> if the same tasks were in the hands of people who wanted CNNs or CNN+RN to work well, the results might have been different. I guess that's an undesirable sy

> if the same tasks were in the hands of people who wanted CNNs or CNN+RN to work well, the results might have been different. I guess that's an undesirable symptom of CNNs (etc.) still being an empirical rather than theoretical science.