adversarial images

[robomont]

http://www.theverge.com/2017/4/12/15271874/ai-adversarial-images-fooling-attacks-artificial-intelligence

a good read with examples of camo in the new digital world of Ai.
more up to date examples.

[ArMaP]

That's an interesting article, and it shows why I prefer to create my own algorithms.

What this shows is that knowing the algorithm used we know how to fool it.