On the use of evolutionary algorithms to find laws from data: Successes and limits

Finding laws of Nature directly from observations, without further
hypothesis nor processing, has been a dream with some attractive to a
variety of scientists and philosophers, as for example the 17th century
British empiricists. Two recent papers [1,2] seem to advance into this
program, and in the related one of automating the process of scientific
discovery.
In this talk I will describe the results on these two papers, putting
them in context, identifying some interesting points and also
limitations. The (optimistic) message is that we human scientists need
not to fear about being replaced by robot scientists ... yet.


[1] Distilling Free-Form Natural Laws from Experimental Data,
Michael Schmidt1 and Hod Lipson. Science 324, 81-85 (2009).
http://dx.doi.org/10.1126/science.1165893


[2] The Automation of Science, Ross D. King et al. Science 324, 85-89
(2009). http://dx.doi.org/10.1126/science.1165620