THE NEWYORKER, SEPTEMBER 30, 2019 39
power users can make “fifteen to twenty
thousand U.S. dollars” on a shopping
holiday like Singles’ Day.
So far, TikTok has concentrated more
on expanding its user base than on offer-
ing opportunities for e-commerce. If
TikTok wants to keep growing, it will
need to attract more people who are no
longer in their teens, and it will need to
hold their attention. Many people are
not terribly interested in even the choic-
est memes the world has to offer; in Au-
gust, the Verge reported that a “signifi-
cant majority” of new TikTok users give
up on the app after thirty days. Bern
thinks that TikTok content will soon
become more mature, as has already hap-
pened with Douyin, which now contains
micro-vlogs, life-style content, business
advice, and videos from local police. Se-
lected users on Douyin can upload videos
as long as five minutes. Fictional mini-
dramas have begun to appear.
“This meme content, people will get
bored with it,” Bern said. “And compa-
nies are, like, ‘We cannot make this type
of content or we’ll damage our brands.’ ”
B
yteDance’s founder, Zhang Yiming,
was twenty-nine when he started
the company, in 2012. Zhang, who rarely
gives interviews, was raised in Fujian
Province, the son of a civil servant and
a nurse, and attended university in the
northern port city of Tianjin. He briefly
worked at Microsoft in China, and
bounced between startups for a while.
He then pitched Chinese investors on
the idea of a news-aggregation app that
would use machine learning to provide
people with whatever they wished to
read. The app, called Jinri Toutiao, was
launched within the year. Its name means
“today’s top headlines.” It’s a bit like
Reddit, if Reddit were guided by A.I.
rather than by the upvotes and down-
votes of its readers.
Like TikTok, Toutiao starts feeding
you content as soon as you open it, and
it adjusts the mix by tracking and an-
alyzing your scrolling behavior, the time
of day, and your location. It can deduce
how its users read while commuting,
and what they like to look at before
bed. It reportedly has around a hun-
dred and twenty million daily active
users, most of whom are under thirty.
On average, they read their tailored
feeds for more than an hour each day.
The app has a reputation for promot-
ing lowbrow clickbait.
In China, daily life has become even
more tech-driven than it is in the U.S.
People can pay for things by letting cam-
eras scan their faces; last year, a high
school in Hangzhou installed scanners
that recorded classrooms every thirty
seconds and classified students’ facial
expressions as neutral, happy, sad, angry,
upset, or surprised. The Chinese gov-
ernment has been assembling what it
calls the Social Credit System, a net-
work of overlapping assessments of cit-
izen trustworthiness, with opaque cal-
culations that integrate information from
public records and private databases.
The government has also set bench-
marks for progress in artificial-intelli-
gence development at five-year inter-
vals. Last year, Tianjin announced plans
to put sixteen billion dollars toward A.I.
funding; Shanghai announced a plan to
raise fifteen billion.
There are two principal approaches
to artificial intelligence. In symbolic A.I.,
humans give computers a set of elabo-
rate rules that guide them through a task.
This works well for things like chess, but
everyday tasks—identifying faces, inter-
preting language—tend to be governed
by human instinct as much as by rules.
And so another approach, known as neu-
ral networks, or machine learning, has
predominated in the past two decades
or so. Under this model, computers learn
by recognizing patterns in data and
continually adjusting until the desired
output—a correctly labelled face, a prop-
erly translated phrase—is consistently
achieved. In this sort of system, the quan-
tity of data is, broadly speaking, more
important than the sophistication of the
program interpreting it. The sheer num-
ber of users that Chinese companies have,
and the types of data that come from the
integration of tech with daily life, give
those companies a crucial advantage.
Chinese tech companies are often
partly funded by the government, and
they openly defer to its requests, turn-
ing over user messages and purchase
data, for instance. Tencent, which owns
WeChat, has a “Follow Our Party” sign
Kenneth Victor Young, Untitled (Abstract Composition), acrylic on canvas, 1972. Estimate $80,000 to $120,000.
Preview: October 3 & 4, 10-6; October 5, 12-5; October 7, 10-6; October 8, 10-12
104 East 25th St, New York, NY 10010 • tel 212 254 4710 • SWANNGALLERIES.COM
African-American Fine Art
October 8
Nigel Freeman • [email protected]
