[PREV - REVIEWERS_LEARN]    [TOP]

RE_EVOLVE

                                                    March 30, 2012

     "It's difficult to get
      everywhere worth being by
      evolutionary increments."

      This is one of those obvious
      points that's worth underlining
      (and expanding on, if possible).


Adaptive, evolutionary approaches
are very good at refining, but they        I'm reminded of Newton-Rapheson
can run downhill into pockets of           iteration, a numerical technique
local minima and get stuck.                for experimentally finding roots
                                           (zero-crossings) of an equation
 Occasionally you need to jump             that uses the slope of the
 out of the ruts with a wild               equation to decide where to look
 mutation, or evolution itself             next.
 can be a dead-end.
                                               It's much faster than just
    Web designers these days are               linearly sweeping through
    doing a good deal of whining               a range of numbers... but
    about Google's insistance on               it's possible for an equation
    A/B testing every tiny little              to have many roots, and if you
    design decision (should that               don't start it up on the right
    line be 5 pixels of 8?  Is                 side of a peak, it'll never
    #8866EE or #8866FF a better                slide down and find the root
    color?).                                   over on the other side.

    I have very little sympathy
    for these designers-- it used to
    be they would regularly push out     I watched some morons
    Grand Designs that flopped.          at Netscape trash
                                         netscape.com's number
    But it is a point that if you        one ranking with an
    rely solely on this accumulation     unusable and unreadable
    of tiny improvements, there's        mess of Javascript.
    a limit to how far you can get.      For the first time,
                                         yahoo.com took the
          You can optimize,              number one spot.
          but not create.


                 What Google actually does is allow a
                 lot of innovation on side-projects--
                 which appear to be frequently starved
                 of resources and left to die lonely
                 deaths.

                 They're also, like many a cash-heavy
                 company before them, quite willing to
                 buy up other software startups with
                 ideas that look potentially useful.




--------
[NEXT - LEARNING_THE_BOARD]