Raphael Roche comments on Nobody is Doing AI Benchmarking Right