Time - USA (2021-03-01)

KHAN: BRANDON THIBODEAUX—THE NEW YORK TIMES/REDUX; JACOBY: ILLUSTRATION BY ALEXIS FRANKLIN FOR TIME; JUMPER: BEN PETER CATCHPOLE

Clementine

Jacoby

29 • Decarceration

through data

More than 2 million people
are imprisoned in the U.S.,
among them hundreds of thou-
sands who experts say don’t
pose a public-safety threat and
could be released. One prob-
lem: the data that could trig-
ger those releases get back-
logged, because they’re often
spread out among different de-
partments. That’s why in 2019,
Clementine Jacoby, a software
engineer, launched Recidiviz, a
nonprofi t that has worked with
more than 30 states to log into
one system key data points—
such as whether an incarcerated
person has served most of their
sentence or has shown progress
by completing a treatment pro-
gram, or more recently, how well
equipped a correction facility is
to handle a COVID-19 outbreak.
It then uses an algorithm to rec-
ommend certain prisoners for
release. “Our hope is that the
people who are succeeding get
off early,” Jacoby says, “and
that frees up attention for of-
fi cers to spend time with the
people who actually need it.”
Of course, no algorithm is per-
fect, and algorithms alone won’t
solve the issues of the criminal-
justice system. But so far, Re-
cidiviz has seen early signs of
success. To date, the nonprofi t
has helped identify as appropri-
ate for release nearly 44,000
inmates in 34 states, including
North Dakota, which last spring
saw its prison population drop
by 20%. —Patrick Lucas Austin

JOHN

JUMPER

36 • Predicting

protein shapes

Being able to predict the shape

of proteins using amino-acid

sequences could one day

inform the development of

new treatments for diseases,

but learning the shape of

even a single protein costs

about $100,000 and takes

a full year of a researcher’s

time. Thanks to John Jumper,

lead researcher for U.K.-

based company DeepMind’s

AlphaFold artifi cial-

intelligence program, it’s now

possible to cut that year of

work to a single day. AlphaFold

leverages neural networks—

or algorithmic systems that

learn by example—to more

quickly compute the shape a

protein folds into. The result: a

technological leap forward that

demonstrates the power of AI

to solve complex problems and

a better ability to use proteins

to understand how diseases

work, develop new drugs and

more. —Jeff rey Kluger

LAYERS

OF WHEAT

10

IN EVERY BITE