He gave an amusing talk at Google in 2007 about creativity[0], where I thought he talked about one of the important things in science is actually believing the data since it seems so outlandish. I thought it was in this talk, but skimming the video I can't seem to find it. The talk is quite amusing regardless.

[0] https://www.youtube.com/watch?v=e7M2l-jzRG8