gwern comments on Real time video face substitution (and the resulting "psychological hacks") - Less Wrong

4 Post author: Icelus 21 September 2011 11:32PM

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

Article Navigation

Comments (5)

Comments (5)

You are viewing a single comment's thread.

Comment author: gwern 21 September 2011 11:37:05PM 4 points [-]

So much for technology helping us be more verifiably honest: http://hanson.gmu.edu/moretrue.pdf

Now it looks much more like an arms race.

Comment author: Icelus 22 September 2011 05:17:04AM 1 point [-]

I see it as a cat-and-mouse game, like the spam problem.

Or even better is the problem of Photoshopping things which people have come up with some I guess fairly good tool to counter:

“Error level analysis (ELA) works by intentionally resaving the image at a known error rate, such as 95%, and then computing the difference between the images. If there is virtually no change, then the cell has reached its local minima for error at that quality level. However, if there is a large amount of change, then the pixels are not at their local minima and are effectively original.”

From http://errorlevelanalysis.com/

Also people could be more skeptical of video from an unfamiliar source, as they must be getting with photos (magazines covers, especially US beauty-oriented ones) and movies (special effects getting better and better, the young Jeff Bridges in Tron 2.0 didn't fool me but I remember reading a comment or two about people saying they were fooled).

My link to the ELA page and inspiration about using one's judgment came from these not-so-technical articles that cam up in a quick Google search:

http://lifehacker.com/5644259/how-to-detect-a-photoshopped-image

http://www.ratchetup.com/eyes/2007/04/detecting_photo.html

Comment author: DanielLC 22 September 2011 06:26:39AM 1 point [-]

I've heard ELA isn't actually that good. For example, notice that they never actually verified what was photoshopped in the example picture.

Also, based on the description of how it works, I'd guess that it only works if the image is compressed before it's photoshopped. This would seem unlikely if the creator has access to the original image.

I'm not remotely an expert in this. The first paragraph is just me echoing something I've heard from someone else, and the second could easily be me misunderstanding how it works.

Powered by Reddit Powered by Reddit