Re:Why deep learning works even though it shouldn’t
by machinelearning on Wednesday October 21, 2020 @12:20AM
(#24844707)
If you want to go more meta, I can point out that your comment is pattern matching to training models and applying them to humans vs. thinking from first principles and realizing that humans and models learn differently. That's exactly the issue with applying functional truths as truths.
This seems super meta, because I am a beginner engineer and your comment is totally pattern matching to distillation, but for humans instead of models. Maybe beginners should start with those “functional“ (read: distilled) truths and train from there :)