A recent article (1) proposes that precision health should prioritize actionable information and long-term user engagement. That's a lot of words but it seems to make sense. Hardware and software companies have been on the prowl to sell "precision health," even though they may not know what it means. This could be a shocker for the statisticians, but health is not precise, not by any stretch of the imagination. As they roam the hallways with pocket calculators and actuarial tables, seeking higher and higher "precision," it is important to recognize that the human remains to be the most complex and enigmatic to figure out.
Health is a difficult construct to define. The regulators have gotten wind of "risk," recently and they are clamoring for "risk stratification." None of these people have had any formal education in risk or economics, but they feel they are experts on policy involving the same. They don't even seek information from other industries and that has been symptomatic of the entire healthcare value chain that includes manufacturers, providers, and payers, who seem to know pretty much everything there is to know. Not so. They may be brain surgeons but there are plenty of rocket scientists outside their domains. It may be better to talk to them, before plunging head down into the abyss.
A century of "development," appears to have increased lifespan by 2000 days. But from a utilitarian perspective, the incremental 2000 days gained by fantastic pharmaceuticals, crazy yoga and jumping up and down the whole day, do not seem to add much value. More importantly, the time gained generally reduces the quality of life, impacted by pain, hospitalization, and a lack of flexibility to make decisions. The human brain appears to deteriorate past allowed time and the individual behind the smoke screen suffers. The manufacturers who plunk down close to 100 billion every year into R&D do not seem to have any acceptable answers. The providers, left in the lurch to care for their patients who may not even recognize them, suffer equally. Meanwhile, the accountants at the payers are cranking up their calculators so that they can meet the quarterly EPS.
It is a painful movie to watch. As the brilliant folks in Washington figure out how to save themselves and perhaps the country, there is significant suffering across the landscape. Losing a life is unthinkable, losing a mind is equally traumatic.
They have been riding high. The abandoned and somewhat less sexy field of Statistics has taken the business world by storm. Bottling old wine in new bottles certainly helped and now both venture capitalists and operating companies may be heading for a hangover. Engineers and statisticians have always wanted to be scientists and now they are crowned as such. There is a .ai company formed every 15 minutes by graduates of prestigious universities and there are capitalists with sacks of money willing to entertain them. As we have seen before, this movie will likely end in tears for many.
Data is certainly a good thing and applying "science," to it could also be good. But those who assert their "scientific credentials," based on regressions and neural nets should be aware that the slide rules they are using have been available for nearly half a century. Mathematics does not fade but asserting old ideas have suddenly sprung to life certainly shows the maturity and age of the emerging "scientists." Consulting firms have always been creative and some of the most famous ones, who could hardly spell "data science," just a few years ago are now pretending to be experts at it. Conferences are plenty where the scientists meet their seekers and the vendors portray their wares almost like the bartering that was routine a few centuries ago. They flow tensors, cognitive networks and even hardware in a Pizza size box, that apparently has solved all the world's problems, already.
Stop hiring "data scientists." They are ordinary human beings with bias and they could do your companies a lot of damage.
As the first biological entities emerged out of the water and into the land, the battle was just beginning. The toxic air was oxygenated over time and as complexity increased, they had to develop sophisticated systems to breathe. The toxin turned into life-giving Oxygen and over a million years, humanoids experimented with systems that could shuttle the magic molecule to power their systems.
Allergies and asthma followed modern humans as they transversed the globe with their badly designed respiratory systems, prone to catastrophic failure and that killed them in large numbers. Later, modern medicine will keep them alive for a few more decades but they often succumb to the inability to oxygenate. Their nemesis, the Virus, attacked their Achilles heel as most died of the common cold and some of the more advanced versions, aptly named Pneumonia. Recent news (1) that claims advances in bronchodilator drugs in asthma is certainly welcome news.
Humans, fragile and badly designed in most systems, do not appear to be robust enough to move to the next stage.
It appears that the completely archaic notion of mass-produced drugs for the average patient is about to change (1). The manufacturers paid lip service to personalized medicine for nearly a century and it was clear that their heart or business models were never in it. The normal function may have done as much damage to humanity as nuclear weapons, for those who adhere to it blindly believe in averages and standard deviations based on a manufactured construct. The only redeeming quality of humans is that they are different and diverse. As the men in power separate the weak from the wealthy, the struggling from those who never struggled, the golfers from those who cannot afford a club, the academics from practitioners, the atheists from the religious, the North from the South, the West from the East, they miss an important point - every human on Earth is different, regardless of the visible features they exhibit or where they originate from.
The design of clinical trials seems to fail this basic notion. Pushing humans through protocols like cattle through a food manufacturing company is not the best way to discover drugs. It is certainly the best way to reduce costs and to prove to the regulators that something important has been done. In the process, they left large underserved populations in the lurch and pumped those who take the medicine with a dose that is suboptimal. Emerging technologies are immensely capable to figure out who will benefit from a drug and who will not and at what quantity. It is time statisticians left the industry as their contributions do more harm than good, not unlike the insurance industry, clinging to actuarial tables.
Now, available technology can titrate every individual to the optimal dose and we do not need, "population statistics," to approve or to disapprove drugs. If the regulators do not return to school to learn what has been happening, they will continue to make bad decisions.
(1) Digitization of multistep organic synthesis in reactionware for on-demand pharmaceuticals
Philip J. Kitson, Guillaume Marie, Jean-Patrick Francoia, Sergey S. Zalesskiy, Ralph C. Sigerson, Jennifer S. Mathieson, Leroy Cronin*
Recent news that a single blood test could provide the diagnosis of eight common cancers with 99% specificity (1) is a constant reminder that medicine is still stuck in archaic and invasive procedures to detect, diagnose and treat ailments. With a high concentration of human resources in provider settings, medicine has been slow in embracing emerging technologies and ideas, outside the domain. And this attitude is shared across the healthcare value chain including manufacturers, payers, and regulators.
It is unfortunate. Granted, Biology still remains to be the arena where humans could not progress exponentially. Their brains, with millions of years of deterministic training, have been well specialized to dominate engineering and chemistry. However, they could not understand the marvelous machines assembled by nature from a single cell organism to somewhat more complex humans, with any level of precision. Nature has had time to perfect designs of such beauty and humans, ever curious, have been trying to walk up to the cup of knowledge. But it has not been. Fossils indicate attempts at brain surgery many hundreds of thousands of years ago and despite higher structural knowledge, we have not advanced sufficiently to a differentiable plateau. In most simpler fields, we have demonstrably shown that humans are the weak links in decision processes - from transportation, energy, manufacturing and even, finance.
It is a conundrum. We are stuck - great strides in deterministic sciences do not translate into domains of high uncertainty and diversity. And, those who practice in these complex domains seem to have their blindfolds on as if they have nothing more to learn. Diagnostics could provide the impetus to move higher - serum and stool harbor such information content, it is a shame we have not figured it out.
(1) Detection and localization of surgically resectable cancers with a multi-analyte blood test
1. Joshua D. Cohen1,2,3,4,5, Lu Li6, Yuxuan Wang1,2,3,4, Christopher Thoburn3, Bahman Afsari7, Ludmila Danilova7, Christopher Douville1,2,3,4, Ammar A. Javed8, Fay Wong1,2,3,4, Austin Mattox1,2,3,4, Ralph. H. Hruban3,4,9, Christopher L. Wolfgang8, Michael G. Goggins3,4,9,10,11, Marco Dal Molin4, Tian-Li Wang3,9, Richard Roden3,9, Alison P. Klein3,4,12, Janine Ptak1,2,3,4, Lisa Dobbyn1,2,3,4, Joy Schaefer1,2,3,4, Natalie Silliman1,2,3,4, Maria Popoli1,2,3,4, Joshua T. Vogelstein13, James D. Browne14, Robert E. Schoen15,16, Randall E. Brand15, Jeanne Tie17,18,19,20, Peter Gibbs17,18,19,20, Hui-Li Wong17, Aaron S. Mansfield21, Jin Jen22, Samir M. Hanash23, Massimo Falconi24, Peter J. Allen25, Shibin Zhou1,3,4, Chetan Bettegowda1,2,3,4, Luis Diaz1,3,4, Cristian Tomasetti3,6,7,*, Kenneth W. Kinzler1,3,4,*, Bert Vogelstein1,2,3,4,*, Anne Marie Lennon3,4,8,10,11,*, Nickolas Papadopoulos1,3,4,*
Deep learning has been in vogue. Combining ideas from the 60s and an insane amount of computing power, the search giant and others have been learning deep - mind and all. This is good news, gentle tricks on established mathematics seem to have reduced overfitting and accelerated "learning." But, technologies based on unlimited resources and computing power, tend to be lazy and deep learning seem to have all the characteristics. Some even call it "Artificial Intelligence," even though there is nothing artificial or intelligent about it.
Humans have been fascinated by their brains forever. They have searched for the mind and soul in a few pounds of messy grey matter they carry on their shoulders but found nothing. When the computer scientists arrived who could create "General Artificial Intelligence," by assembling dumb silicon and using dumber games, their age showed why wisdom is not that easy to attain, Ph.D. or not. The search giant has been on a prowl, picking up anything that ends in .ai for a premium and as the greatest technologist of all times who invented the electric car and electrified space travel proclaimed that only he knew what AI was all about, we seem to have arrived at ego driven emptiness.
Get used to it. Nobody is intelligent enough to create "general artificial intelligence." Those who harbor higher than average brain cells have headed in the opposite direction by proclaiming that knowledge results from understanding and not modeling ideas. Therein lies the conundrum, as the technologists rise without human contact and attempt to travel to Mars, there appears to be a great vacuum between knowledge and know-how. There is a distinct difference between the two, the former conquered by philosophers and the latter by engineers and it is important to distinguish between the two.
It is time to look forward and abolish ego-driven behavior. Those who are prone to it should be told that they are no better than the worst of humans.
For nearly hundred years, every field, life-sciences, manufacturing, high-energy physics, economics, healthcare, and others relied on basic statistics and a rather crude assumption that everything follows the Normal function. There is nothing wrong with the assumption but in a regime that works on the tails, the observation that something works for the population has little practical value. In life sciences, they have been inventing mediocre therapies for over a century, as the clinicians, their regulators, and aiding statisticians have been enamored by the mighty "p-value." They have been striving to prove that the incremental average benefits delivered to a large population are a lot better than life-saving therapies for a few. In manufacturing, they have been optimizing with constraints in an attempt to save nickels and dimes. Lean, mean and mighty, their determinism has led to incorrect decisions in the presence of uncertainty. In healthcare, they have been waiting for the protocols to change based on simplistic observations of small samples. Meanwhile, half the healthcare costs in the World could be attributed to a handful of related disease states. In physics, stuffed with engineers, they have been deploying heavy steel for finding particles and hearing waves, based on basic statistical notions. Even with that, they will be the first to admit that they do not yet know 94% of it. In economics, they have been inventing theories based on regression and even winning Nobel prizes but it is unclear if they are creating insights. Some of them ventured into even making money and some have failed spectacularly as would have been predicted by their own theories. Overall, if one can write down an equation for a process, it is symptomatic of the fact that she has not understood it. The practitioners, who seem to cling to the past are being rendered less effective in the presence of those who look forward.
A generation seems to have wasted their time adhering to basic principles laid out a century ago. Lately, statistics have been made sexier by better naming - now called, "Machine Learning." One has to admit it does sound a lot better, but has anything changed? In a world full of practicing scientists, who have been trained to make equations for everything, we are approaching a significant discontinuity. Machines are certainly marching forward but not because they know statistics but because they do not. Such is the state of affairs that a systematic education delivered by the greatest institutions in the world prepares the next generation to fail with high certainty. Meanwhile, machines can see, hear and make decisions in the presence of uncertainty. As we hunt for fossils to establish our own identity in a process that seems to have taken a long time, machines with no emotions and even less historical baggage, rise. Are humans being rendered irrelevant? As the greatest living physicist warns of ETs, as the world's richest and powerful worry about AI, and as the most powerful man on Earth worry about if his hair is falling straight, we have arrived at the precipice of a great discontinuity.
As they moved out of their homeland in Africa, humans must have made important calculations based on uncertainty. As they descended from the trees into the African Savannah, a few million years prior, they knew the regime was shifting. With dangers all around them, mighty beasts who could maul them in a single swipe, they made decisions based on uncertainty. Their initial journeys into the Middle East and South Asia, closely followed by those who went a bit North, seem to have provided a level of safety. They advanced culture and boredom, the latter most important for the development of human psyche. As the caves in Southern France prove, they could certainly rise above determinism and engineering, very early in their progression.
The regime is shifting again - the opponents are not as gentle as the Neanderthals. Machines are brutal and they are immensely capable. Humans, the victors of past conflicts, are starting from a position of a great disadvantage because of their education of the past. The end of statistics, a figment of the imagination of the most recent generation, is very near.
A new study (1) demonstrates that there are significant common factors that influenced the evolution of past societies. One clear and obvious trend is toward more complex arrangements. The researchers analyzed a large database spanning over 400 societies over 10,000 years. The results show that human societies follow a singular blueprint as they evolve. This appears to have many implications for future designs.
Size, decision controls, information systems, literature and economic development are features that all contribute to a singular measure of social complexity (1). Given the large data set, the researchers may be able to assess the level of development in contemporary societies as well as speculate on eventual outcomes. The fact that most societies show growth and predictable decline means that humans are stuck in a blueprint that was put in place a few million years ago. With complexity grow arrogance and inequality and those climbing to the top of the pyramid seem to lose context and wisdom. Given the data, it appears possible to predict the half-life of the present societies with high accuracy. But it is unclear if such information could have any practical effect on policy that could reverse the predetermined course.
On the positive side, the level of knowledge and sophistication seem to have equalized across countries and societies. Those who were ahead have been arrested by ignorant leaders and those behind are driven by a desire to succeed. In either case, modern humans, already long in the tooth are due for a reset. It is a shame that they could not learn from the abundance of historical data using their nascent tools in "machine learning."