A friend sent me this link to an article describing a way of doing science (making predictions) without any theory or hypothetical model to explain the observed data. Instead, if the data is large enough (petabyte levels of data) then what are required are clever statistical algorithms to find correlations in the data and thereby make predictions. No theories or hypothetical models needed to see which one fits the data best.
These are powerful techniques opened up by having access to huge amounts of data and the writer of the article argues that these techniques will not involve discarding the scientific method but could complement it.
August 1, 2008 at 9:25 pm |
This is very interesting!
About a decade ago, I began thinking about software development, and what a piece of software, ie. application, ultimately boils down to; a series of 1’s and 0’s.
All the software that could ever be created, including the most bug filled, could be produced by simply producing vast permutations of 1’s and 0’s. Of course, finding the “right” applications would prove to be impossible. However, I suspect that there are statistical characteristics amongst various types of well formed programs that could narrow the search for the “right” applications.