Page 103 - Fortune-November 01, 2018

+life ar tifi c i al intelli gence + heal th and wellne s s

7%

PERCENTAGE OF HR
EMPLOYEES WHO THINK
“A ROBOT COULD DO
THEIR JOB,” ACCORDING
TO A STUDY BY IDEALL

MAKING HEALTH CARE

HUMAN AGAIN OUTSMARTING
YOUR DOCTOR
b y eri c to p o l
GROWING THE CURRENT U.S. HEALTH CARE PICTURE INJUST THE PAST

YOUR NEXT is pretty bleak: more than 12 million seri- few years, there
ous diagnostic errors each year, a third of
have emerged cred-
MEAL the $3.6 trillion spent attributed to waste, ible if still-in-the-
works A.I.-powered
reduction in life expectancy for what will
be three years in a row (which is unpre-
technologies that
cedented), and peak levels of physician can read radiology
burnout, depression, and suicide. That’s scans (like Imagen),
ON THE SURFACE, FARMING SEEMS LIKE all happening at a time when there is more identify tumors and
a simple endeavor: Pop seeds in the medical data per individual than ever, imag- track the spread of
ground, water, harvest, repeat. But in ined with wearable sensor physiology, scan cancer (Arterys),
reality, how food is grown is built on a anatomy (above), DNA sequencing, gut mi- detect eye condi-
series of intricate equations. “A lot of tions using retinal
the data we deal with in agriculture is crobiome biology, just to name a few layers. imaging (Google’s
very complex,” says Nate Storey, the Enter deep-learning A.I., with neural net- DeepMind), flag
cofounder and chief science officer of works that will impact every type of clinician, dangerously abnor-
Plenty, an indoor vertical-farming en- from helping to accurately read scans, slides, mal potassium lev-
terprise. Environmental factors (airflow, skin lesions, eyegrounds, and more, to health els via a “bloodless
carbon dioxide, light, and humidity, to systems, promoting the use of remote moni- blood test” (Mayo
name a few), the genetics of the plant, toring that ultimately obviates the need for Clinic Ventures
and the things we do to it, like fertil- regular hospital rooms, and at the consumer and AliveCor), and
izing and watering, are all interacting level, by providing a virtual medical coach to otherwise assist
variables. Now Plenty and a number better manage or even prevent diseases. It’s with the tricky busi-
of other startups are using A.I. to help still early in the integration of A.I. into medical ness of diagnosing,
manage the complex decisions that practice, with far more hype than validation. or even predicting,
go into farming. For example, Plenty But it’s our best shot to deal with all of the disease. Historically,
and its indoor-ag rivals Bowery and formidable challenges: to use the wealth of diagnostic error
Gotham Greens are all building systems data to reduce errors and waste, and the gift rates have been
that collect and analyze data sets of of time to markedly improve the clinician- put at 5% to 20%,
images that can help identify whether a patient relationship. though the rate is
plant has an issue, like nitrogen or iron higher for some
deficiency or a pest problem, through conditions, while the
machine learning and then preemp- Eric Topol, MD, is the founder and health care system
tively treat it. “The software can learn director of the Scripps Research is strained by doctor BR A IN: SPL /SCIENCE SOURCE
what the problems are and do it in an Translational Institute and author of shortage and burn-
automated fashion at a large scale that the forthcoming book Deep Medicine. out—some things
we couldn’t individually do,” Storey says. A.I. may be able to
—Beth Kowitt treat. —E.F.

108
FO R T U N E. CO M // N O V. 1 . 1 8 FEEDBACK LETTERS@FORTUNE.COM