And I think part of what your essay is getting at is say you want to scrub some particular feature from training data such as skin color (consider how google maps blurs out facial features for example), the algorithms will still pick up elements of ethnicity from external factors, whether nose and jawline measurements or even external factors such as media habits, language dialects, fashion choices, etc. Scrubbed data will still have correlations with the stricken factors. A point of deep learning is that we are evaluating at scales of dimensionality beyond those capable by human designed algorithms, but if we’re trying to pare down to essential features then we’re back to requiring human intention in algorithmic design and thus lose the benefit of ML.

Written by

Writing for fun and because it helps me organize my thoughts. I also write software to prepare data for machine learning at automunge.com

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store