SLOANREVIEW.MIT.EDU FALL 2019 MIT SLOAN MANAGEMENT REVIEW 5Sofar,artificialintelligencehasbeenwidely
viewed as a potential game changer for prob-
ing problems and making complex decisions.
It’s well known, for example, that, powered by
AI, DeepMind Technologies’ AlphaZero trained itself in 2017
to beat the best chess players in the world. It accomplished
this feat in less than 24 hours,
not by copying and tweaking
the strategies of chess masters
but by designing and executing
new moves that others found
counterintuitive.
But as AI finds its way into
more and more facets of mod-
ern life, there are a lot of
unknowns: How will it affect
the various ways people think
and interact with one another — and, more
worrisome, how will it affect civilization?
In a recent Atlantic article (“The
Metamorphosis,” August 2019), former
U.S. Secretary of State Henry Kissinger, for-
mer Google CEO Eric Schmidt, and Daniel
Huttenlocher, dean of MIT’s Schwarzman
College of Computing, say “the changes [AI]
will impose on human life will be transforma-
tive,” but not always in a good way.
For example, they note that historically,
nuclear deterrence has been based on “the rationality of par-
ties” and “the likelihood of retaliation deters attack.” But
new weapons and strategies designed with AI, they worry,
could turn things upside down (as AlphaZero did for chess),
overwhelming existing models built on transparency and
creating new levels of risk. They add that smarter and more
sophisticated digital assistants and devices have minuses
as well as pluses: fewer information choices for users, for
instance, and less access to challenging ideas.
Although the three authors differ on how optimistic they
are about AI, they all agree that, as AI becomes more ubiqui-
tous, we need clear systems for keeping it in check. Among
them: “requiring human involvement in high-stakes pattern
recognition,” like reading X-rays; developing simulations
to teach AI to identify ethical issues; and “auditing” AI
decisions that touch on human values.High-Profile Jitters Over AI
Why Weather Will Always Be Uncertain
We’re getting better and better at using data to pre-
dict what’s likely to happen next, and nowhere is
this truer than with weather. Throughout most of history,
people really didn’t know whether it would rain or shine until
the day it happened. But with the introduction of the telegraph in the United
States in the mid-19th century, residents in one town could finally learn if it
was raining in another town 300 miles away. Writing in TheNew Yorker
(“Why Weather Forecasting Keeps Getting Better,” June 24, 2019) and
drawing on insights from journalist Andrew Blum’s new book, The Weather
Machine: A Journey Inside the Forecast, Hannah Fry, a mathematician at
University College London, describes how weather forecasting has
evolved — and why 100% accuracy will never be attainable.
Weather, Fry writes, “is an intricate mathematical tapestry that is far
too intertwined to unpick by hand.”
A century ago, she notes, a British mathematician dreamed of a “forecast
factory” where thousands of people crank out calculations related to weather
conditions in different parts of the world. Fry reports that in a few years, with
supercomputers, the European Centre for Medium-Range Weather Fore-
casts expects “to be able to detect high-impact events two weeks into
the future.” However, she says the effects of global warming will make
extremes more likely and long-range predictions more challenging.
— Bruce PosnerRebooting the Supermarket
For supermarkets, the race is on to develop business models that
are able to compete with the likes of Amazon and, not incidentally,
widen their paper-thin profit margins. One of the most aggressive efforts to
date is coming from Cincinnati-based Kroger, which operates some 2,
supermarkets in 35 states. Eager to leverage the latest warehouse automa-
tion and robotics technology, Kroger entered into an exclusive partnership
last year with Ocado Group, a British grocery retailer that has no stores,
delivering orders to homes directly from its automated warehouses. In
the next three years, Kroger plans to build 20 automated warehouses.
As a recent Bloomberg article by Sarah Halzack
(“Kroger Goes Full Robot to Take On Amazon,”
July 22, 2019) points out, getting a grip on the online
grocery business, where at least some items are
perishable or frozen, is considerably more compli-
cated than other types of online retail. At one of
Ocado’s grocery fulfillment centers near London,
for example, some 1,000 robots scurry around a
mechanical grid moving bins of groceries. The better
the automated system works and the more orders
it processes, the brighter the prospects of higher
profits. But whether these changes can occur fast
enough, says Halzack, is an open question.[ELSEWHERE ]