The TL;DR insight is a good one: if NNs are trained on the concensus status quo, they will generate the OK results, not the effective or outstanding ones. The

The TL;DR insight is a good one: if NNs are trained on the concensus status quo, they will generate the OK results, not the effective or outstanding ones. The author is I guess in product marketing and says that this result should be encouraging. They point out that simply benchmarking your efforts against the competition also gets you just to the consensus status quo. There’s also an amusing subtle shout-out in the opening chick sexing hook: people (correctly) complain that we can’t look into NNs to explain their inner working; the author points out that many skilled people can’t explain their decision making either.