How bad can AI be

Nuri Erginer
2 min readAug 13, 2021
Evil Eye: Home Decor made of Turkish Glass that protects you from Nazar which is believed to be cast by a malevolent glare, usually given to a person when one is unaware.

I came across an article submitted on 21 Jul 2021 to Cornell University Computer Vision and Pattern Recognition archive related to the unexpected outcomes of the AI research.

We all know the most beneficial practice of AI is in medical studies. Identification of cancer cells from the X-ray images are now mainstream. AI is better then the doctors in some cases because of looking at the images far beyond the human eye.

Nobody knows how the internal process works in the AI neurons. This unknown part of the puzzle may give birth to the worst superpower of AI.

You will understand what I am saying when I tell you more about the article I mentioned above. Imagery used in radiology research, including chest, hand x-rays and mammograms can be used to predict the patient’s race.

Images included patients identified as Black, white, and Asian. For each scan type, the researchers trained the algorithms using images tagged with a patient’s self-reported race. They then challenged algorithms to predict the race of patients in different, unlabeled images.

The scientists who wrote this article are not sure how AI can do that but the prediction results’ accuracy is between %80-%99.

Looking at the medical images with the evil eye can produce more unexpected data. You might think that the algorithm looks at the body mass index or bone density to identify certain demographic groups that seem to act as clues. But that is not the case. Believe or not the evil eye does not look like a human and even does not think like us.

If you don’t know how your enemy thinks, you can’t take any action.

Please don’t consider me an anti-science person. AI is the most advanced technology in human history. But the side effects and the potential use cases by ill-intentioned persons can write another history. The history of human destruction.

There are tons of articles about the AI worst superpower. Most of them are exaggerated. But think of this study: can it exacerbate disparities in health care? We do not know. Black patients and other marginalized racial groups often receive lower levels of care than wealthy or white people.

What I want to emphasize here is that ill-intentioned people can use this technology in other ways. In a way no human even thinks about it. With the power of evil thinking.