Watching the Dysons and Tom Standage (the guy who wrote the “Victorian Internet” telegraph book) talk about the future. Cadigan started off by pooh-poohing the singularity — computers are a long, long way off from being as complex as human beings. Charlie and I exchanged glances — doubling curves start shallow and grow FAST.
Aside via Charlie’s blog:
Original ARPA contract to build 4000-ton nuclear powered interplanetary spacecraft in 1958: 27 pages.
Original NASA contract to supply 1800-odd photocopied pages of old NASA files for NASA archives (at six cents per page): 32 pages.
George Dyson sez that if we have 100*10^6 transistors on a chip, we’ll use half for the OS.
Cadigan just asked Freeman about his plan to visit Saturn by ’70. Dyson sez we coulda done it, and used up all of our nukes besides (see Freeman’s plan to make “putt-putt” rockets, exploding nukes off the ass-end of a [well-shielded] rocket to propel it).
Freeman: We thought we’d go to the moon, but nothing happened for 15 years. Then Sputnik went up and we said, “Thank God, now we’ll get moving.” We started thinking about how to use nukes to get into space.
(Aside, Charlie told me about a story he’s working on where the French suboceanic nuclear tests were actually aimed at exterminating the cthuloid sea-monsters — which is why the Brits didn’t really protest)
George: I was 5 years old when the project began and it was a complete black hole of secrecy, Dad couldn’t tell me he was working on a spaceship. Then the feds declassified it and he told me that we were moving to California so that we can go to Jupiter and I became consumed with the project. My most recent book with Penguin is the first public thorough documentation of the rise and slow starvation of that project.
Cadigan: How complicated was the Turk (sham Victorian chess-playing automaton)?
Standage: People like Babbage had argued about whether a machine that could play chess was a thinking machine. In the book, I disinter the old story to explore the ancestry of AI and computers. An automata is a self-moving machine, and so is a computer. Think cellular automata.
Since the Turk appeared, there have been lots of attempts to define machine intelligence: Interactivity (the earliest automata would just do something, wind down, get wound up and do it again). The Turk would respond — it would interact and behave non-deterministically. But by that standard, an ATM is intelligent. By Babbage’s time, intelligence was memory and foresight. Then Turing, who was very interested in chess, so it became a proxy for intelligence. Then conversation — the Turing Test. It’s always about imitation, trickery, games. The Turk was a trick, it was an imitation.
Cadigan: So instead of trying to develop intelligent machines, we’ve been tricked into developing machines that play chess! Lately we’ve been hearing a lot about complexity, and there’s this notion that once the complexity of a machine achieves the complexity of a human brain, something intelligent emerges. It’s fun to imagine this spontaneous transcendance, but this really isn’t good science.
Standage: The more you know about computers, the less likely you are to believe in this. The bigger a computer is, the more brittle it is.
Me: horseshit! The Internet is the most complicated machine we’ve ever made, and its robustness comes from its complexity and size.
Standage: Kurzweil’s arguments are spurious numerical arguments.
Me: Talk about spurious. Ever heard of evolutionary software?
George: There’s a slim possibility that we could revive Project Orion. Arthur Clarke wrote me a letter: I was shocked by the story of the early nuke scientist who lit a smoke off a nuclear blast — doesn’t he know that smoking’s bad for your health?
Cadigan: Freeman, what are you thinking about?
Freeman: If we’re serious about going to space, we should be thinking about it. How do we grow potatoes on Mars? How to we adapt ourselves to live on other planets rather than embarking on terraforming adventures.
Cadigan: Are you still a disbeliever in nanotech?
Freeman: Oh, it exists, but it’s not revolutionary, not like biotech. Most of what nano was supposed to do are being done far better with biotech. Nano is neither as dangerous or useful as biotech.
Audience: You didn’t like Wolfram’s book, Freeman. Is the world designed by mathematics or algorithms?
Freeman: I was quoted as saying it was worthless. I was also supposed to have said that I only glanced at it before pronouncing judgment. But that’s not true, I looked at it rather carefully. But while it’s interesting, it’s mostly not new. Most of the interesting cellular automata was done by Conway with the Game of Life. Wolfram’s elaboration of the Game of Life doesn’t amount to much, despite his completely unfounded claims that CA theory governs physics, biology, etc. His programs are beautiful and interesting toys, but they lack intelligence.
Me: The most complex machine we’ve ever built (the Internet) owes its robustness to its complexity — complexity is NOT brittleness.
Standage: You’re right — the Internet isn’t engineered; it’s grown. It’s more like gardening and less like science.
Freeman: Anything complex enough to be intelligent can’t be understood, anything simple enough to be understood can’t be intelligent, which is why Kurzweil won’t build an intelligent machine.
(me: I don’t think you get Kurzweil. He wants to evolve intelligent machines that he can’t understand, by using raw, brute-force computation)
(Thanks for the new Conway link, Seth!)