So for spot the difference images, I cross my eyes. The differences jump out.
HOWEVER
In this one, and I know it’s two different images, but when I tried the same trick, the image became accurately 3D. Some fuzziness, of course, but there were proper layers to the resulting depth.
I’m very curious as to the implications of how this A.I. may have been trained…
Added the result.