3 ms·

Isn't the biggest structural problem with ML that it can't explain how it came to a specific conclusion? Anyone knowledgeable on this subject?

• viraptor 6y ago
  There's a difference between "we can't tell you exactly why X was classified as it was", and "this phrase was classified as bad". Here, YT knows exactly what triggered the issue and can review the decision.
• lmm 6y ago
  ML absolutely can, if you bother to implement it that way. And non-ML systems can be just as opaque. The problem isn't ML per se, it's lazy organisations that don't care.
• fastball 6y ago
  Not really. The limitation is mostly in the training set. If you train an ML algo on a set of videos that you either banned or didn't, sure, when you run that in prod all you'll get out of it is "ban or no". But if the training dataset includes a ban reason, then the inferences will include reasons too. What is opaque is how the algo decided the thing was violating the policy. That applies to supervised learning. For unsupervised learning, you can still do this by requiring a valid rationale as part of the "win condition".