total: 813, more: 75, less: 48 | diff: 28, punc: 102, sfx: 9, case: 37 [WER: 0.20]
—
Artificial
Visual
intelligence
intelligence
is
is
said
said
to
to
have
have
huge
huge
impacts
impacts
on
our
our
own
lives
lives
and
and
even
even
society.
society's
And
and
some
some
actually
actually
say
say
it's
it's
the
the
new
new
electricity.
electricity
Since
since
the
the
fifties,
fifties
computer
computer
scientists
scientists
have
have
been
been
trying
trying
to
to
program
program
computers
computers
to
to
behave
behave
like
like
humans.
humans
So
so
let's
let's
say
say
for
for
example
example
you
you
want
want
to
to
teach
teach
a
a
computer
computer
what
what
a
cat
is.
it
Well.
kept
How
its
hard
but
can
can
that
that
be,
be
right?
right
I
i
mean
mean
after
after
all,
all
a
a
cat
cat
is
is
just
just
a
a
collection
collection
of
of
shapes
shapes
and
and
colors.
colors
and
So,
also
computer
computers
that
scientists
insists
just
just
thought
thought
okay
okay
we'll
will
just
just
give
give
rules
rules
to
of
the
the
computer
computer
and
and
it'll
still
be
be
able
able
to
to
manage
manage
what
s
well,
and
to
to
recognize
recognise
what
what
a
a
cat
cat
is.
is
So,
so
used
we
we
should
should
just
just
tell
tell
it
through
the
the
computer,
computer
in
in
mathematics
mathematical
mathematical
language
language
of
of
course,
course
that
that
a
a
cat
cat
has
has
a
a
round
round
face,
face
two
two
pointy
point
eight
ears,
years
a
of
chubby
chevy
body,
body
and
and
a
a
long
long
tail.
tail
But.
what
about
about
this
this
cat?
kept
And,
and
what
what
about
about
if
if
the
the
cat
cat
is
is
hidden?
hidden
That
that
doesn't
doesn't
work.
work
Giving
giving
rules
rules
to
to
computers
computer
will
will
only
only
get
get
you
you
so
so
far.
far
So
so
another...
another
another
another
type
type
of...
of
this
is
another
another
type
type
of
of
approach
approach
was
was
developed
developed
and
and
this
new
you
type
type
of
of
approach
approach
developed
development
was
is
called
called
machine
machine
learning.
learning
Machine
machine
learning
learning
gives
gets
computers
computers
the
the
ability
ability
to
to
learn,
learn
without
without
being
being
explicitly
explicitly
programmed.
programmed
And
so,
so
rather
rather
than
than
hand
hand
coding
coding
specific
specific
sets
sets
of
of
instructions,
instructions
the
the
machine
machine
is
is
trained,
trained
using
using
large
large
amounts
amounts
of
of
data.
data
About
about
five
five
years
years
ago,
ago
a
a
specific
specific
type
type
of
of
algorithm
algorithm
was,
was
of
a
machine
machine
learning
learning
algorithm
algorithm
was
was
developed,
developed
called
called
deep
deep
learning.
learning
You
you
might
might
have
have
heard
heard
of
of
deep
deep
learning
learning
algorithms,
algorithms
that
that
they
were
were
used
used
to
to
beat
beat
humans
humans
playing
plane
go,
go
or
or
a
recently
recently
poker.
poker
and
Deep
deep
learning
learning
uses
uses
techniques
techniques
that
are,
there
let's
might
say
say
loosely
loosely
inspired
inspired
from
from
how
how
our
our
brain
brain
works.
works
And,
and
we've
we
had
verve
seen,
seeing
very
very
extremely,
extremely
where
well,
the
extremely
extremely
impressive
impressive
results
results
because
because
they
they
have,
have
a
a
unique
unique
ability
ability
to
to
identify
identify
patterns
patterns
from
from
data
debt
and
and
learn
learn
from
from
it.
it
So
so
this
this
means
means
that
that
these
these
new
new
types
types
of
of
algorithms
algorithms
can...
can
can
do...
do
it
let's
let's
say
say
three
three
main,
main
can
can
solve
solve
three
the
main
main
problems.
problems
So,
so
fed
faced
with
with
data,
are
these
these
algorithms
algorithms
can
can
generate
generate
new
new
stuff.
stuff
So
so
for
for
example
example
if
if
you
you
train
train
a
a
machine
machine
a
a
machine
machine
learning
learning
algorithm,
algorithm
with
with
a
a
set
set
of
of
artist
artists
painting,
painting
like
like
Picasso.
picasso
You
you
feed
feet
of
a
machine
machine
learning
learning
algorithm
algorithm
with
with
Picasso
picasso
paintings,
paintings
he
he
will
will
be
be
able
able
to
to
reproduce,
reproduce
but
that
style,
style
and
and
apply
a
plan
on
on
a
a
new
new
painting,
painting
and
and
actually,
actually
we
we
talked
talked
about
about
Prisma
christmas
very
very
briefly
briefly
earlier,
earlier
priest
Prisma
my
uses
uses
machine
machine
learning
learning
algorithms
algorithms
to
to
do
do
just
just
that.
that
So,
so
you
it
can
can
generate
generate
new
new
images,
images
you
we
can
can
generate
generate
for
for
example
example
new
new
texts,
text
it's
it's
used
used
to
to
do
do
translations.
translations
It
it
can
can
also
also
do...
do
not
that
Its
their
ability
ability
to
to
recognize
recognize
patterns
pattern
is
is
also
also
very
very
useful,
useful
for
for
predictions,
predictions
to
to
make
make
predictions.
predictions
I
i
don't
don't
know
if
if
you've
done
dent
those
those
types
type
of
of
tests,
test
you
you
know,
know
kind
of
cut
IQ
i
q
tests,
tests
or
or
well,
well
this
this
type
type
of
of
tests
where
they
this
well,
further
they
they
give
gave
you
you
a
a
set
set
of
of
of
of
numbers,
numbers
or
or
shapes
shapes
and
and
then
then
you're
your
supposed
supposed
to
to
find
find
the
next's
the
the
next
next
one.
one
So
so
basically,
basic
what
what
you're
you
doing
doing
there
there
is
is
identifying
identifying
patterns,
patterns
and
and
then
then
predict
predict
what's
what's
coming
coming
next,
next
and
and
computers,
computers
and
and
machine
machine
learning
learning
algorithms,
algorithms
are
are
extremely
extremely
good
good
at
at
doing
doing
that.
that
So,
so
that's
that's
very
very
useful,
useful
for
for
example
example
in
in
the
the
business
business
world.
world
If
if
you
you
train
train
an
an
algorithm
algorithm
like
like
that
that
on
on
historical
historical
sales
sales
data,
data
it
it
can
can
help
help
you
you
forecast
forecast
the
the
demand
demand
of
of
a
a
product.
product
These
these
algorithms
algorithms
also
also
very
very
good
good
and
at
music
making
making
associations.
associations
So
so
I'm
i'm
going
going
to
to
share
share
with
with
you
you
an
an
example
example
but
that
I
i
really
really
like.
like
so
A
a
machine
machine
learning
learning
algorithm
algorithm
was
was
asked
asked
to
to
find
find
associations
associations
between
between
between
between
-
sorry
sorry
-
between
between
art
artworks
works
from
from
the...
britain
from
from
the
the
Tate's
tate's
collection
collection
of
of
British
british
art
art
And
and
Reuters
images
images
from
from
two
thousand
2016.
sixteen
and
and
here,
here
is
is
what
what
the
algorithm
algorithm
gave
gave
us.
us
Associations.
associations
I
i
think
think
it's
is
quite
quite
interesting.
interesting
With
with
this
this
work,
art
work,
what
he
approach
typed
type
this
this
photograph
photograph
from
from
Reuters.
reuters
With
this
one,
he
he's
associated
associated
with
this
this
one.
win
So
so
that's
that's
kind
kind
of
of
fun
fun
but
but
translated
translated
in
the
trend
in
in
the
the
business
business
world,
world
these
these
algorithms
algorithms
can
can
use,
use
mean
the
dvd
association
association
ability
ability
of
of
these
these
algorithms
algorithms
can
can
use
use
to
to
segment
segment
-
can
can
be
be
used
used
to
to
segment
segment
consumers
consumers
according
according
to
to
similar
similar
behaviors,
behaviors
for
for
example.
example
Advances
advances
in
in
a
machine
machine
learning
learning
are
are
not
not
only
only
do
it
due
thanks
to
to
the
the
advances
advances
in
in
the
the
algorithms
algorithms
themselves.
themselves
These
these
advances
advances
are
are
due
due
to
to
two
two
other
other
factors
factors
that
that
concurred
concurred
at
at
the
the
same
same
time.
time
That
that
developed
developed
at
the
the
same
same
time.
time
when
One
it's
is
picked
"big
up
data".
so
With
with
internet
internet
we've
we've
been
been
producing
producing
lots
lots
and
and
lots
lots
of
of
data.
data
And
and
this
this
data
that
can
can
be
be
used
used
to
to
train
train
those
those
algorithms.
algorithms
To
to
say
say
the
the
truth,
truth
machine
machine
learning
learning
algorithms
algorithms
or
were
the
already
idea
idea
of
of
machine
machine
learning
learning
algorithms,
algorithms
is
he
i
mean
has
has
been
been
developed
developed
for
for
a
a
couple
couple
of
of
years
years
and
or
in
even
the
decades,
decades
but
but
we
we
were
were
lacking
lucky
the
that
data
that
to
to
feed
feed
them
them
enough.
enough
So,
so
that
that
was
was
a
a
big
big
thing
thing
to
do
to
to
help
help
these
these
algorithms
algorithms
takeoff
take
off
and
and
the
the
second
second
thing
thing
he
needed
needed
was,
was
computer
computer
power,
power
which
which
now
now
we
we
have.
have
So
so
basically,
basically
today
today
AI
takes
takes
off
off
because
because
the
the
stars
stars
have
have
aligned.
aligned