The Mouse Problem: Why A.I. Will Probably Only Be Super Dangerious Just After It's Created

That thing you don't want to write about? That's probably the thing to write about.

I've got a tiny mouse problem in my house. It sometimes happens during this time of year. I tried capturing the little guy in a container, but he's proven just a little too spry. Every time I finally hear him in the middle of the night and make my move, he's scurried off into some corner.

I've only actually caught sight of him the one time. Our eyes locked, he knew either death or capture was coming, and he managed to scurry off before his large and lumbering pursuer could capture him.

So I poisoned him.

Yup. Left out a little cartridge of some mouse poison or another and, the next day, I noticed that it had been nibbled into. That poor little mouse who was probably just trying to get prepared for winter got straight up murdered because it was inconvenient for me to catch him.

Now I imagine him clutching his tiny belly somewhere in the wall, wondering why God has left him.

Man... Human beings are assholes.

And before you start up with that whole 'speak for yourself' sanctimoniousness, let me remind you that even if you're a vegan who bicycles, recycles, and carries a hemp bag to the grocery... You're still killing shit for your convenience. You go to your house that's made out of tree corpses and enjoy your AC that fucks with the environment and flush your toilet and pour harmful chemicals down the drain which screws with ecosystems.

But that got me thinking. You know, one of the main arguments that AI alarmists make is that the AI might decide to destroy us because we became too inconvenient in the same may, say, a mouse is sometimes inconvenient for us.

I think these people have kind of missed the point.

You see, while it's true that your average human is inconceivably more powerful than your average mouse and can rely on tools the mouse could scarcely imagine, it's also true that we don't kill mice because we are more powerful than they are. We kill them because we aren't powerful enough. If you were smart enough, had good enough tools at your disposal, then
you could easily get rid of the mouse or make it go away without hurting it.

We kill because we aren't intelligent and capable enough to solve things another way.

Take that analogy back to AI and you get an interesting picture. An AI that is many, many, many, orders of magnitude more intelligent and capable than a human could probably control us completely without us noticing, get everything it wanted, and give us everything we wanted.

Crazy, huh?

A good, if crude, example is wildlife preserves. There are many that allow limited hunting on these preserves. The hunting licences that are sold fund the preserves operation, stopping the animals from being hunted to extinction. Also, the preserve controls which animals are hunted and when, neatly taking care of old, sick, or 'problem' animals.

Now, do you think the rhinos on a preserve have any conception of what a preserve is?

Unlikely.

So the real problem ain't super, super smart machines. The problem would be A.I. that's smarter than humans, but not so incomprehensibly smart that it could create a mutually satisfying situation. In this case the A.I. might exterminte us simply because it couldn't find a better way.

That's the real danger.

But there's even odds that the issue won't come up in quite a big a way as we're worried about it coming up. But that's another post.

And who knows, the day when we're so powerful that we don't have to kill shit might well be fast approaching. Won't that be neat?

Till then, if any giant platters of generic food suddenly appear in front of your computer, I'd probably avoid eating it.

Source: 1