Why will AI always fail to predict the true meaning of a photograph?

As contemporary AI systems attempt to model emotion through visually visible features in a photograph, this research exposes a fundamental limitation: emotion is not embedded in the image alone.
‍
The same photograph can produce convergence, divergence, or complete contradiction of meaning across individuals.
The variance in the responses is not noise, rather an evidence of untrainbilty.

Heard of Kuleshov effect?
The sequence in which you see a set of images/ videos of images you see makes you feel a certain way, and that's where machine fails fundamentally.

For a meaning to truly emerge, we need information about the viewer's memory, finacial status, emotional state (at the time of viewing), political views, cultural meaning of signs, colors and elements.

It is constructed at the moment of viewing.
The nuances get created on the go.

You can't automate becoming.  
You can only automate what you've already been.

Participate in the research