How AI Will Change Medicine by Bypassing Human Brains
The human brain has been moving medical science forward for centuries, but now that we have fully entered an age of rapid advances in machine learning, the human brain may be beginning to hold us back. As a result, we are about to embark on the biggest change in the scientific method since Aristotle. I’m talking about how artificial intelligence (AI) will alter science by bypassing our human brains. How? Because AI removes the scientific middleman from medical research. I’m going to explain this, but first, you need to understand the concept of reductionism.
All Science Is Reductionist
Reductionism is an approach to understanding the nature of complex things by reducing them to their fundamental or simpler parts. Our human brains seem to comprehend best when we reduce the complex whole into simpler parts. And this is no different in the world of medicine. At a basic level, I’ve explained before that the hamstrings, for example, as a distinct entity really don’t exist. They are one part of the bigger machine (the full interconnected body) that we give a name to in order to break the body into simpler parts. But reductionism can hurt patients and physicians, too, like to focus on the distinct part or system rather than the body as a whole.
This is much broader than one body part, however. Take, for example, a drug that blocks enzyme X, which helps disease Y. Because we only ever understand one of the myriad of complex systems that enzyme X impacts, two new problems then develop in systems we know little about. For example, an over-the-counter anti-inflammatory, like Motrin, blocks a chemical that helps inflammation and pain, but on the flip side, it also dramatically increases heart-attack and stomach-bleeding risks. So all of that Motrin-related morbidity and mortality, I would argue, is a direct result of reductionism. In this case, blocking the enzyme to help pain and inflammation, but, in the process, putting other body systems at risk. So you can see how reducing the patient’s problem to its one simpler part or system may hurt the whole patient. So if the human brain and its reductionist approach to medicine and the human body is holding us back, what is the answer?
Remove the Scientific Middleman?
What if everything from statistical analysis to medical diagnosis to our reductionist understanding of the body has been holding our medical progress back? What if we remove the scientific middleman altogether? Currently, for example, during an MRI we use a computer to collect data on complex changes in a magnetic field. We then assemble this data into an image a human can read. The physician, a human, then places a label, a diagnosis, on that condition. Medical resources are then activated to help the patient based only on that diagnosis.
So what if we remove the middleman in this equation? It might look like this: The raw MRI data is processed through AI, which then tells us what the treatment is, bypassing the diagnosis altogether. In fact, we wouldn’t even need to know what diagnosis the MRI points to in order to be able to deploy a treatment. Really, no diagnosis? That’s medical sacrilege! Yes, and the reason why it could work is because it removes the middleman’s reductionist approach. Keep reading.
AI Removes the Scientific Middleman’s Reductionist Approach
How did I come to such an “out there” conclusion that AI would change science and medicine? We are in the midst of a knee-microenvironment stem-cell-treatment study, in which we are measuring 25 cytokines and growth factors in 200 stem-cell-treated knees. We take a sample of the synovial fluid before and after a stem cell treatment and get a slew of data.
It became very clear very early on that the data we were collecting was chaotic and extremely complex. However, all we really cared about was finding a pattern that was associated with a positive outcome at that stage in the study. So why would we waste time on a reductionist theory of knee osteoarthritis cytokines and growth factors when we could use an AI neural network to associate a pattern of chemicals in the knee to a positive outcome?
Neural networks, or machine learning, are becoming more and more common in healthcare. In a neural network, rather than having to be hard-programmed, the machine learns from the patterns in the data. These connections may be invisible to our reductionist human brains. For example, one doctor only has his or her own experiences, and sometimes he or she has biases, which create flaws. A neural network can look at the experiences of thousands of doctors and pick out trends that traditional statistical analysis would never discover.
After our experience with our study, it hit me: what if medical research looked more like this? As much data as possible is collected. This data is run through an AI neural network, which decides, based on the collective data not a body-specific assigned diagnosis, which treatments are associated with the best outcome. Throughout the research, the system constantly updates itself. This made me realize that AI neural networks are really going to change medicine and science by bypassing what our human brains are capable of.
You can learn more details about our study from this recent blog post: “My IOF 2017 Talk on Stem Cells, Knee Microenvironment, and Machine Learning.”
How Will AI Change Medicine and Science?
Early AI, like we have now in these simple neural networks, can help us recognize complex patterns in chaotic medical and scientific data that our human brains really don’t have the ability to understand.
As the AI technology develops, AI will show us relationships and mechanisms in medicine in science we can’t comprehend. This will mean the rise of what I call “black box” science.
“Black box” science means that understanding, or knowing, the “why” is really more of a luxury, not a necessity. The “why” will become less and less important to advances in medical research. In addition, finding out “why” is crazy expensive.
The upshot? Future medical scientists will help AIs recognize complex patterns in data, but those scientists will likely care less and less about why those patterns exist. As AI changes medicine by bypassing human brains, this will make many academics uncomfortable at first, but once embraced, AI will supercharge medical research. As a result, real-world medical progress will explode as it bypasses our need to break complex systems into specific conditions the human brain can understand.