This is a nice, succinct condemnation of the same idiocy that motivates people to pick up a book on "agent-driven swarm learning systems" instead of understanding how to maximize a posteriori the probability of an observed data set and apply the results to an incoming stream like it.
Wired (Conde Nast) is in the eyeball business: they sell eyeballs to advertisers. The real surprise is why anyone expects them to publish something other than SciFi hand waving which appeals to the "arithmetic mode" of consumers.
This may not be the proper place to express this, but does anyone else find excessive linking in an article obnoxious? Can't they just quote and then cite at the end? Why do I need 9 tabs open to fully enjoy a short article?
Remember though that the human brain came to exist through the processes of random mutation and natural selection.
You _could_ make an airplane by taking a car, making random changes and selecting those that fly the best, but I doubt anyone would have the patience to be successful at this method. You'd probably be better off hiring some engineers.
I think at least optimization of airplane parts by evolutionary algorithms has been done. Also, why start with a car?
Anyway, the point was that there seems to be a "data mining" algorithm that is very successful without forming scientific theories, namely the algorithm the brain is running. So I think the guy proclaiming science to be unnecessary might have a point. The article of this topic reminded me a bit of the ramblings of somebody defending his job that has become superfluous.
The art in using Bayesian statistics is properly choosing and weighting priors, which might explain (assuming your theory is correct) why some people consistently make better decisions than others.
Note also that computers are good at doing stupid things fast, while people are better at doing complex things slowly. Use the right tool for the job.
computers are good at doing stupid things fast, while people are better at doing complex things slowly
It's a good rule of thumb, reflecting our current ability to use computers. I'm looking at the problem from scientific perspective, not trying to get things done.
why some people consistently make better decisions than others
You don't have to look that deep to explain this. These people should just have better understanding of the domain. I don't think there are data saying some people are consistently better than others across domains. If such data exist (wink), you could say they have better modeling of the world in general.
(Example kiped from http://www.autonlab.org/tutorials/mle.html ; Andrew Moore is a hero)