Thursday, September 29, 2011

Virtual Monkeys Write Shakespeare! - Giving Randomness a Boost

There is an old thought experiment, which I've read versions of in Richard Dawkins' The Blind Watchmaker [sic] and Daniel Dennett's Darwin's Dangerous Idea, which imagines thousands or millions of monkeys randomly hitting keys on a keyboard and one of them producing a work of Shakespeare by chance. The point of the thought experiment was to show that it's possible for the appearance of design to arise from randomness without a designer. The thought experiment has now apparently been 'tested' in a virtual experiment that sets millions of virtual monkeys to smashing away at virtual keyboards.

But, given just how immensely, inconceivably improbable it would be for even a million monkeys banging on keyboards for billions of years to produce one work of Shakespeare, the authors of the thought experiments realize that they can't leave it at that. The conductor of the virtual experiment knows this as well -- even the most powerful processors running pure randomness algorithms at astonishing rates wouldn't be able to run enough iterations to ever realistically arrive at Shakespeare, much less the complete works of Shakespeare.

Their solution for salvaging the analogy on behalf of evolution is to introduce 'selection' into the process, which is aware of the result that is being sought. Of course, blind material processes would have no idea what end they were working towards, and so with even the slightest bit of critical examination the analogy falls apart swiftly and fully.

Reading the BBC article this becomes glaring. The program is not a virtual simulation of randomness only. It virtually simulates the random key-strikes of million of monkeys and then compares 9-character blocks of text to a known goal -- the complete works of Shakespeare -- and keeps the blocks of text that appear somewhere in the works and disregard the rest. Where did such an ability come from? Where does the knowledge of what the desired end is come from? And even if the end was known, how would a monkey, or even a team of extremely intelligent monkeys, know how to cross-check the 9-character blocks of text against the enormous database of information that is the complete works of Shakespeare? Surely this is far beyond the capability of any monkey, any group of monkeys or any human, for that matter. It certainly is not a blind, random process of nature. Dawkins and Dennett make similar adjustments to their thought experiments, claiming that evolution can and does do something similar (which David Berlinski similarly and acutely deconstructs).

A more accurate headline would read that "Millions of Virtual Monkeys Striking Keys Randomly Can Re-Create Shakespeare, After They Give the Results Over to a Super-Computer That is Already Aware of What the Works of Shakespeare Look Like Beforehand and Which Aids in The Reconstruction". Not nearly as exciting or interesting, is it?

No comments:

Post a Comment