The Truth About Everything
Originally published 2005 in Atomic: Maximum Power Computing Last modified 03-Dec-2011.
I don't know what turns the conversation took in the schoolyard between you and your nerdy friends (I'm assuming you had nerdy friends. Oh, you were all rugby-team captains, were you? Sorry.), but I and a number of my compatriots in early teen social hopelessness independently came up with the idea of the Truth Machine.
You've got this huge computer with awesomely hypergigantic amounts of storage, and monstrous input and output bandwidth, and processing power to burn, right? And you feed it all human knowledge. Maybe you just let it chew through the Library of Congress, or something, but it's more fun if you also put public input booths all over the place, so anybody can drop in and tell the machine stuff.
(It was always booths for my school friends and I, not Internet input, 'cos FidoNet and CompuServe were pretty much where it was at in consumer connectivity at the time.)
Anyway, every actual statement about the world the computer receives goes into the Truth About Everything database, where it's weighed up against everything that everybody else has ever told the machine. Then, after enough yottabytes of data have been piled up, you can go into one of the booths and ask a question, and the computer will do the relational database lookup to end all relational database lookups, and give you the best answer human knowledge can generate.
Computers like this, and their technotheological implications, are a staple of the late Golden Age of sci-fi. They always had to be at least a mile on a side. (Most fans think of Asimov's "The Last Question" from 1956 as being the canonical example, but they're often actually thinking of Fredric Brown's short-short "Answer" from two years earlier. Geek-lit digression concludes.)
It took me another couple of years of goofing off at school before I learned the fatal flaw in the Truth Machine idea. That flaw is that database searches, even with the best will and the craftiest algorithms in the world, do not take only twice as long when you have twice as much data to sift. They don't necessarily take four times as long - that's where the crafty coding comes in - but you're still talking about an impressive level of growth in search times.
This means that very large databases with very large amounts of relational interconnectedness - which a Truth Machine's memory will, inescapably, have - take a long time to search. A whole-of-human-knowledge kind of database will take an unreasonably long time to search. Like, a large number of times the age of the universe, for each query, even if your fairy godmother has magically cluster-linked all of the planet's current computing power into one smoothly-running ultracomputer for you.
This is analogous to the chess-computer problem. The best modern chess programs, running on not-terribly-expensive hardware, are now playing at strong grandmaster level - and getting better. But the "perfect" chess computer, that just knows every possible game and simply picks moves that lead down the possible-game tree towards it winning, cannot exist. Well, not until we come up with computers that have a lot more bits of memory than there are particles in the universe, anyway. That's combinatorial explosions for you.
Wouldn't you know it, though - maybe we are going to make that kind of computer.
Quantum computers are still at, in transistor terms, the Bell-Labs-in-1947 stage. Despite the press releases, they're not going to be useful for anything for some time. Quantum computers are, currently, way more powerful if you emulate them on a desktop PC.
If we ever manage to build proper quantum hardware, though, the unglamorous but ubiquitous world of databases will be turned upside down. It's been known since 1996 that a working quantum computer will be able to do database searches not in an amount of time governed by the number of records put to some power, but in an amount of time proportional to the square root of the number of records.
This makes a big, big difference. It's even better than twice as many records taking twice as long to search. Twice as many records will, ideally, take only 1.4 times as long.
The '50s writers tended to assume that know-it-all megacomputers would immediately achieve sentience, and a dangerously high opinion of themselves, when you turned them on. Quantum computers, I'm afraid, do not promise to be this entertaining.
With more and more poorly correlated information swamping the world every day, though, what quantum database searching does offer is almost as much fun.
