Skip to Content

Can AI Keep You Healthy?

A Chinese entrepreneur wants to track your health data and suggest ways to improve. But are computers really smart enough to make sense of all that information?
Yann Kebbi

“This smart mirror isn’t very smart,” says Jun Wang, standing in front of a full-length mirror wearing designer jeans ripped at the knees. “It’s just a camera and a mirror,” he says, looking mildly distressed—or as distressed as possible for a man whose face is perpetually unperturbed. “What I want is a mirror that does a 3-D scan of me here,” he says, using his hands to trace the contour of his thighs, “and here.” Wang indicates his belly, which is lean. “We want an exact 3-D figure of you: the fat, the muscle—your entire body shape, plus facial recognition, and what’s going on with your skin.” He points to the top right area of the mirror. “And I want readouts about my health up there, next to where I’m brushing my teeth—my weight, blood pressure, and heart rate, and how does that correlate with my DNA?”

This as-yet-unrealized smart mirror is just one of several gadgets that Wang, a 41-year-old biologist and computer scientist turned entrepreneur, says he is building. The devices will help gather, analyze, and display a crush of health data he wants to collect about himself—and, he hopes, from millions of others. This is why Wang cofounded iCarbonX (ICX), a highly ambitious, if quixotic, personal-health company based here in Shenzhen, in southern China.

ICX wants to capture more data about your body than has ever before been possible. It starts with your DNA sequence and includes data from Fitbit-style wearables that measure your steps, heart rate, and sleep patterns. Add frequent blood tests to measure various proteins and enzymes that can, say, reflect the health of your heart or signal very early signs of cancer. Include monitoring of the ever-changing levels of metabolites produced by the body as it processes food; traditional blood tests on levels of cholesterol and glucose; heart data from an EKG; and information from your medical history. The goal: continuous monitoring of your health and suggestions of adjustments you might make in your diet and behavior before you slip from being healthy into the early stages of an illness.

This sounds a little like personalized medicine, which has been discussed for years. But for Wang, it’s not just about treating disease. It’s also about what might be termed personalized health. “Right now you don’t know about your temperature, or your pulse, or the microbes inside you that affect your emotions,” he says. “Or what to do if you have an allergy, or you want to lose weight because you’re fat.”

The devices will help gather, analyze, and display a crush of health data he wants to collect about himself—and, he hopes, from millions of others.

This vision of personal health monitoring is becoming achievable in part because of dramatic cost reductions for sequencing DNA and measuring the many thousands of biological compounds and processes that regulate the body. What all that means for any one of us, especially when all the readings are combined, is unclear. But ICX is part of a new wave of companies that figure they can find something meaningful in the data and enable medicine to stop merely reacting to an illness you have; these companies want to keep you healthy at a fraction of the cost. Unlocking this puzzle, with its millions of moving pieces, is where AI and other advanced computing techniques will have to come in. “AI is how we can take all of this information and tell you things that you don’t know about your health,” says Wang.

Assuming it works, putting all of this together will not be cheap. As CEO of ICX, Wang has raised $600 million in funding for the effort, a remarkable amount for a project offering high-tech tests for healthy people. “But he’ll need it, and probably more, with everything they want to test,” says Eric Schadt, a computational biologist and mathematician who recently stepped down as director of Mount Sinai’s Icahn Institute for Genomics and Multiscale Biology in New York. Schadt has launched his own health data company, called Sema4, which is scanning genomes and molecular biomarkers.

ICX is using its pile of cash in part to invest in or acquire companies that might contribute to Wang’s holistic vision. This includes a $161 million stake in Colorado-based SomaLogic, which is working on a chip that can measure 5,000 proteins in the blood; more than $100 million in PatientsLikeMe, a company in Cambridge, Massachusetts, that provides an online platform for more than 500,000 patients to share experiences, metrics, and feelings about their health and diseases; and $40 million in AOBiome, also of Cambridge, which sells spray-on microbes that it says make skin healthier. ICX also recently invested in HealthTell of San Ramon, California, which identifies antibodies from a blood sample as clues to the presence and progress of diseases including cancer and autoimmune disorders. Additionally, ICX is also collaborating with several companies in China.

Yann Kebbi

Tying this eclectic alliance together is an aggressive effort to build an artificial-intelligence system that will attempt to analyze all this data. That’s being led by Israel-based iCarbonX-Israel, which ICX acquired last year. Founded in 2005 as Imagu Vision Technologies, the company develops software to interpret CT and other medical images. Now Imagu’s engineers are working with counterparts at ICX to create what they call a “virtual health brain” that will interpret the thousands of data points ICX wants to collect on each customer. “We want to create a tool that not only analyzes data but offers ways to help people improve their health, like how to alter their diet,” says Imagu CEO and cofounder Mor Amitai.

“If this all sounds ridiculously complicated, it is,” says Wang, smiling in a way that blends reassurance—which undoubtedly is appreciated by investors—and bemusement, as if he knows that what he is proposing sounds a bit daft. The question, then: can he use his money and technical savvy to revolutionize medicine?

Precision health

A tall man with short black hair, Wang strolls coolly through his company’s headquarters, a Silicon Valley knockoff with open workstations, glass-walled conference rooms, a gym, and a café always stocked with food, healthy drinks, tea, and coffee. It’s on the third floor of an industrial-park building in a complex of similarly unexceptional structures tucked between two sprawling, wooded theme parks called Happy Valley and the China Folk Culture Village. In the back of ICX’s HQ is Wang’s office, a comfortable niche with deep leather chairs and a private conference room—a business setting that is a long way from where Wang started, as an academic researcher sequencing DNA at Beijing University in the late 1990s.

Wang authored over 100 studies as a professor at the University of Copenhagen and as a bioinformatics whiz at the Beijing Genomics Institute (BGI), which he helped found in 1999. BGI was the organization that led China’s relatively small contribution to the Human Genome Project, a worldwide effort in which several countries worked on different segments of the human DNA sequence published in 2003. BGI later churned out the first complete DNA sequences of an Asian person, a strain of rice, the SARS virus, and the giant panda. During his stint as BGI’s CEO, Wang helped build the company into one of the largest sequencing operations in the world. In 2016, it had revenue of $250 million, and this summer it issued an IPO. Wang remains a major shareholder and a member of the board.

“You also need millions of people—maybe as many as 10 million people—to get meaningful signals for common diseases.”

But he left BGI in 2015 because he was frustrated by the limits of genomics. Specifically, sequencing DNA doesn’t provide much insight into the health of most individuals. Scientists have found countless DNA markers that seem as if they should help determine whether a person is healthy or sick. But those markers have turned out, nearly 15 years after the completion of the Human Genome Project, to make less of a difference than originally thought. With the exception of certain rare genetic mutations, DNA is just one determinant of a person’s medical fate. “It turns out you also need to know about proteins, and metabolites, and all the rest,” says Wang.

Soon after his departure from BGI, Wang formed ICX, knowing he would do something with AI and health. But he wasn’t sure exactly what data besides DNA the company could, or should, collect. To figure it out, he met with a range of experts and companies, including a pivotal meeting in July 2016 at the Original Max’s restaurant in Burlingame, California, near the San Francisco airport. In town pitching ICX to investors and prospective partners, Wang had arranged to see Jamie Heywood, the cofounder and chairman of PatientsLikeMe, who was visiting from Boston. As they sat in an orange-and-yellow plastic booth in the truck-stop-style café, it didn’t take Heywood and Wang long to realize that they shared a fundamental exasperation with the limitations of today’s medical practices. Giving people more data seemed like a promising route. PatientsLikeMe, which runs a service where thousands of members discuss their various chronic diseases in online forums and provide metrics about their health and the progression of their disease, had already shown the value of careful health tracking by individuals. Drinking coffee, Wang and Heywood dissed classic medical testing, which tends to be static, with one test taken at a time—an EKG in a clinical setting every year or two, for example, or when symptoms seem to warrant it. “We got excited about the possibility that we could discover the early stages of when a person shifts from good health to, say, becoming a diabetic,” says Heywood, an MIT-trained engineer. “We both agreed that the technology is there, or is close to being there.”

Heywood, who is a fast-talking bundle of energy where Wang conveys a steady calm, suggested that such a profile should also include the sort of behavioral and personal data collected by his company. Information that people share in the forums of PatientsLikeMe—on such issues as the health impact of stress at work—provides valuable clues to other members on how they can better manage their chronic conditions. Why not help healthy people use similar tools and data? “It took about five minutes for Jun and I to realize that we could do this,” he recalls.

Heywood brought something else to the table: his company had built a computer platform designed to analyze data reported by its half-million users. But it’s not yet clear that combining all the data that ICX and its collaborators want to capture will be meaningful. Nor is it likely that AI will find significant correlations in the data unless ICX lures millions of people to its service—and even that many might not be enough. “ICX will struggle,” predicts Eric Schadt of Sema4. “You also need millions of people—maybe as many as 10 million people—to get meaningful signals for common diseases.”

Yann Kebbi

Wang readily acknowledges the challenges. “To do everything we want will take many years,” he says. When asked about the need to test large numbers of people to discern signals in the noise of all this data, he says that ICX is looking to enroll at least one million people in the next five years. “China has this big population, so I’m not worried about this,” he says. He adds that as disposable income increases in China, people want to spend money on their health.

Wang admits, however, that he doesn’t yet have a clear business plan. “I tend to think about the right thing to do with the science and the product first,” he says. “Then I figure out the business model. Investors are okay with this. They don’t want short term.”

Smoke and mirrors?

Underlying ICX’s challenge are also some fundamental questions about how to integrate artificial intelligence into health care. There’s little doubt that advanced computing will eventually provide a huge boost toward making sense of all manner of health and biomedical data. And Wang is not the only one with business ambitions for the technology. According to CB Insights, which tracks venture capital investments, investors are funding 106 startups in AI and health—up from a handful a few years ago. They’re pursuing everything from mental health and drug discovery to lifestyle management, virtual assistants, hospital management, and medical imaging and diagnostics. While this sounds impressive, AI so far has failed to make a substantial impact on most of medicine and health care. “In certain niches, AI is here and has been for years,” says Marty Kohn, a physician and the former chief medical scientist at IBM, who helped develop IBM Watson Health. “But it’s not happening at scale. And it hasn’t yet helped large numbers of patients.”

One reason is that it’s incredibly difficult to interpret the crush of data. “I think AI has tremendous potential,” says Leroy Hood, the president and cofounder of the Institute for Systems Biology in Seattle. “But the claims for AI and health care are very overblown.” Most companies, he suggests, “don’t do real science.”

A longtime pioneer in finding tools for understanding the body’s complex functions, Hood is a cofounder of Seattle-based Arivale, another health data company. Two years ago, Arivale started offering its own version of lifestyle, wellness, and molecular testing, coupled with personal coaching. In July 2017, Hood and Arivale published a small study in Nature Biotechnology that he says provides a proof-of-concept analysis of what the researchers call “personal, dense, dynamic data clouds” measured in healthy people over time. They used advanced algorithms to make correlations for 108 subjects who took dozens of health tests and measurements. Some of the participants learned that they had vitamin deficiencies; others found they had early signs of inflammatory bowel disease or diabetes that needed tending through diet or supplements. These results, however, are preliminary, and far more of them will be needed to separate real findings from the firehose of data.

There’s little doubt that advanced computing will eventually provide a huge boost toward making sense of all manner of health and biomedical data.

As for Wang, he is experimenting on himself with still more ways to acquire such information. As he continues his tour of ICX’s headquarters in Shenzhen, he points to a toilet just off his office where he collects plastic bags of poop for his daily microbiome analysis. Wang describes plans to build a “smart toilet” that will capture and analyze one’s waste and feed it into an AI-generated personal profile. “We have the technology to do this,” he says. “We have the algorithms. It will be cheap, something like $200.” Wang next lifts up his sky-blue polo shirt to show off a wireless continuous heart-rate monitor.

One wonders, however, if millions of healthy people will be as obsessed as Jun Wang is with collecting so much data on themselves. The question seems to take him by surprise, momentarily roiling his composure. He knits his brow, looking as if he couldn’t imagine that other people might not want smart mirrors and toilets, frequent blood draws to measure thousands of metabolites, and heart monitors taped to their chests. “I’m not asking everyone to do this,” he finally says. “People choose not to know a lot of things. But there are plenty of people who want to know, or can be educated to want to know.” He pauses for another nanosecond and then flashes that smile, looking as if he had just figured out the answer to this literally multibillion-dollar question about his effort and his company. “People used to not want to know their genes; now more and more people want to know,” he says. “I’m sure that this trend will continue.”

David Ewing Duncan is a life science journalist and author of Experimental Man, and the curator of Arc Fusion.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.