Google once pulled a videoconferencing system because it couldn’t recognize people of color, says former exec

Diane
Bryant, the former COO of Google Cloud, speaks on the campus of
UC Davis in August 2018.
Greg
Sandoval/Business Insider

• Former Google executive Diane Bryant said during a
  presentation on Tuesday that she had first-hand knowledge about
  video-conferencing technology that Google had to “pull back”
  because it couldn’t identify people of color. 
• Bryant suggested that the problem stemmed from a lack
  of diversity among the people who built the
  tech. 
• While some of Google’s top execs have warned about the
  problem of bias in creating algorithms,
  the company has created software in the past that was allegedly
  racially biased. 
• Bryant told the gathering of IT professionals:
  “Diversity is a fact. Inclusion is a choice.” 

Google once had to “pull back”
video-conferencing software for employees because of the
technology’s inability to accurately identify people of color,
Google cloud executive Diane Bryant said on Tuesday.

Bryant, Google cloud’s former COO
before departing in July after barely more than half a year, also
suggested that the problem with the technology was at least
partially caused by a lack of diversity among the people who
build it.    

Bryant made the comments during a
presentation at the University of California Davis before a
gathering of several hundred IT managers and faculty who work for
for the UC system during the annual

University of California Computing Services
Conference

.

Towards the end of her
presentation, Bryant began to discuss the need for diversity of
viewpoints within IT departments. She cited as an example
something that occurred at Google during her tenure there as an
example of what can go wrong when diversity is lacking.

“There’s been a lot of news
lately on diversity and AI,” Bryant told the crowd. “For
instance, if an algorithm is trained with incomplete data you’re
at risk of developing a bias, which can be disastrous when you’re
automating decisions, say, behind parole hearings or the approval
of loan applications.

“I sadly saw this firsthand at
Google,” she continued, when “IT’s initial videoconferencing
solution based on AI had to be pulled back” after initially
rolling out to employees. 

“The algorithms that detected the
faces in the room and then zoomed in automatically to center them
in the screen failed to accurately identify people of color,”
said Bryant. 

Through a spokeswoman, Google
declined to comment. Bryant departed the auditorium directly
after her speech without answering questions from the audience
and did not respond to questions sent later to her via LinkedIn.
 

Diversity is a hot-button topic
at Google

The issue of diversity is a
sensitive topic for Google, as it is at many other companies in
the tech sector. For a long time, Google has pledged itself to
diversifying its ranks, though progress has been slow. Google’s
latest diversity report disclosed that more than half of its
workers (53.1%) were white, a drop of 2.4 percentage points from
2017.

Dr.
Fei-Fei Li, Google’s chief AI scientist.
Greg Sandoval/Business Insider

That means the number of staff
members from other ethnic backgrounds increased slightly, but
most figures remained largely unchanged.

In 2015, software engineer Jacky
Alciné discovered that Google’s image recognition algorithms were
classifying black people as “gorillas.” The company apologized to
Alciné and promised to fix the problem.

Meanwhile, Diane Greene, the CEO of Google Cloud and Bryant’s
former boss, as well as Dr. Fei-Fei Li, the unit’s chief AI
scientist, have been vocal advocates for reducing bias in
algorithms and AI. 

As Bryant began to wrap up her
presentation, she received enthusiastic applause after saying
“diversity is a fact. Inclusion is a choice.”

She also warned tech companies
that they can’t blame under-represented groups for the diversity
problem, saying 
“y
ou
can’t put the burden of the marginalized class on the
marginalized class. That simply is illogical.”