I have a new Kindle. This is rather terrifying, because I can now read things. A lot. I think it may be starting to affect my sleep pattern — not that I slept early enough before…
Anyway, I’ve been reading Constellation Games, which is a sci-fi story about a guy who’s a game developer — until one day aliens arrive and start living on the moon. He then ends up reviewing alien games, and it’s all very exciting. (It happens to be written by the guy who wrote Beautiful Soup, a Python HTML parser).
Obviously the aliens (of varied species, called the “Constellation”) are far more technologically advanced than we are; and so the book contains all kinds of machine translation, sentient artificial minds, and so on. There’s a particularly amusing conversation between the protagonist, Ariel Blum, and “Smoke”, a (presumably) sentient AI:
Smoke-ccsspm-6be8 Hello, Ariel. I am a submind of Smoke, Ring City's general-purpose cognition engine. My cognitive address is Smoke-Cursive-Cytoplasm-Snakebite-Singsong-Polychromatic-Musteline. In a recent email, you asked to be matched with a member of the contact expedition. I'm evaluating your application. Please answer questions with YES or NO. Do you understand? ABlum: NO Smoke-ccsspm-6be8: My supermind tells me you're being sarcastic, so I'll continue.
…you get the idea. I particularly like the way Smoke has various subminds, which Ariel gets transferred between during their conversation. (You can find the whole conversation about halfway through the first two sample chapters.) Successive superminds are more intelligent: they can understand English prose, then idioms, and so on.
How is this relevant? Well, reading about fictional AIs tends to get me excited about trying to make a real one!
Some time ago I read I, Row-Boat by Cory Doctorow. The idea of humans “uploading” to the “memosphere” is a little weird, but he has some really interesting ideas — such as the ability of uploaded humans to fork their minds to spawn multiple copies of themselves, and then re-integrate later.
I found one comment the robot protagonist makes rather interesting:
Robbie was used to time dilation: when he had been on a silicon substrate, he could change his clockspeed to make the minutes fly past quickly or slow down like molasses. He’d never understood that humans could also change their perception of time, though not voluntarily, it seemed. The climb to the surface felt like it took hours, though it was hardly a minute…
Of course I’m familiar with the idea of varying perception of time — that enjoyable activities seem to pass faster, while boring ones take forever — but I hadn’t thought about it in the context of AI before. And I have no idea what perception of time would mean for an AI in the first place (or perception, even!), let alone the ability for it to alter its own speed. Presumably a computer-based AI would have to run in some kind of cycles, like in the story — or at least move between discrete states, due to the limitations of computing technology.
I’m excited, anyhow! Hopefully I’ll write something more useful soon…