►
Description
Presented by Terri Burns, Twitter.
Last year, NPR did a story answering the question, can computers be racist? (Yes.) Not soon after, Microsoft launched an AI chatbot experiment, called Tay, which shut down when the software began spewing hateful speech online. One of the known fears of AI and machine learning is the notion of algorithmic bias, which can create or indirectly allow machines to learn prejudiced behavior. In this talk, we will explore what it really means to “teach” a computer to have prejudices, and what can happen product managers, designers, and engineers are not cognizant of their own biases. Appropriate for both technical and non-technical audiences. What Attendees Can Expect to Learn: Examples of “bad” computing, how personal beliefs and prejudice can manifest in code and product development, and ways to avoid personal bias in algorithms and products.
About GitHub Universe:
GitHub Universe is a two-day conference dedicated to the creativity and curiosity of the largest software community in the world. Sessions cover topics from team culture to open source software across industries and technologies.
For more information on GitHub Universe, check the website:
https://githubuniverse.com
A
All
right
so,
tough
morning,
okay,
that's
okay!
I
had
one
of
those
things
happen
to
me
this
morning,
when
you're
like
inching
closer
to
the
weekend
and
so
I
look
up
and
I
was
like.
Yes,
it's
Friday
and
then
I
quickly
realized
that
it
is
in
fact
not
Friday,
so
I
feel
you
I
hear
you
no
worries
but
we're
at
github
universe,
which
is
pretty
awesome.
The
decor
is
fantastic.
A
The
speakers
have
been
amazing
and
I
can
sort
of
just
like
sense,
all
of
the
smartness
kind
of
emanating
from
the
audience
right
now,
because
you
all
seem
like
very
intelligent,
experienced
people,
so
I
think
that
we
were
in
some
exciting
times
right
now
and
I.
Think
maybe
I
should
try
asking
that
question
one
more
time.
So
how
is
everyone
doing
today?
A
Okay,
beautiful
beautiful?
That
was
much
better.
So
welcome
to
my
talk,
bad
people,
bad
computers,
I'll
be
talking
about
algorithmic
bias,
keys
in
AI
and
machine
learning,
as
well
as
just
product
development
in
general,
no
matter
what
it
is
that
you
do,
whether
you
are
a
software
engineer
or
a
product
manager
or
a
designer.
There
are
decisions
that
you
can
make
as
a
technologist
that
go
into
the
products
that
you
build,
and
sometimes
these
decisions
can
have
really
really
amazing
and
fantastic
ramifications,
and
sometimes
not
so
much.
A
My
name
is
Terry
burns,
also
known
as
TC
burning
on
the
interwebs,
but
I
wanted
to
kick
off
today
by
giving
you
guys
a
little
story,
mostly
because
the
alternative
is
me
like
singing
or
dancing
and
I
figured
that
that's
probably
not
the
time
to
do
that
right
now,
although
I
am
a
fantastic
dancer,
so
I'm
going
to
talk
about
my
friend
Jackie
to
get
started,
and
so
Jackie
is
a
really
good
friend
of
mine.
He
and
I
became
friends
a
couple
of
years
ago
via
the
interwebs.
So
it
goes
in
this
day
and
age.
A
Specifically,
we
became
friends
on
Twitter,
so
Jackie
and
I
began
following
each
other
on
Twitter
a
couple
of
years
ago,
and
basically,
we
just
kind
of
like
caught
up
with
each
other's
lives
and
were
able
to
stay
up-to-date
on
what
we
were
up
to
for
a
long
time.
He
since
calm
down
a
little
bit
but
jackie,
was
super
super
stream
of
conscious
on
the
Internet
and
so
anytime,
you
sort
of
like
felt
or
thought
something
he
would
kind
of
tweet
about
it.
A
So
there
was
a
period
of
time
where
I
felt
like
I
was
in
Jackie's.
Brain
parallel
to
this
and
related
to
this
story
is
a
couple
of
years
ago,
a
piece
of
software
came
out
that
I
started
playing
around
a
little
bit
with
really
beautiful,
flexible
and
awesome
piece
of
software,
and
that
is
Google
Photos.
So
here
is
a
fusion
screenshot
of
just
a
pressed
piece
that
came
out
when
feudal
when
google
photos
was
released.
Basically
the
way
Google
photos
works.
A
It's
a
really
awesome
piece
of
software,
it's
in
an
image
recognition,
and
so
it's
basically
able
to
categorize
the
content
of
what
exists
inside
different
types
of
photos.
So
if
you
go
on
an
amazing
vacation
to
Barbados-
and
you
come
back
home
and
you're
hanging
out
with
your
mom,
you
might
say:
Oh
mom
I,
you
know
had
the
most
fantastic
vacation.
I
took
a
bunch
of
pictures
of
flowers
instead
of
having
to
sift
through
hundreds
of
photos
of
selfies
on
the
beach
you
can
like
type
in
the
keyword,
flowers
and
up
pops.
A
A
So
Jackie
had
been
using
this
google
photo
software
and
he
realized
one
day
that
the
Google
Photos
software
was
categorizing
photographs
of
him
and
all
of
his
friends
who
happened
to
all
be
black
in
the
photos
that
he
uploaded
as
guerrillas.
And
then
you
can
see
to
the
right.
This
is
just
a
screenshot
of
a
press
piece,
but
there
was
a
little
bit
of
press
around
this
because
Allah.
A
Obviously,
a
lot
of
people
were
super
upset
by
this,
so
the
ending
of
the
story
is
actually
a
pretty
good
ending,
so
it
so
happened
that
one
of
the
lead
engineers
from
Google
photos
happen
to
see
Jackie's,
tweet
and
responded
on,
and
the
team
got
working
on
it
very
very
quickly.
Super
super
smart
group
of
people.
They
were
able
to
deploy
a
fix
relatively
quickly.
A
What
ended
up
happening
was
they
were
able
to
kind
of
update
the
algorithm
a
little
bit
and
it
was
able
to
mitigate
the
issue
from
happening,
but
not
entirely,
and
so
they
actually
ended
up
removing
the
term
guerrillas
from
their
keywords
entirely.
So
again,
this
is
a
couple
of
years
ago,
so
I'm
not
exact
sure
what
the
implementation
looks
like
now,
but
at
the
time
it
was
a
quick
fix.
A
The
team
was
super
apologetic
about
it,
but
ultimately
it
was
something
that
was
really
jarring
for
a
friend
of
mine
and
it
was
really
jarring
for
me
to
have
to
see
that
and
have
to.
You
know,
be
a
part
of
that
experience
as
well.
So
that's
a
little
bit
about
what
my
talk
today
is
about.
It's
about
algorithmic
biases,
it's
about
programming,
biases,
it's
about
design,
biases
and
sometimes
even
the
smartest
smartest
people
in
the
world
who
are
working
on
really
fantastic
technologies
that
are
scaling
out
to
millions
of
people.
A
Sometimes
there
are
cracks
and
there
are
holes
in
the
software
and
that
ends
up
manifesting
and
negative
experiences.
Real
experiences
for
people
all
around
the
world
so
before
I
jump
jump
into
that
a
little
bit
and
talk
about
that.
What
that's
like
and
how
to
mitigate
this
from
happening
and
some
of
the
things
that
you
can
think
about.
As
a
technologist,
I
figure
I'll
tell
you
guys
a
little
bit
about
myself
because
you
guys
are
probably
like
who
is
this
girl
if
you're
talking
about
all
these
things?
A
So
this
is
me
photo
from
back
in
the
day
it's
Thursday,
so
I
figured
I'd.
Do
a
throwback
Thursday
up
here,
hashtag
TVT,
as
I
mentioned
before
my
name
is
teri
burns
and
I'm
originally
from
Long
Beach
California,
lbc,
Snoop
Dogg
always
born
and
raised
California
girl,
absolutely
loved
California
I
spent
a
couple
of
years
outside
of
the
state,
though,
because
I
studied
computer
science
at
NYU,
and
so
my
foray
into
technology
and
into
the
world
of
programming
was
a
little
bit
of
a
roundabout
way.
A
Company
and
so
I
went,
and
we
had
this
tour
of
the
company
and
it
was
totally
like
you
know.
Technology
is
going
to
be
the
future
of
humanity
and
we
are
changing
the
world
and
humanity
rests
on
one's
ability
to
write.
As
many
lines
of
code
as
possible,
which
now
that
I've
been
in
the
industry
for
a
few
years,
I
have
a
little
bit
more
of
a
nuanced
opinion
on,
but
at
the
time
I
was
totally
totally
sipping.
A
The
kool-aid
and
I
was
like
this
seems
really
cool
and
so
I
went
back
home
after
that
tour
and
basically
spent
the
rest
of
the
summer
sitting
around
on
my
laptop
playing
around
and
teaching
myself
how
to
program,
and
then
in
the
fall.
I
took
my
first
official
computer
science
class
and
then
three
years
later,
I
ended
up
graduating
with
a
degree
when
I
worked
it
like
that
it
sounds
a
lot
smoother.
There
was
definitely
a
bunch
of
ups
and
downs
along
the
road,
but
now
here
I
am
I've.
A
There's
a
reason
why
I
decided
to
add
this
photograph
here,
which
is
a
throwback
to
back
in
the
day.
So
part
of
this
talk,
as
I
explained
before,
is
talking
a
lot
about
thinking
about
how
technologies
work
for
people
all
over
the
world
in
good
ways
and
in
bad
ways.
So
I
have
a
question
here
just
for
the
audience
in
general,
by
a
show
of
hands
who
in
here
has
ever
or
is
interested
in
potentially
writing
a
piece
of
code
that
can
be
touched
by
people
all
over
the
world.
A
I
thank
God,
okay,
good
I'm,
talking
to
the
right
audience
here,
yeah,
that's
amazing
and
I'm
going
to
go
ahead
and
assume
that
all
of
you
are
super
noble
and
hard-working
good
people
that
are
hoping
to
build
technologies
that
can
impact
people's
lives
in
really
positive
ways.
And
if
that
is
the
case,
I
very
much
do
hope
that
you
are
successful
in
this
pursuit.
A
But
in
today's
day
and
age
we
are
building
technologies
at
a
completely
unprecedented
scale,
and
so,
when
you
think
about
the
biggest
consumer
technology
company,
even
when
you
think
about
enterprise
technologies,
directly
or
indirectly,
a
lot
of
the
platforms
that
we're
building
are
touching
people
at
a
massive
scale
that
we've
never
seen
before.
And
even
if
you
work
at
a
small,
tiny,
startup
and
you're
building
something
that
hasn't
necessarily
touched.
A
lot
of
people's
lives.
A
Even
just
the
mental
sort
of
model
of
being
able
to
one
day
touch
people
all
over
the
world
is
something
that's
completely
unprecedented
and
the
issue
with
that
is
that,
as
I'm
sure
many
of
you
know
there
are
over
seven
billion
people
on
the
planet,
which
means
there
are
over
seven
billion
different
lives
that
you
are
trying
to
touch.
Those
are
different
perspectives.
Those
are
different
ways
of
interacting
with
technology.
A
Those
are
different
ways
of
just
viewing
the
world
in
general,
and
so,
when
you're
writing
a
piece
of
software,
you
really
have
to
think
very
carefully.
How
can
I
write
a
piece
of
software
or
how
can
I
create
a
product
or
how
can
I
design
a
product?
That's
durable
enough,
but
also
flexible
enough
to
be
able
to
be
useful
to
all
different
types
of
people
all
around
the
world,
and
that's
a
really
challenging
thing
to
think
about.
A
The
way
I'm
going
to
sort
of
walk
through
this
is
by
throwing
it
back
with
a
very
simple
program
that
you
might
have
written
back
in
the
day
when
you
were
first
learning
how
to
program.
So
as
I
mentioned
for
me,
that
was
when
I
was
the
summer
after
my
freshman
year,
when
I
was
just
learning
how
to
program
for
some
of
you
that
might
have
been
20
years
ago.
Some
of
you
may
be
last
week.
It
could
have
been
in
college
or
in
school
or
at
a
boot
camp
or
late
at
night.
A
In
your
basement,
when
you
were
teaching
yourself
how
to
program
whatever
that
looks
like
for
you
and
whenever
that
looks
like
for
you,
I
am
encouraging
you
to
sort
of
transport
yourself
back
to
that
time
and
think
about
the
types
of
programs
that
you
were
building
and
the
things
that
you
are
working
on
to
help.
You
learn
the
fundamentals
of
programming
and
I'm,
going
to
show
you
how
even
then
you
were
thinking
about
seven
billion
different
people.
So
what
we're
gonna
do
right
now
is
we're
going
to
write
a
guessing
game.
A
So
remember
these
are
back
in
the
early
days
of
programming
or
writing
a
very
sort
of
simple
program.
In
this
case
the
program
is
going
to
be
guess
the
number
one
through
ten.
If
the
user
guess
is
that
number
correctly,
they
win
a
million
bucks,
quick
pies.
Do
we
have
any
volunteers
in
the
audience
for
anyone
who
wants
to
contribute
that
prize
money,
all
right,
we're
going
to
circle
back
on
that
at
the
end,
no
worries,
okay,
and
if
you
don't
win,
then
you
then
you
lose.
A
Then
you
you,
don't
guess
the
number
correctly
so
I'm
gonna
write
up
a
quick
Python
program
here,
so
you
sort
of
generate
a
random
integer.
You
prompt
the
user
to
guess
the
number
one
through
ten.
If
you're
right,
you
want
a
million
bucks.
Oh
my
goodness,
you
want
a
million
bucks,
ask
someone
in
the
audience
for
it
again
we'll
follow
up
with
that.
Otherwise
super
awkward.
A
So
if
you're
like
me,
you
you
might
be
assigned
to
write
a
program
like
this
and
you
go
ahead
and
you
write
it
and
you're
like
oh,
my
goodness,
I
I
know
how
to
write
this
program.
I'm
a
10x
engineer.
This
is
amazing,
I'm
the
smartest
person
in
the
world
and
then
you
might
go
and
like
have
someone
test
it
out.
So
for
me,
oftentimes
I
would
like
go
back
to
my
mom
and
be
like
hey
mom
I
wrote
this
really
cool
program.
Do
you
want
to
play
with
it?
A
It's
just
a
super
simple,
guessing
game.
If
you
want
to
try
it
out,
and
so
my
mom
would
be
super
supportive
and
she
would
go
ahead
and
she
would
play
the
game,
and
so
we
would
run
the
program
got
the
number
one
through
ten.
If
you're
right,
you
win
a
million
bucks
and
what
does
my
mother
enter
twenty?
A
So
you
learn
this
lesson
and
you're,
like
you
know
what
it's
fine,
mom
I'm
gonna
go
back
to
the
drawing
board.
You
add
a
quick
clause
and
you
make
sure
that
the
input
is
between
1
and
10
inclusive.
Otherwise,
you
sort
of
reap
romp
to
the
user
to
re-enter
it.
You're
like
all
right
back
to
being
a
10x
engineer.
Life
is
great
mom,
let's
run
this
program,
she
runs
the
program,
guess
the
number
1
through
10.
A
If
you
get
it
right,
you
win
a
million
bucks
and
what
does
she
enter
two
as
a
string
and
another
error
is
thrown
so
again.
This
goes
back
to
the
fundamentals.
It's
totally
reasonable.
For
someone
to
like
look
up
this
program
and
say:
oh
I
have
to
enter
a
number
okay.
Well,
I'm
gonna
enter
it
in
the
way
that
I
think
is
reasonable
to
enter
it
and
that's
totally
fine.
A
There's
lots
of
potential
things
that
can
go
wrong
with
even
a
very
elementary
program
that
you
might
have
written
back
in
the
day
when
you
were
first
learning
how
to
write
programming
or
how
to
program.
What's
really
interesting
about
this
is
when
you
think
about
even
the
simplest
of
programs.
Forget
super
complex
consumer
software.
That's
being
you
know,
touched
by
lives
all
over
the
world.
If
you
were
to
write
a
simple
guessing
game
program
and
your
goal
was
ultimately
to
touch
seven
billion
people
all
over
the
world.
All
of
a
sudden.
A
What
is
a
super
simple
sort
of
piece
of
software
and
a
super
simple
piece
of
logic,
all
of
a
sudden
becomes
extremely
complicated.
For
example,
you
might
want
to
consider
the
use
case.
What
if
someone
physically
can't
see
the
screen
out
of
seven
billion
people
all
over
the
world?
There
are
certainly
a
good
number
of
people
who
actually
physically
have
a
limitation
where
they
can't
look
at
the
screen,
but
they
might
want
to
be
able
to
interact
with
your
program
in
some
way.
What?
If,
for
example,
someone
can
only
count
to
five?
A
Maybe
this
person
is,
is
really
young
or
has
certain
mental
limitations
and
would
absolutely
love
to
play
this
game
and
have
a
really
good
time
playing
this
game.
But
the
range
up
to
ten
is
just
not
feasible
or
one
if
someone
doesn't
know
how
to
interact
with
the
computer
at
all
and
this
one's
my
favorite,
because
when
I
would
home,
for
you
know
like
Thanksgiving
and
Christmas,
and
all
of
that
stuff
I.
Remember
writing
these
programs.
A
When
I
was
in
school
I'm
going
to
my
grandmother
and
being
like
grandma
I'm
studying
computer
science,
now
I
know
you
don't
know
what
that
means,
but
check
out
this
really
really
cool
program
that
I
wrote.
All
you
have
to
do
is
press
these
buttons
and,
like
something
cool,
will
pop
up
on
the
screen.
My
grandmother,
you
know
she
would
like
look
with
her
glasses
and
she'd,
be
like
keep
it
up
good
job
and
have
no
idea
how
to
even
add
any
sort
of
input
to
a
program.
A
My
grandmother
is
part
of
that
7
billion
people
around
the
world,
and
so
it's
tough
to
think
about
it's
definitely.
You
know
this
example
is
simplified
and
also
overdramatic
at
the
same
time,
just
for
the
sake
of
example
here.
But
if
you
are
someone
who
is
building
software,
you
have
to
constantly
be
thinking,
as
you
are
sort
of
getting
your
next
adopters.
A
As
you
are
trying
to
scale
out
the
number
of
users
that
you're
having
what
types
of
problems
might
users
have
what
types
of
limitations
might
people
have
and
how
can
I
build
software
in
a
way
that
is
inclusive
to
all
of
these
people?
Now
I,
don't
know
about
all
of
you,
but
for
me,
when
I
start
thinking
about
all
of
these
edge
cases
and
when
I'm
like
sitting
down
working
on
a
product
dock
and
trying
to
understand
whether
or
not
something
that
I'm
building
is
super
inclusive,
it
is
exhausting.
A
But
I
think
that
as
an
industry,
it's
important
for
us
to
recognize
when
things
go
wrong
to
remember
when
things
go
wrong
and
be
honest
with
ourselves,
so
that
we
can
hold
ourselves
accountable
and
build
even
better
products
in
the
future,
so
I'm
gonna
do
right
now.
The
first
example
is
about
pay.
So
a
couple
of
years
ago,
Microsoft
released
an
artificially
intelligent
chat
bot.
This
is
like
a
super
cool
artificially
intelligent
machine
learning
based
chat
bot.
Her
name
was
tazed.
She
was
meant
to
be
like
a
cool
young
teen
and
you
could
chat.
A
Latavia
group
me
or
cake
or
Twitter
or
whatever,
and
basically
how
she
worked
at
a
very
fundamental
level
was
the
more
you
spoke
to
Tay
the
more
she
learned
and
the
more
she
learned,
the
more
sort
of
thorough
conversations
that
you
could
have
with
her
so
intake.
This
is
a
screenshot
of
just
a
TechCrunch
article
that
came
out
when
teh
first
released
and
then
within
a
couple
of
hours.
All
of
the
headlines
quickly
changed.
A
So
what
ended
up
happening
was
is
that
there
were
a
lot
of
people
in
this
case.
These
are
intentionally
bad
actors,
but
people
would
have
conversations
with
Tay
and
say
extremely
offensive
and
inappropriate
and
racist
and
really
just
exclusive
language
and
tailored
all
of
that,
and
soon
enough
started
spewing
out
that
language
herself
and
so
Microsoft
very,
very
quickly,
sort
of
had
to
issue
an
apology
and
Shayla
said
shut
down
very
quickly.
A
Another
example
is
YouTube,
so
I'm
sure
every
single
person
in
this
room
has
at
some
point
in
their
life,
gone
very
deep
down
a
YouTube
hole.
I
think
that
speaks
a
little
bit
to
just
how
like
durable
and
awesome.
A
lot
of
the
algorithms
are
with
YouTube,
but
recently
there's
a
little
bit
of
flak,
because
YouTube's,
restricted
mode
started,
filtering
an
oculus,
lgbtq
content,
and
so
YouTube
has
a
restricted
mode.
I,
don't
know
what
the
exact
sort
of
like
Community
Guidelines
are
for
it,
but
essentially
the
restricted
mode
is
meant
to
limit.
A
You
know
overly
explicit
or
sexual
or
violent
content
and
nature
that
is
still
technically
allowed
to
be
on
YouTube
and
a
lot
of
content
that
was
LGBTQ
in
nature,
but
was
ultimately
innocuous
and
not
really
violating
any
of
the
term
for
restricted
mode
and
we're
still
being
filtered
out.
Youtube
also
issued
an
apology
and
they're
constantly
working
on
some
of
the
data
that
they
feed
into
their
machine
learning
algorithms,
that
sort
of
like
make
these
ecosystems
exist.
A
The
last
example
that
I'll
give
is
VR
so
who,
in
here
has
tried
VR
that
make
sense.
I
would
hope
that
a
lot
of
people
in
this
audience
would
have
tried
it
before.
Vr
is
pretty
cool
and
a
couple
of
years
ago,
actually,
honestly,
maybe
even
today
there
were
a
lot
of
conversations
about
the
ways
that
VR
could
become
a
part
of
our
day
to
day
lives
and
by
day
to
day
I
mean
you
know
the
new
way.
We
watch
movies
a
new
way.
We
play
games
a
new
way.
A
Okay,
that
was
a
lot
of
really
negative
stuff,
so
I'm
gonna
pause
real
quickly.
Do
a
quick
check-in
on
the
group
also
throw
up
a
really
really
cute
gift.
This
is
like
the
most
adorable
thing
I've
ever
seen,
so,
hopefully,
this
kind
of
lightens
up
the
spirits
a
little
bit
but
before
I
go
on
I
just
want
to
pause
and
sort
of
recap,
some
of
the
things
that
we
talked
about.
So
the
first
is
that
we
talked
about
how
you've
been
taught
to
be
inclusive
from
the
very
beginning,
from
the
very
you
know.
A
First
days
of
programming,
you've
been
taught
to
consider
edge
cases.
You've
been
taught
to
consider
ways
that
people
might
intentionally
or
otherwise
break
your
programs
and
how
you
can
sort
of
mitigate
that
from
happening.
We
also
talked
about
some
of
the
things
that
you
might
want
to
consider
in
terms
of
inclusivity.
A
So
what
I'm
gonna
do
right
now
is
apologize
in
advance
because
I'm
gonna
read
the
next
slide,
but,
as
I
was
doing,
research
for
this
talk,
I
came
across
this
really
really
awesome,
Guardian
article
that
was
talking
all
about
algorithmic
biases
and
there's
this
one
paragraph
that
really
sort
of
summarized
what
it
is
that
I'm,
ultimately
getting
at
in
a
really
really
compelling
way,
and
so
I'm
gonna
go
ahead
and
I'm
gonna
read
that
piece
and
talk
about
it
for
a
little
bit.
So
it
goes
as
follows.
A
The
the
type
the
title
of
the
article
is
is
an
algorithm
any
less
racist
than
a
human.
Any
algorithm
can
and
often
does
simply
reproduce
the
biases
inherent
in
its
creator
and
the
data
it's
using
or
in
society
at
large.
For
example,
Google
is
more
likely
to
advertise
executive
level,
salaried
positions
to
search
engine
users
if
it
thinks
the
user
is
male.
Well,
Harvard
researchers
found
that
odds
about
arrest
records
were
much
more
likely
to
appear
alongside
searches
for
names,
thought
to
belong
to
a
black
person
versus
a
white
person.
These
aren't
necessarily
malicious
situations.
A
It's
not
that
Google
is
staffed
by
sexist,
for
example,
but
rather
that
the
algorithm
is
just
mirroring
the
existing
gender
pay
gap,
but
in
so
doing
the
algorithm
reinforces
that
gap,
and
as
long
as
we
continue
to
believe
an
algorithm
is
an
unbiased
machine.
We
risk
reinforcing
the
status
quo
in
harmful
ways.
So
what
I
really
like
about
this
quote
is
that
it
includes
very
pertinent
examples.
A
I'm
sure
lots
of
people
in
here
have
googled
stuff
on
the
interwebs
before
and
it's
interesting
to
think
about
the
fact
that
the
search
results
that
you're
giving
can
be
mirroring
the
status
quo,
and
sometimes
positive
and
sometimes
in
harmful
ways.
But
the
part
that's
the
kicker
here.
That
I
think
is
super
important.
Is
that
if
we
continue
to
believe
in
algorithm
is
an
unbiased
machine,
then
we
will
continue
making
these
mistakes.
A
So
you
might
be
wondering
well
how
do
I
do
that?
How
do
I
is
the
technologist
work
on
making
sure
that
the
things
that
I'm
building
are
inclusive
to
lots
of
people
and
I'm
gonna
tell
you
in
a
second
so
Ice,
Cube,
highly-respected,
profit
and
software
engineer?
Let
me
just
pause
right
now
and
say
he
is
not
a
software
engineer.
I
was
asked
that
question
very
recently,
but
he
does
have
a
fantastic
song.
That's
called
that
says
you
better
check
yourself
before
you
wreck
yourself
and
I.
A
Think
that
is
a
fantastic
mantra
to
keep
in
mind
and
I'm,
going
to
explain
why
in
just
a
second,
so
you
better
check
yourself
before
you
wreck
yourself
is,
first
of
all
easy
to
remember,
I'm,
not
entirely
sure.
If
ice
cube
meant
this
when
he
originally
said
it,
but
I
like
to
believe
that
you
better
check
yourself
before
you
wreck
yourself
means
to
be
cognizant
of
your
implicit
biases.
That's
thing
number
one
just
be
cognizant
of
the
fact
that
you
have
biases.
Every
single
person
in
this
room
is
biased.
A
Every
single
person
has
lived
completely
different
lives
than
the
person
next
to
you,
and
therefore
you
have
biases
about
the
way
the
world
works,
and
as
long
as
you
acknowledge
that
fact
that
is
one
of
the
first
steps
in
sort
of
mitigating
some
of
these
biases
from
ending
up
in
the
products
that
you
build.
So
the
first
part
is
being
cognizant
of
your
implicit
bias,
and
the
second
part
is
thinking
about
how
that
manifests
in
code.
So,
in
other
words,
check
your
priveledge.
A
Part
of
this
is
also
thinking
a
lot
about
privilege,
and
so
every
single
person
in
this
room
right
now
certainly
has
a
fair
amount
of
privilege.
So
I
will
talk
about
some
of
the
privileges
that
I
have
so
again.
The
vast
majority
of
us
have
privilege
some
privileges
that
I
have
first
and
foremost
I
have
dashingly
good,
looks,
but
second
to
that
I
have
I'm
kidding
not
kidding.
I
have
some
sense
of
financial
security.
A
I
am
highly
educated
and
I
have
a
support
system
both
personally
and
professionally,
and
these
are
all
things
that
are
privileges
in
my
life
that
might
shape
some
of
the
decisions
that
I
make.
So,
for
example,
bearing
in
mind
or
not
bearing
in
mind
some
of
these
privileges,
I
might
start
a
company
tomorrow.
A
That's
like
the
next
hottest
cool
startup
in
the
world
and
I
might
say
something
like
ok,
let's
have
our
users
pay
five
bucks
to
play
this
game
that
we're
building
for
me,
if
I,
you
know
were
to
walk
down
the
street
one
day
and
lose
$5
that
wouldn't
really
impact
my
life
pretty
much
at
all.
But
when
we're
thinking
about
that
7
billion
again,
all
of
these
people
around
the
world
for
most
people
around
the
world,
paying
$5
to
play
a
game.
A
It's
a
pretty
significant
financial
undertaking,
and
so
considering
that,
when
you're
building
out
products
and
when
you're
scaling
out
products
is
super
important,
I
might
also
say
something
like.
Please
enter
the
college
that
you
attended
to
confirm
your
identity,
so
I
went
to
college.
A
lot
of
my
friends
went
to
college,
although
this
is
changing
a
little
bit
now
that
I
live
in
San
Francisco.
A
lot
of
my
family's
members
went
to
college,
and
so
when
I'm
building
something
out,
I
might
think
Oh
lots
of
people
have
gone
to
college.
A
We
need
more
information
on
users,
let's
just
have
them
put
in
their
their
college
info
again.
Thinking
about
7
billion
lots
of
people
around
the
world.
Lots
of
women
in
particular
around
the
world,
can't
get
an
education
at
all,
let
alone
a
higher
education,
and
so
that's
inherently
a
limitation.
I
might
also
say
something
like.
Please
enter
your
managers
contact
info
to
verify
your
work
history,
so
this
seems
completely
reasonable.
I
actually
see
it
all
the
time
and
I
and
I
think
it
actually
is
reasonable.
A
But
even
the
act
of
having
a
really
good
awesome,
supportive
manager
is
a
privilege
in
and
of
itself,
because
it's
something
that's
your
control,
I'm
sure
that
some
of
you
guys
have
friends-
or
you
know,
you've
heard
of
horror
stories
with
managers
before
and
a
lot
of
times
that
just
our
circumstances
that
you
unfortunately
end
up
in
and
so
having
a
good
manager
somewhat
of
a
privilege.
And
so
these
are
the
types
of
things
that
you
might
put
into.
Your
software
I
am
not
implying
that
these
are
things
that
you
should
never
do.
A
I'm,
not
implying
that
these
things
don't
make
sense.
Sometimes
they
do
sometimes.
You
are
working
really
hard
on
getting
certain
people's
information.
Sometimes
you
are
just
a
small
start-up
and
you
need
to
be
able
to
cut
corners
or
make
certain
decisions
that
are
tough
and
that's
fine,
but
at
the
end
of
the
day,
remember
that
you're
always
trying
to
scale
out
for
as
many
people
as
possible,
and
in
so
doing
you
need
to
make
sure
that
you're,
aware
of
some
of
the
ramifications
of
some
of
the
choices
that
you
make.
A
Bearing
in
mind
people's
experiences
and
people's
preferences
and
how
comfortable
someone
it
might
be
with
a
certain
sort
of
like
stack
or
framework,
might
impact
your
ability
to
build
certain
things
out
to
cut
corners,
to
make
various
product
decisions
and
so
keep
in
mind
that
even
what
your
preferences
might
be
as
a
programmer
or
what
your
preferences
might
be
as
a
p.m.
or
a
designer
or
whatever
can
impact.
What
ultimately
gets
made,
and
that's
something.
A
That's
super
super
important
to
think
about
so
I
just
drew
in
some
of
the
things
when
I
thought
about
a
lot,
especially
as
a
student
Emacs
versus
vim
angular
versus
react,
Cassandra
versus
server
versus
client,
Java
versus
C++.
All
of
these
things
that
these
are
trade-offs,
for
example,
that
you're
thinking
of
when
you're
building
something
out
can
ultimately
impact
what
ultimately
gets
built.
So
I
already
showed
the
slide,
but
I'm
sure
many
of
you
in
here
we're
waiting
for
the
time
that
I
would
get
to
the
d-word
which
I'm
about
to
show.
A
Making
these
decisions
here
are
the
features
that
we
should
add.
This
is
when
we
should
ship
it
X,
Y,
&,
Z,
and
because
we
just
learned
to
be
cognizant
of
your
own
biases
that
you
have
can
pause
and
I
can
say.
Okay,
a
number
of
these
decisions
were
shaped
by
these
specific
experiences.
Let's
validate
whether
or
not
these
experiences
happened
to
other
people
or
how
often
they
happen
to
other
people
etc,
etc,
to
try
and
make
these
decisions
a
little
bit
more
flexible.
A
However,
I
can
do
that,
but
at
the
end
of
the
day,
I
am
only
one
person,
and
if
you
have
someone
next
to
you,
who
has
had
a
completely
different
set
of
experiences,
who
has
a
completely
different
perspective
on
things
that
you
should
be
thinking
about
when
you're
building
products,
then,
ultimately,
you
can
build
a
piece
of
software
and
you
can
build
a
product
that
can
be
able
to
touch
a
lot
more
people.
Basically
it's
all
about.
A
Like
expanding
the
number
of
considerations
you
can
have
when
building
a
product-
and
it's
a
lot
easier
to
do
that
when
you
have
people
that
are
on
your
teams,
when
you
have
people
in
your
life
that
basically
have
completely
different
perspectives,
and
so
diversity
is
something
that's
super
super
important,
because
ultimately,
the
products
that
you
make
are
that
much
better
and
not
much
cooler
when
you
have
different
people,
sort
of
throwing
ideas
around
to
make
decisions.
So,
ultimately,
what
I'm
trying
to
get
at
is
this.
A
For
the
first
time
in
history,
we
are
learning
what
it
means
to
design
an
engineer
product
for
everyone
in
the
world,
and
that
does
not
mean
that
we
are
there.
Yet.
There
are
certainly
not
seven
billion
people
that
are
connected
right
now,
but,
as
I
mentioned
before,
we
are
that
much
closer
to
actually
getting
there.
A
We
need
to
take
a
good
hard
look
at
what
we're
doing
and
make
sure
that
we're
really
truly
trying
to
make
the
world
ultimately
a
better
place
and
again
it
is
exhausting,
and
it's
very
tiring
to
think
about.
It-
can
be
extremely
drawing
to
try
and
consider
every
single
possible
use
case
when
you're
building
out
products,
but
as
I
sort
of
touched
on
with
some
of
the
examples
that
I
gave.
Sometimes
it
can
result
in
really
bad
experiences
for
people.
A
If
you
don't
do
that,
and
there
are
things
that
you
can
do
to
sort
of
mitigate
them
from
happening
being
cognizant
of
your
own
biases,
building,
diverse
teams
and
having
these
conversations
with
people.
So
ultimately,
the
title
of
my
talk
is
bad
people
about
computers
and
when
it
comes
to
bad
people,
we
can
ultimately
be
talking
about
bad
actors.
People
who
have
malicious
intent
and
intentionally
look
at
different
technologies
and
say
how
can
I
break
this?
How
can
I
you
know,
do
bad
things
using
this
technology?
A
We
want
to
make
sure
that
we're
building
things
that
are
very,
very
difficult
for
bad
people
to
use.
There
also
are
good
people.
People
want
much
like
yourselves,
I'm
sure
who
are
super
smart
and
super
experienced
and
are
building
really
flexible,
awesome
pieces
of
technology,
but
sometimes
have
oversights
that
end
up
manifesting
into
really
really
negative
experiences
and
then,
ultimately,
what
happens
is
good
people
get
perceived
to
be
bad
people?
Then
there
is
bad
computers.
A
Now,
ultimately,
computers
aren't
sentient
beings
necessarily,
but
if
you
are
someone
who
has
limited
sort
of
exposure
to
technology
or
even
a
lot
of
exposure
to
technology-
and
you
discover
that
a
lot
of
your
experiences
are
relatively
negative,
then
the
technology
itself,
the
computers
itself
can
seem
like
it
is
inherently
a
bad
thing
that
is
out
in
the
world,
and
we
definitely
don't
want
people
around
the
world.
Thinking
that
technology
is
something
that
is
inherently
bad.
A
So
to
sort
of
close,
my
talk
when
I
had
first
written
this
talk
a
couple
of
years
ago
and
I
was
doing.
Reese
I
had
happened
to
just
read
this
book
called
when
breath
becomes
air
and
I
highly
recommend
it.
It's
a
really
beautiful,
read,
quick,
read
and
the
book
basically
does
this
really
interesting
sort
of
exploration
of
what
it
means
to
sort
of
live
life
and
explorations
of
death
as
well?
A
And
there
was
this
quote
that
really
stood
out
to
me
in
terms
of
how
I
personally
want
to
live
my
life
and
the
decisions
and
the
choices
that
I
make
us.
Someone
on
this
planet
who
hopefully
leaves
it
being
better
off
than
it
was
before.
I
came
here
and
because
I
obviously
work
in
technology
and
spend
a
lot
of
my
life
building
technology,
it
sort
of
guides
the
way
I
think
about
tech
as
well,
and
the
quote
is
as
follows.
A
You
can't
ever
reach
perfection,
but
you
can
believe
in
an
asymptote
toward
rich.
You
are
ceaselessly
striving,
so
there
will
always
be
bad
people
and
there
were
always
the
flaws
and
the
technologies
that
we
build,
but
I
hope
that
after
today
that
each
and
every
one
of
you
ceaselessly
strive
towards
this
asymptote
of
good
people
of
good
computers
and,
ultimately
of
a
better
world.
Thank
you.