General

The Future of AI Depends on High-School Girls

Lauren
Smiley, The Atlantic, May 23, 2018

Women
make up one-quarter of computer scientists. But in the field of artificial
intelligence those numbers are likely much lower.
Salinas
High School sophomore Stephanie Tena, 16, is passionate about coding and AI,
and she is working on a project that uses these skills to detect contamination
in water sources. Alison Yin

During
her freshman year, Stephanie Tena, a 16-year-old programmer, was searching the
internet for coding programs and came across a website for an organization
called AI4All, which runs an artificial-intelligence summer camp for
high-schoolers. On the site, a group of girls her age were gathered around an
autonomous car in front of the iconic arches of Stanford’s campus. “AI will
change the world,” the text read. “Who will change AI?”
Tena
thought maybe she could. She lives in a trailer park in California’s Central
Valley; her mom, a Mexican immigrant from Michoacán, picks strawberries in the
nearby fields. Tena has long black hair, a cheery, high-pitched voice, and an
unflappably professional bearing: She refers to other students as “my peers”
and her mentors as “notable professors,” and she has nailed the language of the
scientific method (“My hypothesis was proven incorrect”). She had been coding
for a couple of years, ever since attending a programming club at the local
community college when she was still in junior high. “I prefer science over
history,” she says. The summer after the 8th grade, she had flown to Los
Angeles for a coding boot camp run by the supermodel Karlie Kloss, (who had gotten
interested
in coding after taking a class herself), where she had
learned some programming languages and developed a website. She saved up for
more than a year, both her allowance and her pay from working at a bubble-tea
shop, and bought a new MacBook.
But even
as Tena applied, and was accepted, to the AI4All program on a full scholarship,
she knew little about artificial intelligence. Nor was she fully aware of
AI4All’s reason for existing: While women make up only a quarter of computer
scientists, their numbers appear to be even smaller in the AI field in
particular. While there are no government statistics on the percentage of women
in AI, women at the Annual Conference on Neural Information Processing Systems
(NIPS), the AI field’s top conference, made up only 17 percent of attendees
last year. The percentage of women has risen for the last four years, and NIPS is
considering changing the anatomically evocative conference title “in the
context of diversity issues,” according to
its website.
Artificial
intelligence is considered the major driver of what’s known as the fourth
industrial revolution (after the steam engine, electricity and mass production,
and the digital eras), with major tech companies like Google, Facebook, Amazon,
and Microsoft realigning around it. Algorithms are driving ever more real-world
decisions: helping doctors detect cancer; suggesting who should be released
from jail, interviewed for a job, or get a loan. While some high-profile
technologists, such as Elon Musk, have expressed
fears
that AI could become an existential threat to humanity, others
in the field have identified a
more immediate concern
: far from some God-like omniscience, AI can
be as biased and fallible as the humans who build it. AI has already made
embarrassing mistakes, like when Google Photos auto-tagged
pictures
of two black people as gorillas earlier this year because
the algorithm, it seems, wasn’t good at correctly labeling some non-white
faces. An Uber self-driving car killed a
pedestrian in Arizona
. While women were fighting for full sexual
agency in the real world, mostly male roboticists were creating AI-enhanced
mostly female sexbots. Bringing people like Stephanie Tena into artificial
intelligence is not simply important for the tech industry; in a world
increasingly driven by algorithms, it’s important for all of us.
Stephanie
Tena, 16, left, teaches middle schooler Bianca Castro, 12, coding and AI during
an after school club at Washington Irving Middle School. (Alison Yin)

In 2014,
Fei-Fei Li, an associate professor at Stanford, started brainstorming about how
to widen the pipeline with Olga Russakovsky, an advisee at Stanford, who was a
research assistant in the AI sub-field known as computer vision, which involves
analyzing images. Russakovsky now teaches at Princeton. In 2015, they held the
first summer camp on the Stanford campus (then called Stanford Artificial
Intelligence Laboratory’s Outreach Summer Program or SAILORS), a two-week
immersion program for girls in the basics of artificial intelligence.
Eventually, Li, Russakovsky, and Rick Sommer, who runs Stanford’s pre-college
summer programs, created a nonprofit with a board of advisors from academia,
nonprofits, and the industry, and funding from the likes of Melinda Gates and
Autodesk. They hired Tess Posner, who had led initiatives aimed at getting more
diversity into the digital economy, to be the CEO. “A lot of the perception of
AI is that it’s so hard to do and exclusive and you need to be a genius,” says
Posner. “And this program is helping to break that narrative and say this is
really for anyone and has applications for helping people.” This summer, AI4All
will take place on six campuses in the United States and Canada—some campuses
focusing on girls, and others on students of color and low-income students.

The first
summer, Li recalls, the students lit up when, in the middle of a dense
computer-science lecture, she described a project in which she and colleagues
had used computer vision to track hand hygiene practices among staff at a
Stanford hospital, so as to minimize the spread of infections. “Even though
they’re only 14 or 15 years old,” she says, “they’re passionate about making a
difference in using whatever they learned.” One alumna went on to work with Li
on a project using computer vision to assess a surgeon’s operating skill; the
resulting paper won best paper at a top AI conference. Another started a
hackathon for female high-school students.
Middle
schooler Karman Kaur learns coding and AI during an after school club run by
Stephanie Tena. (Alison Yin)
The
program was in its second summer, last June, when Tena’s sister dropped her off
at Stanford. Tena walked into a room of girls who hailed from Silicon Valley,
other areas of the country, and as far as Pakistan. She was one of very few
bilingual Spanish speakers, and the only one from the agricultural Central
Valley. She was a little intimidated when she learned other girls already knew
the Java and Python programming languages—“They had a little bit of a head
start”—but she plunged into the curriculum. She was inspired that Li had come to
the United States at the age of 16 without knowing English and had risen to
become a top AI researcher; along with teaching at Stanford, Li is now a Google
executive. Tena also loved the finer points of campus life, namely, the tater
tots at the university cafeteria. By the end of two weeks, Tena and a group of
teammates were programming a mini-autonomous car.
Tena left
the program with an awareness not only of AI but of the way in which the field
threatens to shut out people from communities like hers. “ If there’s a project
going on, and the majority of the people in the project are of one race or one
gender, you’re not really able to have the perspective of others, so it will be
tailored or targeted towards one specific group,” she says. When she returned
to Salinas, she started an AI club at the local public junior high. She walks
over each Monday afternoon to teach the basics of Java—which, while once so
foreign to her, has become the one she’s most skilled at—and about major
figures like Alan Turing, and the process for coding websites. The composition
of her club, as she calculates it, is more than half girls, and more than
three-quarters racial minorities. In her own public high school, she is one of
only a few females in the school’s inaugural advanced-placement
computer-science class. “I have an A,” she says.
Under the
supervision of her former AVID teacher Estefany Reyes, left, Stephanie Tena
leads an after school club to teach AI to middle schoolers. (Alison Yin)

Tena also
signed up for a new program AI4All debuted in the spring, called the Alumni
Research Fellowship Program, which paired students with industry mentors to
pursue real-world AI projects. Tena decided on a data-science project to map
the water toxicity in her area, given that the Central Valley’s water is
contaminated from fertilizer and manure run-off. She wanted to see how that
data correlates with the demographics of the surrounding towns—to test if, for
example, poor areas have dirtier water. One requirement of the program was that
Tena meet twice with a mentor in Oakland. Tena’s older sister, who lives near
Sacramento, would drive two-and-a-half hours south to Salinas to pick her up,
take her to the meetings, then chaperone her home again—eight hours of driving
each time.

On a
recent Saturday afternoon, the research fellowship’s 13 students gathered in
the Wozniak Lounge in Berkeley’s engineering building to present their
projects. A procession of teen girls, and one boy, approached a podium in the
front of the room, wearing blazers and dresses, as their parents filmed them on
their smartphones, and a photo collage of Steve Wozniak, an Apple founder,
looked on. One girl was working on a tool to detect wildfires using a drone.
Another had developed a speech bot that detected and responded to abusive
language. One team of three had created a triage system for paramedics to
respond to the most serious calls first; one team member explained that her
grandmother had died after suffering a stroke last year when the ambulance took
longer than 20 minutes to reach her. At the end, Russakovsky said, “I was
coming here expecting we kind of played around with some data, and, no, these
are, like, real research presentations!”
Speaking
after the ceremony, the group on the paramedic project giggled that they’d just
gotten their algorithm finally working the previous night. “I’m not going to
lie, I was also surprised by the quality of our presentation,” said Esther Cao,
a student from Palo Alto High School. “We didn’t have working code at all; it
was calling everything a diabetic emergency.” Their mentor, a female software
engineer from IBM’s Watson group, sat down next to the team and handed them
congratulatory laptop stickers.
Most of
the students were Asian Americans from high-achieving Silicon Valley high
schools with rigorous science and math curricula; a few told me their parents
were doctors and engineers. Some said they’d been in computer-science camps and
classes before AI4All, but that still didn’t mean they’d seen themselves going
into the computer-science field. “The only person I knew who codes is my
brother,” Trisha Sengupta, a student from a San Jose public high school, told
me. Until recently, she’d thought she’d probably go into medicine. Now, after
the project, Sengupta is pondering a double major in biomedical science and
computer science. “Now there’s like a repository on GitHub which has my name on
it,” she told me. “Okay, that’s cool.”
Programs
like AI4All are no doubt drawing dozens of girls, people of color, and
low-income teens into a field they otherwise wouldn’t have considered—which, in
combination with other coding-focused camps, such as Black Girls Code, may
start to improve diversity. Still, privately-run coding camps are not as
scalable or omnipresent as, say, getting a basic coding curriculum into all
junior high and high schools: The tech industry isn’t missing women and
minorities in the dozens, but the tens of thousands.
At the
end of the ceremony, Li stood and asked for feedback from the mentors. In the
spirit of programming, she is keen on rapidly iterating to improve the program.
“I see high- school students today being a lot more overscheduled than I was
when I was in high school,” said one Pandora data engineer, eliciting knowing
laughs from the parents. “It was difficult to make progress when there’s so
many other competing things going on.” The engineer wondered whether it would
be better run as a hackathon, and one mom said she wished the program had taken
place during the summer, with less going on. Another mentor chimed in that she
liked the multi-week format of the program: Since the students on her team
arrived with differing levels of AI knowledge, it gave her time to help get
them all up to speed.
Stephanie
Tena is working on a project to detect contamination in water sources as a
result of runoff from fields in her hometown. (Alison Yin)
One
student wasn’t there to chime in: Stephanie Tena. Her mentor had fallen absent
as the weeks wore on, and she didn’t have enough help to complete the project.
So she was glumly stuck in Salinas. Still, AI4All was working on transferring
her to another mentor, and she would present her water project to the
high-school girls at this year’s Stanford camp. She’ll already be a familiar
face: On the AI4All web page Tena first
landed on, her photo is now at the top.