There are 2 secrets to the ML hype, and I’m not sure you’ve figured out either yet.
- ML techniques, beyond trivial classifiers, are rarely, if ever, better than ‘hand coded’ algorithms.
- Significant ML, where the machine actually derives relationships, is almost never used in actual products. (why is this? that’s a better article you could write…)
I’ve read about Spotify and Netflix recommendation engines, and neither seem to have significant ML aspects, in fact they seem like the opposite, they contain an enormous amount of fine tuning coded by humans. Can you demonstrate otherwise?