mca blog

i've been reading dawkin's the blind watchmaker and am struck by the *simplicity* of darwin's original theory - cumulative complexity. it got me thinking about software and other things computer.

most real-world demonstrations of cumulative complexity share some basic aspects:

• limited set of information points
• a clear reward for a 'successful' morph
• simple repetitive looping
• pre-stated completion goal

i wonder how difficult it would be to apply cumulative complexity morphing to the world wide web. in other words. can a simulation be devised that demonstrates darwin's theory using the internet and a browser?

one possible approach would be to supply a browser with a set of text (word-phrase, paragraph, etc.) and set the browser the task of finding that text on the internet. possibly building the phrase one word at a time; storing each word as it is found. this kind of game avoids the difficulty of more interesting versions that involve media (images, video, sound). for example, find a picture of the mona lisa or a recording of jumpin' jack flash, or a video of the backstreet boys, etc.

however, this just sets up a task-completion pattern. what would be far more interesting is a model that *improves* the browser's performance of a task over time. not only does it complete something, but it gets better at completing similar tasks. now *that* would be interesting.

is this just a learning bot?