Comments by "Dmitry Kazakov" (@dmitrykazakov2829) on "Oriol Vinyals: Deep Learning and Artificial General Intelligence | Lex Fridman Podcast #306" video.

  1. In AI there are concepts of overfitting and coverage. The size of the data set neither precludes nor guarantees anything.
    2
  2. This is called "explanation system." I.e. if a machine gives an answer you can ask how the algorithm came to it. This depends on the learning system. The major critique of neuronal networks is their explanatory weakness. They stay black box so that you basically win no knowledge from training such a system.
    2
  3. Any classifier is nearest neighbor with some metric induced.
    1