Science - USA (2022-01-14)

SCIENCE science.org 14 JANUARY 2022 • VOL 375 ISSUE 6577 129

downplay the degree of deaths.” (Three

Indian government sources that Science

contacted did not reply to requests for

comments on the new study.)

The new estimates for India come as

little surprise to Laxminarayan. “My start-

ing point is that unless you can tell me why

India is different, I’m going to assume that

India is the same as any other country,”

he says. “I don’t believe in exceptionalism

of any kind unless it’s well justified.” His

team last month published a study in The

Lancet that focused on the Indian district

of Chennai and concluded that reported

deaths “greatly underestimated pandemic-

associated mortality.”

Virologist Shahid Jameel of Ashoka Uni-

versity says the countrywide estimates by

Jha’s team are also “in broad agreement”

with two other independent studies that

examined a similar time frame. “India paid

a heavy price for not having good real-time

data on deaths, especially during the first

wave. That led to complacency and a ter-

rible toll in the second wave,” Jameel says.

The work nicely triangulates data from

different sources, each of which has its own

limitations, says Samira Asma, a WHO as-

sistant director general who works on data

and analysis. “The study design is robust,”

Asma says. “Countries can learn from this

approach to ... produce country-specific

estimates.” WHO is now updating its esti-

mates of excess deaths caused by COVID-

and plans to release them soon, she says.

A worldwide comparison of all-cause

mortality before and during the pandemic,

published in eLife 6 months ago, suggests

undercounting is widespread. Russia had

4.5 times more deaths than normal, far

beyond its official COVID-19 tally, and

the trend has continued, the researchers

recently tweeted. Tajikistan, Nicaragua,

Uzbekistan, Belarus, and Egypt also had

profound disconnects.

Omicron has begun to run riot in In-

dia, and Jha warns the country should not

count on it causing milder disease than ear-

lier variants, as data from other countries

suggest it might. “I would be really careful

about those assumptions because they’re

based on selected populations that you

can’t take from South Africa or the U.K. or

Canada to India,” he says.

And Jha is wary of “wishful thinking”

that the prior high levels of infections in

India and wide-scale vaccination will cre-

ate population level immunity and reduce

severe disease from the variant. “We just

don’t know enough about how these differ-

ent variants behave in immunized popula-

tions,” he says.

Spoken like a researcher who doesn’t

want to get it wrong again. j

NEWS

Software that identifies unique styles poses privacy risks

COMPUTER SCIENCE

T

hink your bishop’s opening, queen’s

gambit, and pawn play are unique? A

new artificial intelligence (AI) algo-

rithm has got your chess style pegged.

AI software can already identify

people by their voices or handwriting.

Now, an AI has shown it can tag people based

on their chess-playing behavior, an advance

in the field of “stylometrics” that could help

computers be better chess teachers or more

humanlike in their game play. Alarmingly,

the system could also be used to help iden-

tify and track people who think their online

behavior is anonymous.

“Privacy threats are growing rapidly,”

says Alexandra Wood, a lawyer at the Berk-

man Klein Center for Internet & Society at

Harvard University. She says studies like

this one, when conducted

responsibly, are useful be-

cause they “shed light on

a significant mode of pri-

vacy loss.”

Chess-playing software,

such as Deep Blue and

AlphaZero, has long been

superhuman. But Ashton

Anderson, a computer sci-

entist at the University of

Toronto and principal in-

vestigator of the new proj-

ect, says the chess engines

play almost an “alien style”

that isn’t very instructive for those seeking to

learn or improve their skills. They’d do bet-

ter to tailor their advice to individual play-

ers. But first they’d need to capture a player’s

unique form.

To design and train their AI, the re-

searchers tapped an ample resource: more

than 50 million human games played

on the Lichess website. They collected

games by players who had played at least

1000 times and sampled sequences of up

to 32 moves from those games. They coded

each move and fed them into a neural net-

work that represented each game as a point

in multidimensional space, so that each

player’s games formed a cluster of points.

The network was trained to maximize the

density of each player’s cluster and the dis-

tance between those of different players.

That required the system to recognize what

was distinctive about each player’s style.

The researchers tested the system by

seeing how well it distinguished one

player from another. They gave the sys-

tem 100 games from each of about

3000 known players, and 100 fresh games

from a mystery player. To make the task

harder, they hid the first 15 moves of each

game. The system looked for the best match

and identified the mystery player 86% of

the time, the researchers reported last

month at the Conference on Neural Infor-

mation Processing Systems (NeurIPS). “We

didn’t quite believe the results,” says Reid

McIlroy-Young, a student in Anderson’s lab

and the paper’s primary author. A non-AI

method was only 28% accurate.

“The work is really cool,” says Noam

Brown, a research scientist at Meta (the

parent company of Face-

book) who has developed

superhuman poker bots.

He looks forward to chess

bots that mimic Magnus

Carlsen, the reigning

world champion, and says

style-aware AI could trans-

form other computer in-

teractions. “There’s a lot of

interest in chatbots, where

you can have a chatbot

that would speak in the

style of Albert Einstein or

something,” he says.

The researchers are aware of the privacy

risks posed by the system, which could be

used to unmask anonymous chess players

online. With tweaks, McIlroy-Young says,

it could do the same for poker. And in the-

ory, they say, given the right data sets, such

systems could identify people based on the

quirks of their driving or the timing and loca-

tion of their cellphone use.

NeurIPS organizers found the study tech-

nically impressive but ethically fraught, and

accepted it on the condition that the re-

searchers elaborate on the privacy risks. (It

could be “of interest to marketers [and] law

enforcement,” one reviewer commented.)

Anderson says they’ve decided, for now, not

to release the software code. j

Matthew Hutson is a journalist in New York City.

By Matthew Hutson

Artificial intelligence unmasks

anonymous chess players

PHOTO: XIJIAN/ISTOCK.COM