Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
user:jenda:challenges [2018/11/07 21:04]
jenda [Crazy/insane stuff]
user:jenda:challenges [2019/01/12 08:21] (current)
jenda [Crazy/insane stuff] +DLmind
Line 51: Line 51:
   * There exist image/audio patterns and [[https://​en.wikipedia.org/​wiki/​Denn%C5%8D_Senshi_Porygon#​Strobe_lights|video sequences]] that trigger seizures in some people. Take such sequence, show it to someone while recording EEG and measure how much "​unusual"​ brain activity it has generated - voilà, you now have an utility function (video) → (how good it is at causing seizures). Now use some kind of evolutionary algorithm/​optimizer/​hill-climbing to maximize this function. You might find a real-world [[https://​en.wikipedia.org/​wiki/​BLIT_(short_story)|BLIT]]! (yes, I know that optimizing a costly black-box function across video sequences is hard, but hey, the AI field solved similar hard optimizations in the last few years)   * There exist image/audio patterns and [[https://​en.wikipedia.org/​wiki/​Denn%C5%8D_Senshi_Porygon#​Strobe_lights|video sequences]] that trigger seizures in some people. Take such sequence, show it to someone while recording EEG and measure how much "​unusual"​ brain activity it has generated - voilà, you now have an utility function (video) → (how good it is at causing seizures). Now use some kind of evolutionary algorithm/​optimizer/​hill-climbing to maximize this function. You might find a real-world [[https://​en.wikipedia.org/​wiki/​BLIT_(short_story)|BLIT]]! (yes, I know that optimizing a costly black-box function across video sequences is hard, but hey, the AI field solved similar hard optimizations in the last few years)
     * [[https://​www.biorxiv.org/​content/​early/​2018/​11/​04/​461525|This paper]] should be a good start.     * [[https://​www.biorxiv.org/​content/​early/​2018/​11/​04/​461525|This paper]] should be a good start.
 +    * And [[http://​perceive.dieei.unict.it/​deep_learning_human_mind.php|here]] they kind of build this model for humans, but don't test if it can be used to generate adversarial examples.
 
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Noncommercial-Share Alike 4.0 International
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki