Is Machine Learning good or bad for the physical sciences?

I show that there are new biases introduced into the physical sciences by the use of machine-learning methods. At the same time, there are places where measurements and scientific conclusions can in fact be made more conservative and believable by the introduction of machine-learning methods. I present some ideas for improving machine-learning methods for physical-sciences applications.

Bio

David W. Hogg is a physicist and astronomer with interests in statistics, precision measurement, and applied mathematics. He works on finding planets around other stars, measuring the dark matter using stellar dynamics, and cosmology.