Jetson Fallacy

I was re-reading this tonight and finally came up with a way to explain why I think Marshall Brain is a complete fucking idiot. I think he falls for something I'm going to call the Jetson Fallacy.

If you look back at the early and pre-history of computing, you get a lot of quotes like "by 2010, computers may weigh as little as half a ton!". Even the people who got it right, like Vannevar Bush, didn't really get the implications, so if you want to make predictions about computing, it would seem safe to go wild and predict the most optimistic changes you can imagine (and you'll still miss the implications, but everyone always does, about everything). So you predict that androids will take over all industry and herd humans into giant storage compartments.

Ironically, if you told the same android-predictors that in ten years we'd all have antigravity-powered shoes and time machines, they wouldn't buy it. "Those are physically impossible". Interesting phrase, that. There's a physical law that says you can't travel in time, or have an antigravity field. But there's no physical law that says you can't have, say, an airliner that seats 500 people, so that they'd find plausible.

The Jetson Fallacy is the failure to impose the "physically impossible" filter on computation. We've seen changes like going from biplanes to 747s, with computer speed increases and miniaturization. The android-predictors don't know or understand that there's a fundamental difference between progress in that and things like strong AI: one is a 747, amazing but possible, the other is antigravity shoes, totally impossible (without some major and unpredictable breakthroughs).

Computers, every one from the ENIAC to the Macbook, do essentially the same thing, just faster and in less space and with less power. And unless you understand what that thing is, and the kinds of things you can make with it, it's not obvious that strong AI isn't one of them. Data and C3PO seem just a couple RAM upgrades away.

So I guess this means I don't believe in strong AI.

Idea for thing

You know what I want that doesn't exist?

It's about the size of a USB stick, maybe a little bigger and fatter, but roughly that same form factor. It's a 16 gig USB stick, among other things.

But the "other things" are where it shines. It's a tiny computer, running Linux, system-on-a-chip. It boots up as soon as you plug it in, and then you can ssh to it, or use VNC, and run programs on it.

For network access it would be able to route packets through the machine it was plugged into.

So, you plug it in and it shows up as a hard drive, and a little orange LED turns on to indicate that it's booting. The drive contains some client software for Mac and Windows, an SSH client and a VNC client, maybe an app to let it share network or something.

After a few seconds it turns green to show that it's up and running. You connect to it and see a desktop with all your stuff on it. Or, you use tools on the host machine to edit files on the drive, and then you use an ssh window to the stick to compile them.

What's it good for? That, I don't know. Portable dev environment is one, but that's less and less important now with Git and things. It would be really fun to play with though.