http://mndoci.disqus.com/At the interface of science and computingenWed, 02 Jul 2008 10:56:09 -0000http://mndoci.com/2008/07/01/danny-hillis-gets-science-george-dyson-doesnt/#comment-797827<p>My problem is that I am not sure what point he is trying to make, cause the apparent one doesn't make any sense.</p><p>The problem with experimentation is scale. Where you can (genomics) prices go down rapidly. In other cases, for whatever reasons, the costs have remained high for years, perhaps due to over regulation, perhaps due to inefficiencies and a lack of scale. Lots of data is a good thing. Rethinking our approach is a good thing. Forgetting the fundamentals; not so good</p>mndociWed, 02 Jul 2008 10:56:09 -0000http://mndoci.com/2008/07/01/danny-hillis-gets-science-george-dyson-doesnt/#comment-797423<p>Being an article from Chris Anderson, I thought I'd find a discussion about the inefficiency of doing expensive one-off experiments to refute hypotheses. When every experiment you run is expensive, you take fewer chances, do fewer experiments, are driven toward a dogmatic mindset, and learn less. You stagnate.</p><p>Look no further than human clinical drug trials.</p><p>When the data to refute hypotheses become cheap - really cheap as with genetics data, you can take wild chances, do orders of magnitude more experiments, adopt a very open approach to competing hypotheses, and learn so much more as to make what's happening seem fundamentally different than what came before.</p><p>Petabytes of data are to science as Netflix is to a movie fetish. I suspect this is the point Anderson is trying to make, but the article doesn't read that way because he left out the part about expensive experimentation being the root of scientific stagnation.</p><p>Almost makes me want to write an article of my own...</p>Rich ApodacaWed, 02 Jul 2008 10:00:13 -0000