Thursday, August 25, 2005

What goes around

The announcement of the death of Robert Moog, creator of the Moog synthesizer today set me thinking about the circular nature of invention.

Forty years ago electronics in music was analogue. The idea of the Moog was to let you take various sound waves and literally mix them together to produce a new one. It was really: “here's a box with lots of inputs and outputs, here's a bunch patch cables, now connect an output here to an input there and see what you get”. For some, like Wendy Carlos, you got interesting sounds, the rest of us would produce aural mush. It was very difficult to use and certainly very expensive. It turned out to be easier to grab a sound sample as a starting point, hence the $80,000 Fairlight of 20 years ago. Now it was more a case of concentrating on the end result.

As an aside you can now get a digital simulation of the Moog with a screen representation that includes the patch bay and all the analogue bits but the sound is digital. Whether its any good
I do not know but it does seem a weird thing to do.

Anyway back to invention. In computer games and animation starting from the ground up was fine but characters did not move right. Feet were never really in tough with the floor, things like that. It was easier to use motion capture, in effect sampling of real action meant you could concentrate on what the character did not how it was achieved.

Finally there is AI. It has not really progressed since the early days of game. If you applied the Turing test (can you tell the difference between an AI and a human) it was too easy to spot the difference. So what was the answer, on-line gaming and use real people as opponents. In other words its back to sampling again.


At 9:20 AM, Blogger Luke said...

Your last comment about on-line gaming raises an interesting point about the role of human beings within 'intelligent agents' or 'bots'. The history of AI was traditionally bound to the view that intelligence had to be pre-programmed or 'front-loaded' into agents in order to second-guess the whims of the human beings it would serve. One important figure who helped to break that mould is Pattie Maes. Coming out of the MIT Media Lab, she is largely responsible for creating the (ingeniously simple) idea of agents that (a) learn and acquire intelligence by monitoring what their users actually do (think of email junk filters) and (b) acting in concert with other agents that are learning from their respective users (think Amazon recommendations: "other users who bought this book also bought..."). In the 'real world' these intelligent agents (sometimes called collaborative or social agents) have had more impact than the high-falutin fantasies of sophisticated AI manifested in sci-fi, chess-playing supercomputers, and the visions of the "digital butler" made famous by Maes's former colleague (and boss?) at MIT, Nicholas Negroponte (see his 1995 book Being Digital). Here's an article from Wired back in 1997 when this stuff was just beginning to change the world. The idea of human beings and machines working in tandem - even though it can go badly wrong - is something I find both more appealing and plausible than the fantasy/nightmares of human obselescence... which is way off your original topic, but you got me thinking!!


Post a Comment

<< Home