Vivent les ordinateurs!
8 o'clock, March 10, 2006
Brain=computer is so done. (Are you listening, Ray Kurtzweil? No? Well, when some sub-aspect of your uploaded future self is ego-surfing the post-Omega Point simulated blogosphere in the simulated universe, post some simulated comments telling Simulated Me that I’m wrong.) Bruce Sterling nails it, as usual. This is from his speech at O’Reilly’s Emerging Technology 2006, but if I tell you to read the whole thing, you probably won’t. So.
When it comes to remote technical eventualities, you don't want to freeze the language too early. Instead, you need some empirical evidence on the ground, some working prototypes, something commercial, governmental, academic or military . . . . Otherwise you are trying to freeze an emergent technology into the shape of today's verbal descriptions. This prejudices people. It is bad attention economics. It limits their ability to find and understand the intrinsic advantages of the technology.
A good example of freezing the language too early is, I think, Artificial Intelligence. We very early got into the lasting bad habit of referring to computers as “thinking machines.” I suspect this verbal metaphor seriously harmed technical development. Even the word “computing” sounds too much like human mathematical thinking. We might have made a much better language choice if we had called computers something like the French called them, ordinateurs, “ordinators.”
Computers are not “smart,” in any useful sense of that term. They don’t “think.” They don't have “intelligence.” Computers don’t “know” things and they don’t have any literal “memories.” They’re not artificially intelligent sci-fi beings like HAL 9000. Computers are boxes of circuitry, with strings, and slots for the strings. They are not alive and mentally active, they are just sitting there, ordinating.
What is “ordinating,” exactly? Well, if we’d invested our attention in figuring that out, instead of awkwardly struggling to make these devices think like a human brain does, then we would have successfully explored the very large set of interesting problems that computers turned out to be really good at.
If you look at today’s potent, influential computer technologies, say, Google, you’ve got something that looks Artificially Intelligent by the visionary standards of the 1960s. Google seems to “know” most everything about you and me, big brother: Google is like Colossus: the Forbin Project. But Google is not designed or presented as a thinking machine. Google is not like Ask Jeeves or Microsoft Bob, which horribly pretend to think, and wouldn't fool a five-year-old child. Google is a search engine. It’s a linking, ranking and sorting machine.
Linking, ranking and sorting don't sound very sexy, glamorous or philosophically crucial. Instead of nostalgically clinging to the words — the neologisms of the past, which are now archaeologisms — we should pay more attention to the facts on the ground. What works? What matters?
When I think about it: do I really want some classical Artificial Intelligence computer that can talk to me just like Alan Turing? Or do I prefer Google? Imagine two start-up companies. One of ’em has got Alan Turing’s disembodied talking head inside a box, but no search engines. In the other company, they have no AI, but they get to use Google.
Which company out-competes the other? One company asks: where do I find a cheap supplier? In response, they get a really genius math lecture by Alan Turing. Alan is really sincere about it, he’s really thinking hard about the problem of supply, there inside his box. The other company has Google, so in about ten seconds they not only find a supplier but all kinds of massively popular links to other suppliers. Which company wins?
I guess you could argue that Alan Turing's super-smart metal head might invent Google, but do you need this roundabout approach? All it takes is a couple of grad-students to invent Google.
After doing that, the folksonomy kicks in, so you get all the linking and ranking and sorting. So, you can keep Alan Turing in the box, and I'll even throw in Richard Feynman. They can artificially think, while the rest of us will be getting on with the revolution.
I mostly agree, but I think Bruce is way off when he later says,
A tech world that talked about ordinators, instead of Artificial Intelligence, probably would have produced Google in about 1980.
While I can imagine a different historical timeline that would have resulted in Google in 1980, the absence of the notion of "artificial intelligence" is neither sufficient nor necessary to bring that about. Not by a long shot.