Piponi on Automatic Differentiation

Dan Piponi of Neighborhood of Infinity has written a paper about how he has successfully used automatic differentiation in the movie industry. Unfortunately, in the course of the article he gives the best argument against the use of automatic differentiation I’ve heard:

In this paper we will present one approach to automatic
differentiation and describe one application that was used with considerable success during the post-production of Matrix Reloaded and Revolutions.

I rest my case.

7 Responses to “Piponi on Automatic Differentiation”

  1. I don’t get it. What case are you trying to make, and on what do you see it resting?

  2. Oh — that you didn’t like the Matrix movies?

  3. sigfpe says:

    Dude! You weren’t meant to be watching it for the story or the acting. You were supposed to be watching how well the CGI lined up with the live action.

  4. Analogy: I don’t like Shakespeare he was bad at sharpening his quill pen.

  5. Walt says:

    Conal: Yes. I have no substantive point.

    sigfpe: But don’t you see how dangerous that makes automatic differentiation? Once CGI lines up with live action perfectly, both story and acting become literally invisible to the human eye. Soon, every movie will consist of two-hour montages of robots that turn into trucks.

  6. Sam says:

    “Soon, every movie will consist of two-hour montages of robots that turn into trucks.”

    Awesome!

  7. Justin says:

    “Once CGI lines up with live action perfectly, both story and acting become literally invisible to the human eye.”

    …as recently demonstrated by Avatar.

Leave a Reply