Tuesday, September 21, 2010

Nature is bad; intelligence is good

No CEO of any business would abide by the methodologies of natural selection for refining existing products or producing new ones. The manner in which "nature" works when selecting what's "best" is quite dumb, really: make a bunch of copies of stuff that doesn't actually fix or accomplish anything, then throw all of them into the trash except for the one that seems to work the best in that particular niche.

That would be like if a major corporation attempted to determine which product was best suited for a given market by creating ten variations on that product at random -- with two being horribly inoperable -- spending millions of dollars on each product line, and then terminating all of the lines but the one that appears most profitable. Nature is worse, though, because its "products" can scream in pain and terror, and usually do at least once in their lives.

The more "natural" that something is by conventional definitions, the stupider it is to adopt, revere, or support it. Foresight, planning, and goals are far superior to blindness, idiocy, and apathy.


  1. Sorry I haven't written in a long time.
    You're wrong in saying that mindless processes are never useful. Genetic algorithms are good for finding effective (if not optimal) solutions in very little time. Since we have limited time and storage space it's not always rational to use planning and foresight; it's better to prioritize it for important problems. Unless there is a massive revolution in general AI, at least.

  2. Using computer algorithms to process and select for advantageous traits within milliseconds or less is not the same thing as putting an entire lifetime of energy and resources into ten different models where it's obvious that only one of them is optimal. Yeah, I wouldn't go so far as to say that there is never an advantage in random variation, but in the case of computer models, the resources and processing power in use is so minimal -- and so quick -- that it doesn't waste nearly as much as the conventional method, which is not to generate models, but to generate the end products themselves. If a business were to actually produce millions of several different variations of a model and then sell them (rather than select the best one before getting to that stage, where the actual waste occurs), only to realize after the fact that only one made economic sense, then they'd be proving my point.

  3. What's that? I was too busy being blind, apathetic, and idiotic to start hating your anthropomorphized bitch-goddess.