Grey and Green Goo

I'm familiar with the name Ray Kurzweil thanks to some reading I've done on artificial intelligence. Provocative thread about him on slashdot yesterday. Found this buried in the comments:

I'm a grey goo skeptic as well, but only because I think it is unlikely for practical reasons, not impossible for theoretical reasons. The comparison of Grey Goo with naturally evolved microbes ("Green Goo") misses the point that the man- (or machine-) designed nanobots can explore areas of design space that are blocked to natural selection, perhaps by being hidden behind local minima in the fitness landscape.

A well known example of such a feature (albeit on the macro device scale) is an eye with telescopic vision: an engineer can produce this feature trivially by placing one lense behind another and allowing the distance between them to vary. However, in spite of being independently evolved in the animal kingdom as many as a dozen times, no known animal eye has managed to find its way to this two-lense arrangement. Neither has any gradient-descent-constrained simulation of eye evolution managed to achieve this configuration (to the best of my admittedly-amateur knowledge).

(For those unfamiliar with the "local minimum" problem, it's this: a creature with, say, a misshapen doubled-up lense would be outcompeted by its already-extant fully-sighted peers long before its hypothetical future descendants could have time to happen upon the further refinements needed to implement the long-term-superior telescoping design.)

In the nano realm, there may well be enough yet-unknown technological tricks left in the bag to make sunlight-powered dirt-eating replicators a theoretical possibility.

But I wouldn't lose any sleep over the possibility, either.

-- Saucepan

Good example. Intriguing discussion.