You're looking at Less Wrong's discussion board. This includes all posts, including those that haven't been promoted to the front page yet. For more information, see About Less Wrong.

solipsist comments on Open thread, Nov. 16 - Nov. 22, 2015 - Less Wrong Discussion

7 Post author: MrMind 16 November 2015 08:03AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (185)
Tags: open_thread

Comments (185)

You are viewing a single comment's thread. Show more comments above.

Comment author: cousin_it 16 November 2015 02:23:04PM *  9 points [-]

I've been hearing about all this amazing stuff done with recurrent neural networks, convolutional neural networks, random forests, etc. The problem is that it feels like voodoo to me. "I've trained my program to generate convincing looking C code! It gets the indentation right, but the variable use is a bit off. Isn't that cool?" I'm not sure, it sounds like you don't understand what your program is doing. That's pretty much why I'm not studying machine learning right now. What do you think?

Comment author: solipsist 16 November 2015 03:27:13PM *  0 points [-]

It sounds like you do not understand what your experiments are doing. That's pretty much why I'm not studying electromagnetism.

-- Letter from James Clerk Maxwell to Michael Faraday, in the setup of a Steam Punk universe I just now invented

Comment author: solipsist 16 November 2015 03:45:11PM *  2 points [-]

Here's how I read your question.

  1. Many machine learning techniques work, but in ways we don't really understand.
  2. If (1), I shouldn't study machine learning

I agree with (1). Could you explain (2)? Is it that you would want to use neural networks etc. to gain insight about other concrete problems, and question their usefulness as a tool in that regard? Is it that you would not like to use a magical back box as part of a production system?

EDIT I'm using "machine learning" here to mean the sort of fuzzy blackbox techniques that don't have easy interpretations, not techniques like logistic regression where it is clearer what they do

Powered by Reddit Powered by Reddit