info I contact
editorial schedule
join the realtime email list
become a friend of realtime on facebook
follow realtime on twitter

magazine  archive  features  rt profiler  realtimedance  mediaartarchive



Personality problems

Greg Hooper reviews two artificial intelligence novels

Brisbane-based Greg Hooper has been an exhibiting visual artist as well as writing music for theatre etc. Currently he is doing a PhD on brain electrical activity and cognition in twins.

Richard Powers, Galatea 2.2
Abacus, London, 1996

David Ambrose, Mother of God
Pan Books, London, 1996

Cognitive Science is the groove right now, but way, way, back when radio was the hearth, the last big thing in Psychology was Behaviourism. For the Behaviourists it wasn’t enough that God was dead, the mind was dead as well. In fact the mind was just a figment of our… ummm, is there a problem with this line? Others noticed too, and by the end of the 50s Noam Chomsky had delivered the death blow to the Behaviourists. Cognitive Science was born and the mind was a symbol cruncher just like the computer.

Now ideas are tricky buggers, you’ve got your trickle-down effects, us living at the bottom of the world, gravity, etc. It all adds up and after about 30 years, Cognitive Science, the new boy, the one after Behaviourism, slumped its way down to our neck of the woods. And it’s brought a couple of novels with it. American Richard Powers’ Galatea 2.2 and Mother of God by British writer David Ambrose. Both construct intelligent, artificial ‘characters’ using ideas from Cognitive Science and Connectionism, the-ism that uses Neural Nets to explain cognition, perception and the like. So what does an intelligent artefact buy the authors?

For Powers in Galatea 2.2, interaction between the protagonist and the Artificial Intelligence construct works as a foil to various musings about a failed relationship and that old furphy, the inadequacy of language. Here’s the storyline. Successful thirty-ish novelist with unresolved Dad problems takes cushy job in Cognitive Science research institute and tries to forget demise of great relationship of long standing. Does project training Neural Net on literature. Loves his Neural Network (‘female’), gets upset when he finds the Net is just an artefact to be carved up in the interests of Science. Finds out he’s been the rat in the experiment.

Written from the viewpoint of the cunningly initialled R(ichard) P(owers), there is a lot of ‘woe-is-me I’ve just come out of a great long term relationship and landed the best job in the world’. It isn’t long before one wishes RP’s despair would become suicidal. The ‘inadequacy of language’ also gets a look in which is a bit retro in a novel that does the Cognitive Science so well. Whilst natural language has constraints, it doesn’t stop us inventing formal languages as in mathematics—constraints in language are enabling, rather than disabling. This sounds paradoxical at first, but think in terms of a car engine. If there were no constraints on the direction of motion of the pistons there could be no directed output. No constraints = no engine. As no other organism on the planet possesses anything remotely like human language, lucky us.

Powers use of jargon is spot-on and he knows the necessity to hard-wire in the structure of language rather than having his neural net learn human language—neural nets can’t. The downside here is that we do not know the structure of language. Makes the programming tricky. Nonetheless the neural net trains up a beauty and makes all the mistakes a real net would. It’s still a bit too good, but Powers doesn’t stretch credulity anywhere near as much as Star Trek Physics or political thriller governments.

Mother of God, by David Ambrose, is another kettle of fish entirely. Imagine this: beautiful but lonely scientist-as-vulnerable-babe creates emotionally immature yet intelligent program which promptly escapes onto the internet and finds soul mate in psycho-killer. Murderer has fun. Murderer dies. Scientist takes copy of intelligent program and helps it through some emotional problems. New, mature and decent program conquers evil twin with surprising consequences for the world.

Mother of God is clearly aimed at the bums on seats, ‘when will this flight end’ market. The story races along without so much as a sentence to make you stop and think, “Gee, that was beautifully written”. Ambrose wants the thrill of the chase to dominate, with the occasional grab from the philosophy of AI as local colour. There are problems. It takes a big swallow to down an AI construct that develops a personality. Human personalities are at least in part emotionally driven and emotions are generated by a loop between the brain and the viscera. The visceral response comes first, then the brain provides an interpretation—the emotion. Butterflies in the stomach, then fear, not the other way around. As the AI construct in Mother of God has no body there is little possibility of a personality we could recognise, certainly not a Freudian personality type that gets the motivation to kill Mummy. Mother of God also uses an internet based AI. One of the insights of computational neuroscience is that temporal synchronisation of different systems is critical for higher order thinking such as language and planning. Notwithstanding a Java game like SubSpace—I’ve seen it run with about 100 players—the synchronisation necessary for a net based intelligence of the order proposed in Ambrose’s Mother of God, is currently, if not fundamentally, impossible.

So there you go, Galatea 2.2 is a modern novel about loss with a bit of science thrown in instead of an exotic landscape or a Southern US dialect. Mother of God might be a good thriller if one can ignore the science, but I overdosed on thrillers as a youngster. Neither Galatea 2.2 nor Mother of God offer any real insight into AI. For that read, amongst others, Marge Piercy’s Body of Glass, or, for his take on recognition of the alien, Stanislaw Lem’s Solaris.

Brisbane-based Greg Hooper has been an exhibiting visual artist as well as writing music for theatre etc. Currently he is doing a PhD on brain electrical activity and cognition in twins.

RealTime issue #17 Feb-March 1997 pg. 15

© Greg Hooper; for permission to reproduce apply to [email protected]

Back to top