we could never make computers that could do this and describe *why*. That's the central point of artificial intelligence as it currently stands: perhaps human-style responses are only 10 years away, but it will take at least 50 before we have a machine that can explain like a person (and a full century before they can ever lie like one).

Join us tomorrow for the resolution of this revolution.