AI Kills (but that's not certain)

Friends told me that to increase my reach, I should write longer, insightful thoughts. So, here you go, enjoy reading.

Picture this: you are in a very advanced hospital, and robots along with AI systems are diagnosing and treating patients with complete precision. Great, am I right? Great, because AI offers more rapid and precise diagnoses, more personalized treatment, and, thus, less human error. It is almost going to be like having Dr. House on the speed dial — well, without the sarcasm. However, here is the problem: Artificial Intelligence is not perfect. Just like my grandma's recipe for borscht, some details can only be valued by a grizzled human. And AI is only as strong as the data on which it is trained. And if you feed it biased or incomplete data, it starts regurgitating all kinds of frighteningly incorrect results.

Now, a little personal experience: one of my friends had a mole on his back. Trusting blindly in technology, he decided to get a diagnosis from an AI-powered app. It stated that the mole was not dangerous. Months later, visiting a dermatologist on an unrelated issue, the doctor pointed out that the mole was actually melanoma. Luckily, caught in time, but the point here is the danger of relying only on AI. In fact, the black box problem is the biggest for AI in medicine. Now, many AI algorithms, especially deep learning models, can be incredibly complex. Even the people who develop them often don't have a complete understanding of how they make certain decisions. Now, picture that you ask a doctor why he reached a diagnosis, and you get nothing but a shrug. Safe to rely on a black box for your health? No.

However, bias is not just a human glitch — it can affect AI as well. For example, if an AI is mostly trained from data on a certain group of people, then its accuracy for other groups goes down. Actually, this can make health disparities worse; especially in already underserved populations, this is a major problem. It's like training a parrot to say "Hello" and expecting it to recite Leo Tolstoy all of a sudden. Recently, some AI systems have been found to be less accurate in diagnosing certain conditions for women compared to men. That's not because AI is inherently patriarchal; instead, it's often because women are so dramatically underrepresented in historical medical data. The upshot? Mistakes that could be pretty serious.

AI can't empathize, which is the quality in humans that defines one's ability to understand and share the feelings of another. One can never replace the comforting touch of a doctor or a nurse or the empathy you feel when they ask you about your problems. An AI can be programmed to diagnose a set of symptoms, but they cannot tell you to relax as they take your hand and comfort you through a diagnosis. The value of human judgment is enormous, and it has been crafted over years of training and experience. There is a difference between a cook having a recipe and a chef making a masterpiece. Doctors have personal experience and patient history that an AI just could not have.

The regulation of AI in healthcare is a big ask. Standards and protocols for the technology are still being developed, let alone the myriad ethical questions that are raised. For example, if an AI system makes an incorrect diagnosis, who is liable: the developers or the doctors? It's something of a legal and ethical minefield. Laws like the GDPR have put problems of privacy and consent at the forefront. AI systems require massive quantities of data, which raises major questions of patient confidentiality and the safety of sensitive information. It's like trying to write a diary in a glass house.

Ongoing education and training are important to minimize such risks for healthcare professionals. The strengths and weaknesses of AI need to be really appreciated. Besides, the development of robust AI assimilation frameworks contributes to positioning technologies in ways in which they must remain a tool, not a replacement for human expertise in any situation. Transparent, fair, and accountable AI assimilation worth conversation is only possible as a result of cooperative work between AI developers, medicals, and policymakers. Kind of like building a bridge — each part is integral to the whole.

The future of medicine will be one where AI and human intelligence will take up tasks in equal measure. While AI will be busy dealing with data analysis and routine cases, making the life of physicians much easier and moving them to more complex cases and interactions with patients, it's kind of like putting up a perfect partnership and not a capture of pirates. As promising as AI is to revolutionize healthcare, at the same time, overdependence can be harmful. Healthy skepticism and regard for human expertise remain indelible. Medicine is, after all, as much an art as it is a science. Next time one is awed by the wonders of AI, remember the sage words of my Russian grandmother: "Technology is great, but never underestimate the power of a human touch" (possibly, she never said these words, and I made them up for effect).

Image is created by ChatGPT.