[zackchase]

These are strong arguments. Unfortunately, while sometime interpretability addresses a fundamental (not frivolous concern), it's also true that often interpretability is sometimes contrived to serve some irrational bias or political motive.

I think this is expressed in the paper's discussion section:

"We should be careful when giving up predictive power, that the desire for transparency is justified and isn’t simply a concession to institutional biases against new methods."