What Is Really Wrong With AI?

In AI, We Trust …

Humans Imitating The Parrot

There is grave danger when humans take anything as a divine source of reality. It hijacks every sensory and sensual organic interfaces where they source and ground base their decisions. Humans don’t do well with logic. The truth should be an instinctive thought of lemma. If we have to think about it, we haven’t yet grasped it, for example the understanding of a common human towards time as a non-linear property. Yet few are humble enough to recognize their inability to identify their mortal biases, flaws and fallacies.

AI is trained on our biases. Models sometimes correct for these. I’d even argue that AI magnifies our biases. I was having this back and forth with someone this past week. Unfiltered text image generation models seem to represent a white male when prompted for a CEO 100% of the time. In reality 26% of CEOs are women. It takes an already existing bias and magnifies it even further. — Mike T., 2023–10.31

We can correct logical fallacies using logic, there are SAT solvers and theorems for those, but we cannot correct preferences, e.g. aesthetics and cultural sense of aesthetics, with no logical references. — 2023–10.31