If you read his latest book, he describes what I think is a very healthy view of science and the notion of expertise in general. For certain classes of problems, an expert will offer the right solution 98% of the time. But for edge cases it's more like 50%.

I think there is a lot of conflicting science about nutrition (exacerbated by advertising), and people assume it falls in the 98% certain category when oftentimes it falls in the 50% category.