Rethink the current epistemological approach to medicine and the biological sciences.
-
What's broken:
These fields are awash in a sea of data and knowledge. Much of it unfortunately rests upon a shaky empirical foundation. Reconciling new information with existing information is extremely difficult. Ensuring the integrity of the empirical tree, atop which virtually all new research rests—even more difficult.
The process by which scientific knowledge is shared and reviewed—research papers—is at best an antiquated and inefficient mode of collaboration.
Modern drug discovery is extremely expensive. While advances in bioinformatics and computational power certainly help, they are not a magic bullet. Absent complete confidence that a particular drug will achieve perfect efficacy while not causing unintended consequences, extensive study is required. Studies are extremely expensive to conduct and represent a significant portion of drug discovery costs.
The entire premise of pharmaceuticals may be fundamentally flawed. Expending insane amounts of money in search of magic molecular combinations that target very specific conditions, without complete efficacy, via mechanisms that are often not fully understood, and are burdened with a myriad of interactions and adverse events.
The terminology used in medicine and biomedical sciences—cells, proteins, antibodies, antigens, receptors, viruses, fungi, bacteria, paralysis, inflammation, encephalopathy, neurons, dendrites, axons—all of these aren't actual things but rather human-created constructs purposed to assist us in grasping what we're trying to understand. These constructs, while often quite accurate and helpful—do not necessarily map neatly to what's actually happening, nor do they necessarily lend themselves to understanding complex interactions between other constructs.
Empiricism works great until the tree becomes very deep and complex. Then it's problematic. Especially so if the methods for maintaining and growing that tree are far from perfect. At worst, the entire system becomes a hindrance.
-
How to start fixing it:
This isn't a rant against modern medicine. On the contrary, achieving the best results right now with an incomplete understanding, a probabilistic approach (to steal Peter Thiel's terminology) is the best course of action. Studies for example are probabilistic.
Medicine as it exists today has done an admirable job of making the world a far less miserable and deadly place. However, I fear the current approach may not be suitable for achieving mastery of the human body. The depth of empiricism and complexity of knowledge at that level may extend far beyond human mental capability.
To illustrate this point, assume therapeutic nano bots suddenly popped into existence today, straight out of science fiction. No built-in behavior, but fully programmable at a low level. Using them to cure cancer would be an admirable goal, except for the fact it'd probably be incredibly difficult if not impossible to do considering our current knowledge.
Mastering the human body isn't much different. It might require eschewing the probabilistic approach and replacing it with a mechanical approach. Surgeons largely use a mechanical approach in their work, and that's a major part of why they're so successful. At the scale of bones, muscle, and organs the human body really isn't much different from say, a turbofan engine in terms of complexity. Dealing with systemic disease processes on a microscopic scale however, things become far more difficult. While there are many remarkable successes in this area, there's still a long way to go before achieving complete mastery.
As counter-intuitive as it sounds, it may also be necessary to eschew—or at least cease to completely rely upon—the vast tree of existing empirical constructs we've created. Just as a convolutional neural network isn't aware that the photograph it's uncannily redrawing in the style of Picasso contains human faces, a biomedical application utilizing a similar principle may not need to know what cells are—even as it kills cancer.
Conversely, pretend a bunch of really smart people are locked in a room with just a microscope and an encyclopedic amount of samples. These people somehow have zero knowledge of biology, none at all. After a sufficiently long period of time they'd probably come up with an epistemic model somewhat similar to existing biology, but one that would be completely alien in its terminology, perhaps radically differing in certain key areas—possibly for the better. If they had a way to collaborate from the beginning that was far less cumbersome than research papers, it's probably a good bet their resulting epistemic model would far exceed that of existing biology.
The human body is an incredibly complex product of evolution, and as such it is not easily understood by human minds. We're basically compiler output that's trying to reverse engineer itself.
I'm not suggesting AGI as a solution—just that it would be prudent to apply state-of-the-art weak AI in a fashion that's as decoupled as possible from the current epistemic model, because such constructs impose far too many assumptions that might be very wrong in some fundamental way. Ironically the use of weak AI to achieve mastery of the human body would constitute a probabilistic approach, albeit in an extreme form. If a mechanical understanding did follow, it might be very simple or elegant in nature.