Sometime ago I read a paper where hacker can put noises into a deep learning models and/or images so that the network can see fully different objects in the images whereas the human can not detect those differences. This algorithm is called Houdini.
From the current experiments with HTM at least from me, HTM is very robust against noises. Maybe it is a good chance for us for testing if HTM will overperform DL for those problems?
Any comment?
I think the exploits you are referring to are not exploits that would trick the brain, or an HTM system based on the brain. Of course, I’m sure there are other exploits that could affect your brain and HTM systems (see other illusions and brain tricks).
Well I saw this one on twitter yesterday: