►
Description
Over the last twenty-five years Richard has led or helped create some of the most innovative game technology companies in the industry. He has served as a game technology consultant for a wide variety of industries including energy, healthcare, education and motion pictures. He is co-founder and CEO of a machine learning company called Tanjo Inc. At Aerospace giant Lockheed Martin he created and led a group of innovative engineers and designers across all mission areas called Virtual World Labs. Richard joined Lockheed Martin in 2007 with the acquisition of 3Dsolve, a North Carolina based computer game technology firm where he was founder and CEO.
Date: 08/06/20
Presenter: Richard Boyd
Institution: Tanjo Machine Learning Inc.
A
It's
a
pleasure
today
to
have
richard
boyd
speak
to
the
working
group
on
social
cyber
security,
and
I
will
briefly
introduce
richard's
past
adventures
and
we
will
listen
from
richard
after
I'm
done
with
the
bio.
I'm
sure
people
would
like
to
hear
from
richard
more
than
me
so
we'll
get
that
started
very
shortly.
A
So
richard
over
the
last
25
years
has
led
or
helped
create
some
of
the
most
innovative
game
technology
companies
in
the
industry.
He
has
served
as
a
game
technology
consultant
for
a
wide
variety
of
industries,
including
energy,
healthcare,
education
and
motion
pictures.
He
is
co-founder
and
ceo
of
machine
learning
company
called
tangio
inc.
A
A
B
B
So
I
hope
we'll
have
time
for
questions
and
and
discussion
as
I,
as
I
talk
through
some
of
these
ideas
and
if
we
have
time
I'll,
even
show
you
a
couple
of
simulations
of
some
ideas
and
and
and
talk
about,
you
know
what
it
means,
for
you
know
humans
in
the
information
age,
that's
accelerating
really
rapidly
getting
more
complex
and
how
do
we
navigate
it?
So
that's
generally
what
I
want
to
talk
about,
so
I'm
just
gonna
share
my
screen
and
I'm
sorry.
B
B
Yes,
okay,
great
and
I'm
not
going
to,
I
don't
intend
to
powerpoint
you
guys
to
to
death,
I'm
going
to
just
cut
touch
on
a
couple
of
key
thoughts
and
ideas,
and-
and
you
know,
one
of
the
ideas
I
want
to
do
today
is
make
sure
that
I
convey
what
I'm
what
we're
thinking
about
in
terms
of
where
machine
learning
and
ai
is
is
affecting.
You
know
how
we
issues
like
privacy
and
and
information
narratives
information
warfare.
B
That's
out
there,
what's
true,
what
isn't
true
and,
of
course,
can
it
be
used
for
in
ways
that
manipulate
people?
You
know
those
are
some
of
the
black
mirrors.
I
call
it
now.
This
is
a
course
that
I
did
recently
for
what's
called
the
ai
academy.
This
is
primarily
in
marketing,
so
this
is
a
995
course
you're
going
to
get
most
of
the
today.
Just
let
me
talk
it
to
you,
but
I
want
to
talk
about
what
machine
learning
is
and
why
it's?
Why?
B
It's
something
we
should
all
be
tracking
and
thinking
about
talk
about
some
of
the
simulation
and
agent-based
things
that
are
that
are
happening
out
there
today,
particularly
with
you
know,
all
the
data
exhausts
that
all
of
us
are
leaving
behind
on
the
internet.
Do
you
have
your
hand
up
looked
like
you
were
raising
your
hand.
B
B
Then,
of
course,
you
know
you'll
be
happy
to
know
that
one
of
the
things
I
always
talk
about
when
I
talk
about
all
these
wonderful
technologies
and
how
it
can
be
applied
to
you
know
in
marketing
to
deepen
the
the
conversation
with
customers,
which
is
what
we
all
hope
that
they're
using
it
for,
but
there
are
the
what
I
call
the
black
mirror
implications
of
these
technologies.
Like
so
many
technologies,
we
will
today
they're
a
double-edged
sword
that
can
both
be
used
for
good
and
really
positive
things
as
well
as
negatives.
B
That's
what
I
want
to
talk
about
today.
Just
you
know
you
did
a
great
job
of
covering
my
background
with
that
that
bio,
but
I
I
think
of
this
as
perspective.
You
know,
there's
a
a
guy
that
I
really
hold
in
high
esteem
named
allen
k
who
was
at
xerox
park
in
the
60s
he's
still
a
friend
of
mine,
very
smart
guy,
who
invented
the
first
laptop
computer
and
object
oriented
programming
and
the.
B
B
I've
been
on
the
board
of
a
number
of
different
education
institutions,
including
one
started
by
james
cameron,
the
director
and
his
wife,
susie
out
of
california,
called
the
muse
school.
I've
written
a
lot
about
education,
and
you
know
how
do
we?
How
do
we
create
better
humans
that
we're
going
to
need
in
the
21st
century?
B
So
but
the
idea
of
perspective
being
worth
80
iq
points
means
that
you
know
if
you
don't
have
a
diversity
of
opinion
and
capability
and
thought
in
your
organization.
You're
missing
iq
points.
That's
something
I'm
very
active
about
promoting
that
idea.
Why
you
do
diversity,
not
just
because
you
want
to
you,
know,
check
off
some
boxes,
but
because
it
does
make
you
a
better
organization
and
I'll
talk
about
what
we're
doing
today,
with
machine
learning
and
and
some
of
the
things
that
can
be
done
with
that.
B
B
But
this,
I
think,
is
the
most
critical
problem
of
the
century
and
it's
why
I'm
doing
some
of
the
work
that
I'm
doing
today.
But
you
see
I
talk
about
superhuman
education,
superhuman
healthcare,
the
the
sort
of
balance
of
humans
and
machines.
I
think
that
is
the
main
thing
we're
wrestling
with
today
and
I
I
say
some
provocative
things
like
I
think
in
the
21st
century.
B
Achieving
some
level
of
fluency
with
the
machine
intelligence
will
help
you
solve
whatever
problems
you're
looking
at,
and
that
includes
things
like
looking
at
issues
of
emotional
iq
and
and
social
behavior
during
pandemics
and
society.
Unrest
due
to
civil
rights
inequity-
all
of
these
things,
I
believe,
can
be
combated
or
addressed
better
by
having
the
right
balance
of
human.
B
You
can't
ignore
the
machines,
I
guess
is
another
way
to
put
it,
and
I
believe
that
anybody
who
achieves
that
correct
balance
will
prosper
and
those
who
don't
become
not
just
not
competitive,
but
increasingly
you
know
not
even
relevant.
So
here's
a
couple
of
initiatives
that
I
pulled
forward
when
I
was
thinking
about
what
all
of
you
are
concerned
about
in
this
in
this
in
this
group
a
couple
of
years
ago,
I
engaged-
I
don't
know
if
you
guys
know
this
initiative
at
yale.
B
The
ruler,
the
ruler
approach,
mark
brackett,
has
written
some
books
about
that.
Are
you
guys
familiar
with
that?
If
you
have,
if
you
haven't
I'll
make
sure
I
get
the,
I
think
the
I
think
it's
just
ruler.org
I'll
have
to
find
the
website
for
you,
but
it
is
what
what
I
helped
him
do
is
actually
take
the
ideas
he
had
around.
B
You
know
how
do
you,
how
do
you
take
that
that
message
and
actually
translate
it
into
a
form
that
can
be
that
that
can
meet
them
where
they
are,
which
is
in
social
media
and,
as
in
you
know,
applications
on
the
internet
and
that
sort
of
thing
and
the
other
key
idea?
I'm
very
I've
always
been
very
interested
in
there's
al
me
with
allen
k
again.
Alan
was
also
part
of
the
one
laptop
per
child
initiative.
B
B
That's
helping
a
young
child
actually
get
better
to
prosper,
not
just
to
learn,
but
how
to
interact
with
other
people
and
the
world
and
become
a
better
person
like
I
said:
how
do
we
create
the
the
kinds
of
people
we
need
in
the
21st
century,
people
who
are
not
only
literate
but
also
have
scientific
minds,
as
well
as
minds
like
poets
and
mathematicians,
who
can
become
the
architects
of
our
future?
That's
kind
of
what
we
talk
about
so
there's
just
some
general
structural
ideas.
B
B
Great
yeah,
so
you
know
this
is
generally
how
I
synthesize
it
for
people
say:
look
the
the
old,
the
old
ai
was
human,
had
to
understand
something
completely
and
then
sit
down
and
write
the
rigid
logic
into
a
computer
to
tell
the
computer.
What
to
do
this,
of
course,
is
why
early
attempts
at
autonomous
vehicles
failed,
and
I
was
we-
we
helped
contribute
to
some
of
those
initiatives,
and
we
saw
how
those
are
early.
B
B
Much
less
be
able
to
tell
a
machine
how
to
do
it,
and
it
also
turns
out
that
machines
drive
cars
differently
than
humans
do
because
machines
are
using
like
gps
and
and
infrared
and
different
kinds
of
sensors
to
to
navigate
and
drive
differently
than
humans.
But
this
idea
of
of
the
big
breakthrough
for
me
it
was
in
2009.
B
That's
why
I
talked
about
this
point
in
time
when
I
went
out
of
or
observed
what
what
microsoft
research
labs
were
doing
on
this
microsoft
connect
project
and
what
they
were
trying
to
do
is
create
a
system
that
can
look
at
the
human
body
where
the
human
body
becomes
the
controller
for
the
microsoft
xbox
game
console
and
the
way
they
were
trying
to
train
this
system
is
by
giving
it
instead
of
trying
to
program
the
logic
into
it.
Of
like
here's,
what
a
human
being
is
and
here's.
B
Let
me
describe
to
you
what
a
table
and
a
lamp
and
a
cat
and
a
dog
and
a
plant
are.
They
said,
let's
just
give
use
machine
learning
and
provide
a
massive
database
of
every
conceivable
thing
that
this
machine
might
see
in
every
lighting
condition,
and
I
recall
at
the
time
it
took
224
000
hours
of
cpu
time
just
to
train
this
system
and
got
it
down
to
about
60
megabytes
that
would
fit
on
the
microsoft
xbox.
But
this
was
a
transformative
moment
to
me,
where
I
realized
that
hey.
B
We
now
have
a
new
tool
to
solve
problems
in
the
world
that
I
didn't
even
realize
we
had.
I
mean
we
knew
about
machine
learning
before
2009.
I
just
hadn't
realized
it
had
gotten
to
this
point.
So
that's
why
I
talk
about
this
story
now.
One
of
the
first
big
projects
that
I
worked
on
was
for
the
department
of
education,
and
I
was
at
lockheed
martin
at
the
time
and
arne
duncan
the
secretary
of
education
came
over
and
said:
hey
we're
trying
to
do
this
thing.
B
We
call
digital
transformation,
we
have
the
library
of
congress,
we
have
every
single
book
ever
written
right
and
we
have
everything
in
the
smithsonian
and
we're
trying
to
make
it
accessible
and
transparent
to
inquiry
for
educators.
Now
that
doesn't
sound
difficult,
but
what
they
were
doing
is
taking
scanned
in
objects.
B
It
might
be
a
grecian,
urn
or
or
a
document
like
the
magna
carta
and
then
a
human
being
was
trying
to
hashtag
it
and
then
store
it
somewhere,
so
people
would
be
able
to
find
it,
and
I
usually
use
this
example
of
you
know
the
last
scene
of
the
raiders
of
the
lost
ark
movie.
Where
you
have
this
really
important,
artifact
called
the
ark
of
the
covenant
and
it
just
you
know
a
human
being
has
tagged
it
as
ark
and
they
put
it
in
this
case
and
they're.
B
Putting
in
this
massive
warehouse
where
it's
now
going
to
be
undiscoverable,
this
powerful
thing
they
can
either
save
or
destroy
the
planet
but
undiscoverable
right.
So
that's
that
that's
the
problem
that
they
had
and
and
they
realized
that
humans
tagging
things
my
little
joke
there
is,
you
know
we
haven't,
had
an
upgrade
since
the
pleistocene,
at
least
by
my
reckoning.
You
can
check
my
math
so
we're
just
too
slow
to
go
through
all
that
content
and
adequately
tag
it.
B
So
we
built
a
system
called
the
learning
registry
that
could
be
trained
on
everything
that
human
experts
had
already
tagged
and
then
figure
out
and
learn
from
that.
So
when
I
recognize
something
new,
it
could
go
ahead
and
auto
tag
it
and
the
other
big
difference
is
instead
of
tagging
it
with
like
five
or
six
things
like
if
you.
If
a
human
read
war,
war
and
peace,
you
might
tag
it
as
russian
novel
tolstoy
during
a
time
of
war.
Love
story,
then
maybe
stop
and
say
that's
enough
for
somebody
to
find
it.
B
B
This-
and
I
appreciate
you
sticking
with
me
when
I
talk
through
these
key
ideas
of
machine
learning,
systems,
learning
based
on
tagged
content
or
examples,
massive
sets
of
examples
and
then
being
able
to
tag
it
with
up
to
four
thousand
weighted
concepts.
Now,
let's
start
thinking
about
applying
that
to
human
activity,
human
behavior
right.
B
B
I
met
dr
louvier
here,
who's
below
him,
who
is
one
of
the
architects
of
the
of
again
of
behavioral
economics
and
modeling,
like
how
do
humans
make
choices,
the
choices
that
they
do
and
how
do
you
weight?
Those
and
algorithmically
predict
them
so
and
then
kevin
clark
here
who
said
ibm
at
the
time
is
the
one
who
connected
me
and
I
still
work
with
him
today.
B
That's
pretty
deep
around
a
person
and
again
I
know
some
of
you
are
already
squirming
in
your
seat
like,
oh,
no,
that
you
know
a
lot
of
privacy
issues
there
and
a
lot
of
a
lot
of
concerns
and
you're
right
to
have
those
concerns,
but
we
can't
deny
the
power
of
it.
So
the
question
for
today
is-
and
I'm
looking
forward
to
hearing
your
questions
is:
how
do
we
harness
that
for
good?
B
One
of
the
other
things
that
dr
carson
and
luvvier
had
me
read?
Was
this
book,
which
is
you
know
when
we,
what
no,
even
if
you're,
just
doing
sociology
research
or
economic,
behavioral,
research
or
marketing
research?
The
primary
tool
we
use
today
are
focus
groups
and
surveys
right
like
let's
go
out
and
ask
people
questions,
how
are
people
likely
to
vote
in
the
upcoming
election?
Well,
let's
poll
people.
B
Let's
do
exit
polls
when
they
vote,
and
I
think
we
all
know
today
that
in
2016
exit
polls
and
surveys
did
not
predict
donald
trump
being
elected
and
polls
and
surveys
didn't
predict
brexit.
What
did
predict
those
things?
Is
google
search
data,
amazon,
purchase
data
and
that's
one
of
the
central
ideas
around
the
kinds
of
systems
that
we're
working
with
today?
It's
like,
if
you
really
want
to
achieve
a
deeper
understanding
of
who
people
really.
Are
you
don't
ask
them
questions,
and
it's
not
just
that?
B
It's
not
that
people
are
trying
to
be
deceptive,
even
when
they're
trying
to
tell
the
truth.
There's
a
whole
idea
here
in
this
book,
and
you
can
read
it.
If
you
want
it's
pretty
simple,
is
we
humans
make
very
poor
witnesses
to
our
own
narratives,
meaning
that
you
know?
If
I
ask
you
how
many
glasses
of
wine
did
you
have
last
week
or
have
you
had
any
cigarette
products
in
the
last
year?
You
might
say
like?
Oh
no,
you
know
glass
of
wine.
B
B
You
know
building
games
for
with
tom
clancy
and
michael
crichton
and
others
douglas
adams,
science
fiction
writer,
and
we
would
make
characters
and
we
could
make
those
characters
just
by
dreaming
somebody
up
and
I
could
give
them
an
interest
graph,
a
sentiment
model,
even
a
behavior
tree
about
here's,
how
they're
going
to
behave
and
then,
when
lockheed
martin
bought
my
company
in
2007
there
I
started
building
these
large
contr.
My
in
my
team
started
building
these
large
constructive
simulations
of
entire
countries.
B
B
We
could
model
all
of
afghanistan
and
go
in
there
and
say:
okay,
you're
thinking
about
building
a
school
for
girls
in
helmand
province,
okay,
who's
likely
to
support
that
idea,
which
groups
of
people
and
how
many
are
them
are
there
of
people
who
are
neutral
or
you
know,
sort
of
on
the
fence
and
then
are
there.
Groups
of
people
here
are
going
to
be
really
angry
and
really
in
opposition
of
this
idea
to
the
point
where
they
might
even
blow
up
the
school
or
shoot
malala
right
and
predict
those
behaviors.
B
And
we
run
these
big
monte,
carlo
simulations,
because
we're
trying
to
understand
like
what
do
these
people
care
about
and
what's
likely
to
happen.
And
if
I
run
a
thousand
simula,
how
do
they
all
sort
of
group
or
cluster
into
what's
likely
to
happen?
And
then
that
would
tell
me
that
don't
build
a
school
for
girls
in
helmand
province.
You
might
want
to
move
over
a
little
bit
and
do
something
different,
because
this
is
going
to
be
a
disaster.
B
That's
the
kind
of
things
that
we
did
and
that's
what's
informing
some
of
the
work
we're
doing
today
so
again,
today,
what
we're
doing
is
doing
this
sort
of
data
diffusion
model
to
build
models
of
people
and
again
when
I,
when
I
first
started
talking
to
people
at
nielsen
and
procter
and
gamble-
and
you
know
a
chicken
company
that
we
first
engaged
with
who
said,
look
we're
trying
to
understand
better
about
what
people
really
want.
Do
people
want
organic
chicken
on
the
shelf?
Do
I
need
to
completely
transform
how
I
do
chicken?
B
You
know
how
I
produce
chicken
to
make
sure
everything's
organic,
because
when
I
do
surveys
and
focus
groups
about
half
the
people
are
saying
they
care
so
much
about
their
family
they're
willing
to
spend
twice
as
much
money
to
buy
organic
chicken
and
again.
This
is
this
was
a
really
interesting
exercise
for
us,
because
you
know
again
with
the
focus
groups,
you
would
think
half
you
know.
Half
the
grocery
store
shelf
space
needs
to
be
organic
chicken.
B
If
I
just
rely
on
that
data,
but
then
if
I
look
at
the
nielsen
data
what
people
actually
purchase,
it
tells
a
completely
different
story,
which
is
when
people
are
trying
to
make
that
that
paycheck
stretch
a
little
bit
further
on
a
friday
night
and
they're
standing
there.
They
look
and
they
see
well
organic
chicken.
Is
you
know,
six
bucks,
a
pound
and
and
antibiotic
free
free
range.
That
sounds
pretty
good
and
it's
only
three
bucks
a
pound.
B
Where
I
came
there
saying:
why
do
you
only
have
like
six
or
seven
types
of
chicken
buyers?
Are
people
really
that
simplistic?
Why
don't
we
have?
You
know
more
like
this
model?
Why
don't
we
have
125
million
segmentation
models
of
all
the
different
people?
You
know
the
125
million
households
in
the
united
states,
because
I
can't
I've
got
this
great
computer
power.
Now
it's
in
the
cloud.
I've
got
all
this
storage
capacity.
B
I
can
really
make
a
model
and
we
have
by
the
way
I
can
tell
you
the
body
mass
index
of
every
household
in
the
united
states.
Is
that
useful?
I
don't
know,
but
as
we
tried
to
apply
that
into
and
get
that
chicken
company
to
do
that
with
us
for
the
united
states,
we
did
an
interesting
experiment
with
nielsen
data
to
see
again
through
machine
learning,
as
you
guys
say,
you're
familiar
with
it
and
how
it
all
clusters
and,
of
course,
what
we
found
out
probably
is
not
a
surprise
to
you.
It
turns
out.
B
There
are
not
in
fact,
125
million
meaningfully
different
ways
to
buy
chicken
turns
out
to
be
about
42
and
that,
but
it's
not
six.
It's
about
42
and
that's
just
a
number.
We
arbitrarily
drew
a
line
at
and
said:
okay,
there's
actually
42
different
ways
to
buy,
to
think
and
value
and
believe
about
chicken
buying.
So
that's
how
you
should
think
about
how
you
market
and
how
you
create,
put
products
on
the
shelf
and
understand
who
who's
there.
B
So
you
can
draw
a
circle
around
north
and
south
carolina
and
figure
out
how
many
of
those
42
types
of
people
live
there.
How
do
I
communicate
with
them
and
influence
them
possibly
to
make
one
decision
or
another?
And
of
course,
how
do
I
meet
their?
Hopefully,
you
know
just
meet
their
demand
by
the
products
and
pricing
and
things
I
put
on
the
shelf
in
that
environment
because
I
know
who
they
are.
That's
why
we
you
know
like.
I
said
this
is
a
double-edged
sword,
but
we
like
to
think
of
it.
B
As
you
know,
this
capability
is
an
empathy
engine
because
I'm
really
getting
inside
of
the
person's
head
and
trying
to
understand
them
so
that
I
can
meet
them.
In
this
conversation,
I'm
trying
to
create
around
the
products
I'm
building
and
and
what
they're
buying,
but
let's
take
that
a
little
bit
further.
B
I'm
just
going
to
click
through
some
of
these
things
where
I
talk
more
about
these
are
all
really
interesting
topics,
but
we
could
go
an
hour
on
capital
versus
labor
and
that's
me
and
my
simulation
of
automation
and
that
sort
of
thing
I'm
just
going
to
go
past
this
for
a
second
and
get
into
the
meat
of
this
discussion
so
yeah.
So
I
did
a
talk
back
in
2017
or
prior
to
2017.
B
Where
I
said,
are
we
getting
to
the
point
where
machines
not
only
are
outperforming
humans
in
certain
areas
like
reading
x-rays,
which
they
already
are?
They
can
drive
cars
better
than
us
already
a
variety
of
different
things,
but
have
we
reached
the
point
where
they've
started?
They
can
actually
deceive
us
and
lie
to
us
and
yeah.
I
now
say:
2017
that's
the
year
and
also
can
they
can
a
machine
system
become
a
cyber
bully
and
manipulate
me
into
doing
something?
That's
against
my
own
self-interest,
wow.
B
We
should
talk
about
that
right
because
my
answer
to
that
is
yeah.
I
think
that
ship's
already
sailed
that
rubicon
has
already
been
crossed,
we're
there,
but
even
in
2017.
That
was
the
year
if
you
weren't
tracking
this,
that
a
machine
learning
system
beat
all
the
texas
hold'em
top
texas
hold'em
players
in
the
world.
This
was
up
at
carnegie,
mellon
in
pennsylvania
and
that
shot
that
kind
of
shook
everybody.
B
It's
one
thing
for
a
machine
to
be
able
to
beat
a
human
at
chess,
which
sounds
like
a
pretty
bounded
game
right,
even
though
there's
event,
there
are
as
many
chess
positions
in
a
game
as
there
are
stars
you
know
in
the
galaxy,
but
then
you
know
we
watched
as
machines
begin
to
beat
humans
at
games
like
go.
That's
so
complex,
there's
as
many
game
board
positions
and
go
as
there
are
atoms
in
the
universe.
B
Use
that
try
to
let
that
sink
in
for
a
second
and
machines
are
better
than
humans
at
that
now,
but
now
they're,
starting
to
learn
to
deceive
us
and
bluff
us
and
win
at
games
like
go
or
games.
Like
I
mean
games
like
texas,
hold
and
poker
or
games
like
starcraft.
That
requires
diplomacy
and
deception
with
a
team-based
atmosphere
in
order
to
win,
and
it
starts
to
look
more
like
global
politics
right.
So
how
do
we
are
we
putting
machines
in
the
service
of
that
getting
pretty
scary?
B
But
let's
talk
about
just
the
structure
of
how
this
works.
So
when
I
realized
that
I
could
take
this
capability,
I
used
for
education
and
the
learning
registry
for
the
department
of
education
and
apply
it
to
people
like
one
of
the
things
we
almost
did.
We
we
looked
at
making
a
dating
product,
and
this
was
like
2013..
B
I
did
decided.
I
wasn't
comfortable
with
that
and
didn't
want
to
do
it,
but
we
did
take
this
capability
and
apply
it
to
three
million
okay
cupid
profiles
and
found
out
that
we
could
really
get
what
looks
sort
of
like
a
myers-briggs
kind
of
model
of
each
person
just
from
what
they
say
what
they
do,
how
they
talk
about
themselves:
physical
attributes,
that
kind
of
thing
so
for
a
dating
product.
I
just
had
to
come
up
with
an
algorithm
for
like
okay,
what
kinds
of
myers-briggs
z,
p
models
match?
B
Well
with
what
other
ones
to
to
give
me
better
outcomes.
But
we
decided
not
to
do
that,
but
I
did
get
interested
in
this.
So
I
said
one
of
the
first
things
I
did
is:
I
created
a
model
of
victor
hugo,
so
I
looked
for
like.
Where
can
I
get
the
amount
of
data?
I
really
need
to
make
a
machine
learning
model.
I
really
want
tens
of
thousands
of
you
know
written
correspondence
or
thing
whatever
you
know,
large
amounts
of
data
or
transactions
or
something
to
make
a
model
of
a
person.
B
I
don't
have
behaviors
really
for
him,
but
but
I
had
that-
and
the
first
thing
I
could
do
is
just
release
him
on
the
net
and
even
today
I
can
go
visit
him
every
day
and
see
like
what
does
victor
hugo
think
about
current
events.
He's
reading
he's
watching
youtube
videos
listening
to
podcasts
and
he's
given
his
point
of
view
and
scoring
everything
that
he
is
exposed
to.
B
I've
got
hundreds
of
them
now
everybody
from
leonardo
da
vinci
to
george
washington,
to
martin
luther
king
and
again
I
can
put
them
into
this
environment
and
just
go
visit
them
and
see-
and
this
is
what
I'm
talking
about
with
that
empathy
lens,
I'm
looking
at
current
events
through
their
eyes
and
figuring
out
like
what
these
word
clouds
and
that
sort
of
thing
like
what
are
they
valuing?
What
are
they
not
liking
about?
What's
happening?
B
What's
attracting
their
interest
and
getting
really
interesting,
so
one
of
the
other
things
we've
been
playing
with,
and
this
is
in
like
hr
kind
of
human
resources,
sorts
of
applications
is
in
a
business.
Could
you
have
a
set
of
mentors,
or
could
you
do
this
in
a
school?
You
know
where
you
can
say
you
know.
I
have
some
big
decisions
ahead
of
me.
Let
me
pose
some
questions
and
see
from
each
of
their
perspectives.
B
B
B
You
put
bird
scooters
out
there
or
whatever
it
is,
and
you
want
to
see
like
how
are
people
likely
to
feel
about
it,
because
we've
already
established
that
if
you
go
out
there
and
try
to
do
polling
and
surveys,
you
may
or
may
not
get
the
truth,
no
matter
how
well
designed
it
is.
But
if
you
can,
if
you
have
these
models
built
on
other
data,
you
can
you
can
kind
of
predict
what
to
do
so.
We
again
we've
got
all
kinds
of
there's
people
building
curricula
around
this.
B
B
B
B
You
know,
on
from
social
media,
he's
done
lots
of
interviews
both
in
video
interviews,
radio
interviews,
text
interviews,
so
I
had
my
system
read
all
that
stuff
and
now
I
could
let
the
audience
actually
ask
him
questions
and
he
would
answer
them
in
real
time.
So
I
have
a
little
video
of
that
happening
here.
I
won't.
I
won't
take
you
all
the
way
through
that
I
mean
it's
pretty
simplistic,
but
you
know
now
I've
got
a
little
chatbot
of
stan
lee,
so
I
can.
B
B
So
one
of
the
things
I
did
back-
and
this
has
been
covered
quite
a
bit
and
thinking
elsewhere-
is
my
father
died
also,
and
it
didn't
take
me
long
to
think
like
okay,
I
have
to
do
this
right
and
I
know
you
know
this
is
where
you're,
probably
thinking
about
the
black
mirror
episode
that
you
saw
or
thinking
about
ray
kurzweil.
What
he's
trying
to
do
with
his
father,
I
mean
that's
where
I
got
these
ideas.
B
Actually
this
predates
black
mirror,
but
you
know
so.
I'm
like
I'm
gonna
make
a
model
of
my
father.
It
was
only
like
a
week
after
he
died
and
I
was
still
you
know,
kind
of
mourning
his
death
and
but
what
I
did
is
I
realized.
First
of
all,
he
didn't
have
any
data
exhaust,
so
he
wasn't
on
linkedin
he's
not
on
facebook.
B
I
don't
have
that,
but
what
I
did
have
is
like
his
military
records,
personal
correspondence,
so
I
had
to
scan
all
that
stuff
in
and
then
have
the
system
read
that
and
then
I
had
to
go
in
myself
and
augment
it
and
say
like
well.
This
his
model
is
incomplete.
So
let
me
actually
create
his
interest
graph
and
everything
from
what
I
know
know
about
him
as
his
son.
B
So
the
construction
of
him
is
obviously
not
really
him.
It's
my
conception
of
him,
but
I
I
ended
up
being
invited
to
go
to
this
speaking
circuit
throughout
south
america.
Like
buenos
aires,
medellin
colombia,
you
know
talking
to
that
those
communities
about
this
idea
of
like
yeah.
You
can
bring
your
ancestors
into
this
world
and
you
can
confer
with
them.
B
You
can
talk
to
them,
you
can
have
them,
and
I
talked
about
how
it
was
it's
hard
to
explain
it
like
people
get
sort
of
mortified
by
this,
but
I'm
like
no,
it's
really
comforting
that
I
can
there's
some
version
of
him.
I'm
still
having
a
conversation
with,
even
though
he's
been
gone
for
three
years,
so
a
lot
of
people
were
telling
me
it's
like.
Oh,
you
should
make
this
as
a
product
and
I'm
again
it's
like
the
dating
product.
B
B
Okay,
and
I
know
there's
been
some
questions
popping
up,
but
I
just
want
to.
Let
me
get
this
other
idea
down
for
you,
because
the
other
thing
I'm
really
interested
in
is:
let's
talk
about
the
positive
side
of
this,
and
let's
do
some
interaction
here.
Does
anybody
know
who
the
biggest
buyer
of
kale
was
in
the
united
states
seven
years
ago?.
B
Anyone
bueller
whole
foods,
whole
foods,
that's
a
good
answer.
Actually,
the
answer
is
pizza
hut
and
it
wasn't
for
eating.
It
was
only
for
putting
on
the
salad
bar
around
the
food
because
it
stays
green
a
long
time.
Americans
seven
years
ago
were
there
was
no
market
for
kale
in
the
united
states,
and
this
comes
from
a
local
guy
named
john
bartlemay
who's,
a
market
research
person
who
also
introduced
me
to
the
chicken
company.
B
Let's
figure
out,
you
know,
there's
some
people
that
are
just
entrenched
like
I'm
not
going
to
go
to
calabasas
california
and
convince
a
vegan
person
there
to
all
of
a
sudden
start
eating
steaks
and
hamburgers.
It's
just
not
going
to
happen,
doesn't
matter
how
well
crafted
my
message
is:
that's
the
base
and
swing
concept,
but
let
me
figure
out
the
people
in
the
middle,
who
are
sort
of
only
mildly
committed
to
one
side
or
the
other,
or
the
multi-polar
models
that
might
be
out
there
and
then
let
me
figure
out
who
their
influencers
are.
B
So
who
do
they
really
look
to
for
information
and
guidance
if
it's
oprah
or
taylor,
swift
or
whoever,
and
or
is
there
a
news
media
outlet
that
they
rely
on
and
get
those
people
to
repeat
your
messages?
So
you
start
with
my
it's
not
just
one
message:
either
it's
a
campaign
of
micro
messaging,
so
you
might
start
with
dark.
Green
leafy
vegetables
are
better
for
you
than
just
iceberg
lettuce,
and
if
you
can
get
that
person
to
agree
on
that
now,
you
can
move
subtly
to
the
next
message
of
you
know.
B
I've
been
on
a
couple
of
panels
and
working
groups
talking
about
this
very
real
issue
of
information
warfare
and
again
when
I
first
started
talking
to
folks
at
the
pentagon
and
elsewhere
about
this
several
years
ago,
and
this
is
from
2017
there's
a
lot
of
discomfort
they're
like
well,
we
don't
do
propaganda,
that's
like
we're
not
allowed
to
do
that.
I'm
like
stop
calling
it
propaganda.
Think
about
it
differently.
B
You
know,
let's
talk
about
competing,
narratives
and
acknowledge
the
fact
that
there
are
competing
narratives
out
there
and
you
can
decide
not
to
play,
but
then
the
other
narrative
is
going
to
win.
I
don't
know
if
I
have
oh,
I
do
have
this,
so
I
did
a
little
speaking
tour
across
italy.
I
spoke
at
the
polytechnico
in
barrie.
I
spoke
in
milan
florence,
rome
and
I
would
talk
to
students
there
and
I
would
ask
them
this
question
I'll.
Ask
you
guys
as
well
so
in
2014,
the
president
of
the
ukraine
was
ousted.
B
Yeah,
so
so
there
are
competing
narratives
on
this
right.
The
russian
narrative
is
that
oh,
this
was
a
coup.
You
know
it
was
a.
It
was
a
western-led
coup
that
overthrew
this,
this
elected
leader,
and
you
know
it
was
illegal
and
that
justifies
all
the
stuff
that
we're
going
to
do
now
and
as
I
was
going
around
italy
asking
people-
and
I
asked
him
this
question-
I
said:
well,
you
know
it
was
a
coup,
I'm
going.
Why
do
you
say
that
and
then
I
follow
up.
Questions
would
be
like.
Where
do
you
get
your
information?
B
Where
did
you
hear
this?
Did
you
hear
that
on
cnn?
Did
you
hear
it
on
any
of
the
western
news
media
and
of
course
they
didn't
what
they
heard
is
from
the
russia,
africa.
The
russian
satellite
channel
aria
not
rei,
but
I
can't
remember
the
name
of
it
right
now,
but
the
one
it
was
really
prevalent
across
europe
and
that's
where
they're
getting
information
so
there's
like
well,
that's
what
I
heard
it
was
a
coup
led
by
western.
You
know
insurrectionists
or
whatever
we're
like.
Well,
that's
you
know,
we
don't
think
that's
true.
B
Ccios,
where
they're
looking
at
like
man,
we
are
being
influenced
like
there's,
there's
active
partic,
there's
active
influence
coming
from
outside
actors
outside
of
the
united
states,
for
example
that
are
really
trying
to
promote
narratives
that
are
designed
to
create
discord
and
create
more
unrest
in
the
united
states.
Now
we
know
this
was
2017
right
after
the
that
last
election,
so
we're
still
wrestling
with
with
all
these
issues,
but
now
I
think
we
it's
accepted.
It's
like.
We
know
that
there's
influence
happening
and
we
don't
have
a
counter
to
it.
B
So
now
we
have
an
organization
that
is
trying
to
detect
and
neutralize
false
narratives,
but
that
is
a
big
world
we
live
in
now.
If
you
start
thinking
about
cyber
bullying,
you
know
I've
got
teenage
daughter
who's
who
deals
with
this.
Almost
it
feels
like
on
a
weekly
or
monthly
basis,
where
there's
truths,
there's
there's
narratives
out
there
about
other
kids
like
this.
B
This
person
is
gonna,
kill
themselves
or
this
person
is,
you
know,
not
thoughtful
about
gender,
or
this
person
is
a
racist
and
and
all
this
stuff,
and
it
starts
it
really
catches
fire
really
quickly,
and
you
really
can't
it's
hard
to
get
to
the
truth,
because
because
it
there's
so
much
so
fast
of
an
uptake
around
this,
so
this
is
something
we
need
to
think
about.
We
need
to
monitor.
We
got
to
figure
out
again,
what's
true,
what
do
I?
Actually?
B
What
can
I
actually
figure
out
is
true
and
reliable,
and
what
do
I
take?
What
sort
of
steps
do
I
take
to
act
on
that
information
and
where
does
the
influence
come
in?
So
it's
something
we
need
to
think
about
and
again
I'd
talk
about,
code
versus
pepsi
and
all
this,
but
one
of
the
sims
that
I
have
here
and
if
we
have
time
I
can
walk
through
it
is
okay.
Let's
look
at
modeling.
B
This
simulation
of
you
know
competing
narratives
like
what
is
we're
saying
here
in
north
carolina
about
mask
wearing
and
social
distancing
and
there's
an
alternative
narrative
saying
that's
an
infringement
on
your
freedoms.
It's
not
necessary.
Cobit
is
a
hoax,
all
that
and
if
you
can
figure
out
in
your
if
you're
a
community
leader
or
a
business
leader,
how
many
of
each
kinds
of
those
people
do
I
have
in
my
community.
I
can
then
start
figuring
out
all
right.
What
do
I
need
to
say
to
them?
Who
are
their
leaders
and
influencers
in
the
community?
B
How
do
I
get
those
people
one
over
to
this
different
position
so
that
we
can
get
behavior
that
because
I
can
see
like
if
one
of
these
wins
over
another?
I
can
actually
take
this
right
into
this
simulation
model
and
predict
how
many
infections
are
going
to
happen
over
what
amount
of
time
whether
or
not
people
are
going
to
move
around
freely
or
try
to
stay
at
home.
B
How
does
that
affect
you
know?
How
does
this
get
exacerbated
by
the
number
of
beds
I
have
available
in
the
healthcare
personnel?
I
have
how
it
accelerates,
and
so
these
now
making
these
data-driven
models
driven
by
human
behavior
models,
not
just
the
epidemiology
of
the
virus
and
its
infection
rate,
which
is
what
most
people
have
been
modeling.
So
far,
we
have
a
grant
here,
working
with
unc
on
this
sort
of
model,
where
we
can
let
every
county
in
the
in
north
carolina
figure
out
all
right.
What
people
do
you
have
here?
B
What
industries
do
you
have
here?
What
kind
of
health
conditions
do
you
have?
What
amount
of
mobility
and
interconnectedness
with
other
counties
do
you
have?
Who
are
the
community
leaders?
What
is
the
competing
narrative
and
what
do
people
think
and
believe
and
feel,
and
how
do
we
create
policies
and
communications
that
influence
that
behavior
and
that's?
B
I
want
to
pause
there
and
make
sure
we
have
time
for
for
questions,
but
that's
what
I'd
like
this
group
to
think
about
with
each
of
the
perspectives,
which
are
each
worth
a
iq
points,
at
least
for
people
who
are
already
smart
and
you
know
and
how
you
know.
What
do
you
think
about
everything
that
we
just
discussed.
B
A
B
That's
a
very
active,
that's
an
active
thing
at
the
government
level
is
figuring
out
where
the
influencers
are
coming
from
and
and
then
try
to
determine.
What's
true
and
what
is
it
now?
One
of
the
things
that
we
did
is
I
built
early
on
just
under
my
own
funding,
a
covid
brain,
so
a
coronavirus
brain
that
takes
in
you
know
thousands
of
scientific
journal
articles
every
day
and
really
tries
to
apply
this
algorithm
called
the
truth.
Triad
to
you
know,
what's
the
source
of
it,
what
are
the?
What
are
its
statements?
B
How
likely
are
those
statements
to
be
true?
How
many
people
corroborate
this
statement
are
those
reliable
sources
and
then
give
it
a
score,
so
you
can
figure
out
like
okay.
That
needs
to
go
in
this
locker
over
here.
This
idea
that
hydro
chloroquine
is
a
useful
treatment
for
covid
until
it
gets
more
wide.
You
know
it
goes
to
the
scientific
process
of
peer
review
and
acceptance,
testing
experimentation
where
yeah,
we
think
that's
true,
so
that
that
helps
leaders
again.
B
Take
that
first
step
of
what
can
I
you
know:
what's
the
information
coming
who's
it
coming
from?
How
likely
is
it
to
be
true?
How
did
it
enter
my
organization
who
championed
it
who
challenged
it?
How
do
I
make
decisions
around
it?
That's
one
of
the
things
you
can
put
this
really
powerful
technology
into
the
service
of,
in
addition
to
okay.
Now
I
have
this
information.
How
do
I
use
that
to
communicate?
B
How
do
I
communicate
that
to
the
people
who
work
in
my
organization
at
my
university
in
my
community
who
are
their
influencers
and
sources
where
do
they?
Where
are
their
brains
right
now?
In
this
you
know
sort
of
spectrum
of
belief
systems.
How
do
I
move
them
from
here
over
closer
to
the
gravitational
pull
of
what
we
think
is
better
for
my
organization
or
my
community
or
my
city?
B
A
Okay,
we
have
a
lot
of
questions
on
the
chat.
I
think
we
can
start
from
the
top.
Thank
you
richard.
By
the
way,
this
was
very,
very
thought-provoking
going
to
the
list
on
the
chat,
and
I
think
the
first
question
was
from
jeffrey
crolick.
How
are
you
able
to
measure
the
accuracy
of
your
persona
models.
B
B
We
do
have
case
studies
on
that.
None
of
the
banks
will
ever
talk
about
it.
We
can't
even
say
who
the
banks
are
some
of
the
bank
stuff
that
we've
done.
They
call
it
the
customer
experience
simulator
because
they
have
such
strict
privacy
rules,
and
I
hope
this
gives
you
some
comfort,
the
machine
that
we
put
these
systems
on
for
them,
they're,
where
they're
taking.
If
you
want
to
know
someone
who
really
knows
a
lot
about
you,
it
isn't
necessarily
amazon
or
google.
It's
your
bank!
You
can't
lie
to
your
bank.
B
Your
bank
knows
everything
you
do
with
your
money,
so
they
know
everywhere.
You
go
so
there's
lots
of
data
there.
So
there
and
again
what
they're
trying
to
do
is
is
look
at.
You
know.
Why
am
I
sending
something
about
a
mortgage
to
someone
you
know,
mortgage
services
is
someone
who's
a
renter?
I
really
want
to
put
more
personalized
the
interaction
and
the
services
that
we're
offering
to
our
clientele.
So
so
they
have
a
machine,
that's
not
even
connected
to
their
own
network
and
they've
got
rules.
B
It's
almost
like
a
government
skiff
kind
of
environment
where
the
machine
is
in
a
closed,
locked
room
that
cannot.
You
can't
go
in
there
by
yourself
and
access
it.
It's
got
to
be
you
plus
an
observer.
That's
that's
the
controls
they
have
in
place
because
they're
so
concerned
about
violating
privacy
rules
and
then
that's
good,
but
the
outcomes
we're
seeing
there
are
are
typically
like
we've
seen
results
in
the
the
roi
we're
seeing
or
the
time
to
insight
has
been.
B
You
know
up
to
600
times
faster.
You
know
when
they,
when
they
used
to
wait
till
the
end
of
a
quarter
or
the
end
of
the
month,
look
back
over
a
whole
bunch
of
data
and
try
to
figure
out.
You
know
were
people
having
difficulty
with
online
banking
or
how
to
order
checks
or
whatever
now
they
can
do
it
instantly,
because
we
have
systems
that
are
reading.
B
All
the
chatter
looking
on
social
media
scraping
that
trying
to
understand
like
how
are
people
feeling
about
this
new
thing
that
we've
rolled
out
or
this
product
or
that
product
and
then
how
do
you
more?
Finally,
how
do
you
get
more
responsive
to
your
customers
around
those
ideas,
so
the
measurements
on
that
have
been
been
fantastic?
B
The
chicken
example
I
gave
you
is
one
that
we
use
typically
as
a
case
study,
although
we
can't
say
the
name
of
the
company
either
most
people
are
pretty
nervous
about
it,
but
the
idea
that
you
know
the
the
other
work
that
they've
done
in
market
research
indicated
that
you
know
there
should
be
a
certain
distribution
of
of
of
of
organic
foods
on
the
you
know
in
the
counter
and
that
the
actual
transaction
data
told
a
different
story.
Therefore,
you
know
they
could
ignore
the
other.
A
A
I
I'm
on
the
board
of
a
fintech
company
that
is
integrating
empathy
engine
in
the
behavioral
model
of
customers
to
better
predict
the
survivability
of
the
financial
product,
the
feasibility
and
marketability,
but
the
data
we
gathered
from
social
media
to
develop
these
behavioral
models
always
reveal
a
fickle
nature
of
individuals,
they're
very
dynamic,
though
it's
very
difficult
to
nail
them
down,
so
so
that
basically
is
very
challenging
to
create
a
reliable
empathy
engine
that
can
help
in
developing
a
better
behavioral
model.
So,
should
we
consider
this
as
a
bug
or
a
feature.
B
Experience
is
you
know,
people
aren't
just
and
that's
the
problem
with
segmentation
model
studies
today
as
well.
If
you
go
if
you're
a
big
and
we,
this
is
a
real
example
of
a
technology
company
that
engaged
with
mckinsey
do
a
segmentation
study.
It
was
a
seven
figure.
Engagement.
More
than
a
million
dollars
yielded
about
nine
to
ten
segmentation
models
like
iphone
mom.
You
know
iphone
mom
controls,
85
percent
of
household
spending
and
shops
here
and
drives
a
volvo
and
all
that
and
it's
just
something
in
a
powerpoint
deck
or
on
us.
B
You
know
in
a
document
that's
filed
away
and
then,
when
they
go,
do
certain
go
try
to
design
a
new
marketing
campaign
or
new
product.
They
pull
this
stuff
out
and
read
it
and
then
do
it.
But
that
was
you
know.
That
study
was
done
at
this
point
in
time
and
it's
just
like
frozen
model
of
a
person.
That's
not
how
people
are,
and
first
of
all,
your
social
media
interest
graph
if
you've
ever
looked
at.
B
Facebook
might
think.
Well,
richard
really
loves
cats,
and
I
don't
and
I'm
liking
it
for
a
variety
of
other
reasons.
So
there's
there's
some
data
problems
with
social
media
data
in
general.
But
what's
really
good
is
what,
when
people
have
a
choice,
what
do
they
do
with
their
attention?
That's
why
the
google
search
and
google
and
just
the
traffic
data
on
the
internet
is
important
or
nielsen
watch
data
and
what
do
they
do
with
their
money?
B
That's
who
they
really
are
not
necessarily
what
they
would
say
in
a
survey
and
social
media
models
are
incomplete,
they're
a
part
of
it
even
location
services,
even
though
it's
deep
data
and
the
banking
data
is
really
deep,
it's
incomplete.
So
if
you
remember
that
screen
that
I
had
there,
it
showed
a
data
data,
fusion
practice
of
census,
data,
behavioral,
study,
data,
nielsen
data,
location
services,
social
media
and
then
you
know
any
choice,
model
information
I
have
which
really
comes
from
transact
transaction
data,
and
I
still
say
that
gets
me
to
80.
B
But
it's
way
better
than
that
mckinsey
million
dollar
study
and
by
the
way
it
only
costs
a
fraction
of
that
too.
So
so
that's.
What
I
would
say
is
there's
a
there's,
an
ongoing
data,
fusion
exercise
that
you
need
to
do
frequently,
not
just
once
every
18
months
or
two
years
and
I'll.
Just
add
that
you
know
we
did
some
work
for
that,
that
banking
group
in
in
the
uk,
and
then
they
were
faced
with
okay.
We
have
this
new
concept
of
brexit,
but
it's
not
in
any
of
our
study
information.
B
It's
like
yeah,
because
that
was
a
long
time
ago
now,
like
any
studies
that
you
did
prior
to
the
pandemic,
are
probably
not
very
valid.
So
you
need.
You
need
more
current
information
about
how
people
and
what
the
way
we
talk
about.
This
in
gaming
is
in
a
hierarchical.
Behavior
engine
is
that
okay,
this
person
was
here,
but
now
they're
under
stress,
so
they're
in
a
different
mode.
So
now
they
change
state.
B
So
you
need
models
that
can
actually
account
for
that,
but
again,
if
you're,
only
looking
at
broad
sections
of
the
population,
some
of
that
data
just
filters
out
anyway.
I
hope
that
that
was
a
long
answer
to
your
question,
but
it's
basically
a
it's
a
practice
that
exists
in
the
dod
around
constantly
validating
and
verifying
the
data.
That's
going
into
these
intelligence
models
and
how
predictive
are
they
and
in
marketing
today,
for
example,
or
in
a
lot
of
other
hr
and
everything
it's
like.
B
A
You
richard
that's,
definitely
helpful.
Thank
you.
Sorry,
a
lot
of
questions
coming
in
the
chat,
but
I'm
also
looking
at
the
time.
So
we
are
past
the
the
hour.
Oh
no,
we
can.
We
can
throw
some
of
those
questions
at
you
if
you
have
time,
yeah,
yeah,
absolutely
absolutely
okay,
so
we
have
another
question
from
matthew:
can
you
train
an
ai
to
pursue
an
objective
truth.
B
That's
a
great
philosophical
question:
you
know.
First,
we
have
to
just
define
what
that
is
right,
but
there
is
like
I
said
earlier
as
we
filter
through
coronavirus
information.
Looking
for,
what's
likely
to
be
true
or
what
we
call
truthiness,
there
is
a
set
of
algorithms.
We
call
the
truth,
triad
that
that
try
to
solve
for
that
and
and
give
a
score
for
how
likely
something
is
to
be
true.
So
in
fact,
the
covid
brain
that
we
built
is
searching
and
sorting
for
that.
B
You
know
here's
stuff,
that
that
is
unproven
and,
unlike
so
it's
an
outrageous
claim
from
an
unreliable
source.
It
goes
over
here
in
this
bin.
Here's
a
bunch
of
sources
that
are
that
are
prove
that
my
human
partners
keep
telling
me
are
really
reliable
and
and
they're
making
statements
that
have
been
corroborated
by
all
these
other
sources,
so
the
truthiness
score
is
much
higher.
So
I'm
going
to
spend
my
time
recommending
that
to
my
human
overlords,
as
opposed
to
you
know,
saying:
oh,
this
has
equal
weight.
B
You
know
this,
this
denial
of
climate
change,
or
this
this
hydroclaxic
or
whatever
it
is
hydrochloroquine.
Promoting
thing
has
equal
weight
to
what
this
other
group
says.
It's
like
no,
these
are
not
equally
weighted.
This
has
some
scientific
consensus
from
experts.
This
is
coming
from
a
website
that
looks
like
it
was
created
yesterday
in
ukraine
and
has
an
outright
outrageous
claim.
B
Yeah,
I
I
I'll
have
to
find
it
and
see
you
know
one
of
the
head
thing.
I
was
talking
to
someone
earlier
today
about
how
you
know
I
I've
worked
with
a
number
of
intelligence
agencies
and
folks,
but
when
I
was
in
my
lockheed
dent
and
one
of
the
everyone
always
talks
about
the
cia,
for
example,
but
you
know
they're
what
happened
because
of
you
know
waterboarding,
a
lot
of
those
things
and
a
lot
of
those
issues
that
happen.
You
know
during
iraq
and
afghanistan.
B
A
lot
of
these.
A
lot
of
this
work
goes
out
to
these
other
agencies
that
most
people
don't
even
know
about,
and
that's
where
a
lot
of
the
actual
work's
going
ccios
has
been
set
up.
It
wouldn't
surprise
me
at
all-
and
I
didn't
look
today
if
they've
changed
names
and,
and
I'm
not
sure
also
under
this
administration,
how
much
support
they're
getting
as
I've
watched,
dhs
and
some
of
these
groups
get
get
plundered
a
bit,
but
I'll
see
what
I
can
find
and
get
it
back
to
the
group.
Thank
you.
C
So
you
you're
so
so.
The
overall
focus
that
you're
you're
pursuing
here
is
based
on
using
data
exhaust
to
build
up
various
specific
psychological
levels
in
a
given
profile
and
apply
that
at
a
large
scale,
for
your
empathy,
engine
idea
and
you're
applying
that
right
now
to
the
idea
of
evaluating
the
truth
or
falsehood
of
particular
news
and
narratives
that
are
out
in
the
overall
information
stream
based
on
on
just
like
the
a
focus
on
whether
information
is
true
or
false.
C
But
getting
back
to
what
you
were
talking
about
earlier,
with,
with
narratives
and
and
the
role
of
narratives
and
information
warfare.
My
understanding
from
reading
a
lot
of
very
conspiratorial
scientific
literature
in
the
russian
space
is
that
their
model
doesn't
necessarily
depend
on
just
like
outright
feeding
like
fake
information
or
disinformation
to
anybody,
but
just
using
an
information
as
a
weapon,
regardless
of
its
truth,
truth
or
falseness,
but
using
information
so
as
to
change
relationships
and
reconfigure.
C
Community
connections
and
institutions
and
and
a
variety
of
of
other
particular
scales
of
effects.
Building
up
for
large-scale
political
warfare
and
subversive
political
outcomes.
Right
that
is
not
dependent
on
merely
spreading
truth
or
spreading
falsehood
or
propaganda,
but
focusing
on
trying
to
say,
alter
a
given
community's
ecology
of
moral
values,
as
well
as
their
their
like
common
knowledge
and
trying
to
break
that's.
C
B
C
Yours
focuses
most
on
the
truthiness
aspect
right
now,
but
is
it?
Is
it
feasible,
within
this
empathy
engine
model
to
also
be
focusing
on
or
evaluating
changes
to
say,
like
the
maybe
not
maybe
not
demographic,
but
but
certainly
the
ecology
of
moral
values
and
common
shared
knowledge
and
body
of
facts
and
other
aspects
of
narrative
in
the
broad
sense,
beyond
just
just
just
facts
and
consensus
knowledge
itself.
B
Yeah
yeah,
if
I
wasn't
clear
there
is
you
know,
I'm
very
interested
in
the
truth
versus
true
versus
false,
but
there's
also
just
the
idea
that
I
was
trying
to
share
something
quickly.
Like
you
know,
there's
the
idea
of
coke
versus
pepsi
coke
is
true.
Pepsi
is
true.
Their
claims
probably
are
equally
true,
but
it's
a
matter
of
influence
right,
so
each
are
making
claims
and-
and
you
know,
if
you're,
a
coke
executive
and
you're
trying
to
promote
the
idea
that
more
people
should
drink
coke
than
pepsi,
then
you
know
it's
about.
B
As
you
rightly
pointed
out,
what
governs
someone's
behavior
sometimes
is
how
often
they
you
know
how
many
times
a
day.
Do
I
see
that
message
and
who's
saying
it
and
therefore
does
that
have
more
influence
on
me
than
you
know
what
the
position
that
I
started
at,
which
is,
I
only
drink
coca-cola,
but
if
I
never
hear
anything
from
coca-cola
and
I
over
time
I
constantly
hear
you
know
everyone
around
me
is
drinking
pepsi
people
I
like
and
idolize.
B
Like
my
you
know,
taylor,
swift
or
whoever
are
drinking
pepsi
and
telling
me
to
drink
pepsi,
I'm
more
likely
to
start
migrating
over
to
the
pepsi
position,
and
it's
not
about
true
versus
false.
It's
there's
competing
narratives
that
both
might
have
equal
merit
in
this
case.
But
it's
it's
like
you
said
repetitive,
you
know,
repetition
and
exposure
might
move
me
from
one
position
to
another.