00:00:07
Oftentimes, innovations
solve practical problems,
00:00:11
but the advancement of A.I.
00:00:12
might bring new tools
00:00:14
to chip away at the larger,
even existential questions.
00:00:18
Are we alone in the universe?
00:00:21
Can we create lifelike,
intelligent machines?
00:00:24
Maybe they're all moonshots,
but imagine, one day,
00:00:27
having a second,
synthetic version of you.
00:00:32
-How's it going, brother?
-Oh, not bad.
00:00:34
Just spent the last hour
mapping half the cosmos.
00:00:37
I'm looking for
a constellation
to name after us.
00:00:39
You mean "me," yeah?
00:00:41
Whatever. Semantics.
00:00:44
I'm doing all the work.
00:00:46
Touchy.
00:00:51
The starry night sky has been
a source of fascination
00:00:54
and curiosity for centuries.
00:00:58
Is there something out there?
00:01:00
We've got all these suspect
places to look for life
00:01:03
in our own solar system.
00:01:04
And we're just
one little solar system
00:01:06
in a large galaxy,
00:01:08
which is one of many,
many galaxies in the universe.
00:01:11
And so you realize
pretty quickly
00:01:13
the chances of life elsewhere
are pretty high.
00:01:25
[tuning radio]
00:01:27
[man] ...we hope we have a
number of listeners out there.
00:01:29
Most of you are probably
soft and squishy humanoids.
00:01:31
In case any artificial
intelligence is listening,
welcome as well.
00:01:35
[Bill Diamond]
You'll appreciate this,
00:01:36
being a data scientist,
00:01:37
you know we're generating
about 54 terabytes of data
00:01:40
every day, so...
00:01:42
See, that's music to my ears,
right there.
00:01:43
[Diamond]
That's music to your ears.
00:01:44
That's a nice playground
for your algorithms.
00:01:46
[Downey]
In remote northern California,
00:01:48
two scientists are on their way
to collect data
00:01:50
in hopes to answer
a cosmic question...
00:01:54
one that's as old
as humankind itself,
00:01:57
or at least, Galileo.
00:01:59
[Graham Mackintosh]
When it comes to the search
00:02:00
for extraterrestrial
intelligence...
00:02:01
[Diamond] Right.
00:02:03
...there is decades
of scientific discovery
00:02:06
and progress
00:02:07
which is relentlessly telling us
00:02:08
life is more likely
than we thought.
00:02:11
Yeah, the body of evidence
is becoming--
00:02:13
That's right.
00:02:14
...overwhelming,
but can we find it?
00:02:16
[Mackintosh]
When I was ten years old,
00:02:17
I was determined
to have my own computer,
00:02:19
and I found out there was
a kit that you could buy
00:02:23
and put together for yourself,
00:02:24
so I earned enough money
to do that,
00:02:26
and that got me hooked.
00:02:27
I've been obsessed
with computers ever since.
00:02:32
And I hope, I believe, that
A.I. can help us dig deeper,
00:02:35
and hopefully come to the answer
we're looking for.
00:02:39
Is there life beyond the Earth?
00:02:46
[Diamond] Ever since
humans have been able
to gaze up at the sky
00:02:48
and look at the stars,
00:02:50
we've wondered,
00:02:51
"Are we alone?
00:02:52
Is this the only place
where life has occurred?"
00:02:56
The SETI institute is trying
to answer this question.
00:03:05
SETI institute was founded
00:03:06
by Frank Drake, and Jill Tarter,
00:03:09
and Carl Sagan.
00:03:11
I co-founded this institute
back in 1984
00:03:14
as a way to save NASA money.
00:03:16
...see if we can backtrack
00:03:17
to see if we can figure out
what's venting...
00:03:19
Since then,
00:03:21
it has grown far beyond
any of my expectations.
00:03:25
We have nearly
80 PhD scientists here.
00:03:27
Our research really starts
with, "How does life happen?"
00:03:33
What are the conditions
under which life takes hold?
00:03:40
We're trying to understand
that transition
00:03:42
of how the universe
00:03:43
and how our own galaxy
and solar system
00:03:45
went from chemistry to biology.
00:03:49
The number of civilizations
00:03:51
that there might be
in the galaxy
00:03:52
is of the order of a million.
00:03:54
[Downey] Carl Sagan
helped bring the cosmos
00:03:56
down to Earth,
00:03:57
but he wasn't the first
to popularize it.
00:03:59
Ever since Orson Welles
scared our pants off
00:04:02
with War Of The Worlds,
00:04:03
pop culture has had its eyes
on the skies.
00:04:06
Little green men,
extraterrestrials,
00:04:09
contact with aliens continues
to capture our imagination.
00:04:13
[Diamond] We're interested
in all kinds of life,
00:04:16
but of course we have
a special interest
00:04:18
in intelligent or technological
life beyond Earth,
00:04:21
hence, SETI.
00:04:25
Hello, this is Seth Shostak
speaking to you
00:04:28
from Big Picture Science.
00:04:29
Today we're going to talk
about artificial intelligence.
00:04:32
The machines of today
are a lot smarter, if you will,
00:04:35
at least more capable,
00:04:36
than the machines
of 50 years ago, incredibly...
00:04:39
There's vast amounts of data
coming from space,
00:04:41
and A.I. can, um...
00:04:43
allows us to understand
that data better
00:04:45
than we have been able to
in the past.
00:04:47
It's this new capacity we have
to see patterns in data...
00:04:56
[Tarter] We are trying
to find evidence
00:04:58
of somebody else's
technology out there.
00:05:01
We can't define intelligence,
00:05:03
but we're using technology
as a proxy,
00:05:06
so if we find some technology,
00:05:07
something that's engineered,
00:05:09
something that nature didn't do,
00:05:10
then we're going to infer
00:05:12
that at least
at some point in time,
00:05:14
there were
some intelligent technologists
00:05:16
who were responsible.
00:05:17
[Diamond] So, Graham,
we call it Area 52.
00:05:21
[chuckling]
00:05:23
[Mackintosh] We are headed
to the Allen Telescope Array,
00:05:25
and tonight we are going
to be doing an observation
00:05:27
which really
is looking for signs
of extraterrestrial life,
00:05:31
and we're gonna be
using A.I. models
00:05:33
in a way that's
never been done before.
00:05:35
[Diamond] All right,
we are good to go.
00:05:40
So I gotta turn
my cell phone off,
no Bluetooth, nothing?
00:05:43
Nope, we need to be
in a place that is radio quiet,
00:05:46
so you don't have interference,
00:05:48
or at least,
you minimize interference.
00:05:50
We're gonna come around
another bend a little up ahead,
00:05:53
and you'll see the dishes.
00:05:57
[Mackintosh exclaiming] Oh!
00:05:58
[Diamond] There we are.
00:06:01
Welcome
to the Allen Telescope Array.
00:06:06
[Downey] The sole mission
of the Allen Telescope Array,
00:06:09
or A.T.A.,
00:06:10
is to search
for extraterrestrial life.
00:06:13
Past telescopes
were basically toy binoculars
00:06:16
compared to the A.T.A.,
00:06:17
which was built in 2007
00:06:19
with support from
Microsoft's Paul Allen.
00:06:22
Part of what makes it
light-years ahead
00:06:24
is its wider field of view,
00:06:26
and ability to capture
a greater range of frequencies.
00:06:30
It's also an array,
00:06:32
which basically means
00:06:33
it's a group
of many small dishes
00:06:34
working together
to cover more ground,
00:06:36
or sky.
00:06:40
Welcome to the A.T.A.
00:06:42
Fantastic.
00:06:43
Okay. Looks like
Jon is out there.
00:06:45
I think he's manually
turning those dishes
to get 'em lined up.
00:06:48
[laughing]
00:06:52
-Hey, Jon.
-Jon!
00:06:54
-Good to see you, man!
-Good to see you, yeah!
00:06:57
My name is Jon Richards,
00:06:58
and I'm the Senior
Software Engineer
00:07:00
at the Allen Telescope Array.
00:07:02
Radio astronomy is similar
to optical astronomy,
00:07:06
except the radio wave
frequencies
00:07:08
are much lower than visual,
00:07:10
so to receive radio waves,
you need an antenna.
00:07:14
Take a look, Graham.
Under the bell jar,
00:07:16
you see the actual antenna
00:07:18
that's picking up the signals
coming from space.
00:07:21
This is spectacular.
00:07:22
[Diamond] It's kept
00:07:23
below the temperature
of liquid nitrogen.
00:07:25
That brings
the noise level down,
00:07:27
exactly what we want
for deep space observation.
00:07:29
Just amazing.
00:07:30
[Richards] The radio signals
from each one of these dishes
00:07:33
are brought
into our control room,
00:07:34
digitized, made into
binary ones and zeroes,
00:07:39
and combined together
00:07:40
to create the effect
of having one large dish,
00:07:43
so we can actually
map out the sky
00:07:46
much like you would
00:07:47
with a regular
optical telescope.
00:07:49
All right, let's head back.
00:07:51
Let's go.
00:07:54
The observation
we're gonna do tonight
00:07:56
is with the Trappist-1 system.
00:08:00
This is a star
that has planets circling,
00:08:03
and at 8:00 tonight,
00:08:04
two of those planets are gonna
align perfectly with Earth,
00:08:07
which makes it exactly
the right moment
00:08:10
to do an observation.
00:08:11
We're gonna be listening in
00:08:13
for signs of any kind
of communication
00:08:15
between these two planets,
00:08:17
even if that's not communication
directed at us.
00:08:21
[Diamond] We're counting down
to 8:01 p.m.,
00:08:23
which is when
the orientation of these planets
00:08:26
are going to be lined up
in our line of sight,
00:08:29
the so-called conjunction.
00:08:33
[Downey] It's a little like
an intergalactic stake-out.
00:08:35
The guys are waiting
00:08:36
till the two planets
are closest together,
00:08:38
and then plan to eavesdrop
on their conversation.
00:08:41
They have no idea
what they're listening for,
00:08:44
or if there's even gonna be
a conversation.
00:08:47
[Richards] So we can
take out this board here.
00:08:49
We're gonna repurpose it.
00:08:51
-So that's ready to go?
-Yeah, let's go put it in.
00:08:52
All right, let's get it in.
00:08:54
[Richards] Since the site's
getting close
to 20 years old now,
00:08:56
my job is to get all this data
coming in cleanly
00:09:00
and recorded cleanly,
00:09:02
and that is a challenge.
00:09:03
Here's the computer
which is sending all the data
00:09:06
that we receive
from all of our dishes
00:09:07
to our 48 terabytes
of data storage,
00:09:10
so we need to replace a card.
00:09:12
This card will control
our data storage.
00:09:15
[Mackintosh] You know, often
00:09:16
when people think of the search
for extraterrestrial life,
00:09:19
they're thinking of someone
with headphones
00:09:20
listening in on something
that is sent to us,
00:09:23
something that's obvious.
00:09:24
It's really not like that.
00:09:26
It's a lot more subtle,
00:09:27
and that's why we're
going to be collecting
00:09:29
enormous amounts of data.
00:09:31
All of the different parameters
we might have to explore
00:09:34
set that volume,
that exploration volume,
00:09:37
set it equal to the volume
of all the oceans on the Earth.
00:09:42
So how much have we done,
in 50 years?
00:09:45
Well, we've searched
one glass of water
00:09:49
from the Earth's oceans.
00:09:50
The technologies
that we've had to use until now
00:09:54
were not big enough,
not adequate to the job.
00:09:57
Okay.
00:09:58
[Mackintosh] That's why
we need computer systems
00:10:00
and artificial
intelligence systems
00:10:02
to really turn that search
on its head.
00:10:05
[Parr] When we think
about traditional software,
00:10:08
we think about human beings
writing lines of code.
00:10:10
What's extraordinary about A.I.
00:10:12
is that we're teaching machines
how to learn.
00:10:15
This is why
it's a quantum leap,
00:10:17
because for the first time,
00:10:18
instead of human beings
writing the software,
00:10:20
the computer's actually building
an understanding itself.
00:10:33
[Richards]
We have to keep in mind
00:10:35
that the Trappist-1 system
is 39.4 light-years...
00:10:39
39.6.
00:10:39
39.6 light-years away,
00:10:42
so this actual positioning
was 39.6 years ago.
00:10:46
So not only are we, uh,
00:10:48
are we doing
SETI research tonight,
00:10:50
we're time-traveling.
00:10:51
[Downey] That's right.
00:10:52
Because of how far away
these planets are,
00:10:55
and how long it takes
radio waves
00:10:57
to travel through space,
00:10:58
the guys are listening
to a conversation
00:11:00
from about 40 years ago.
00:11:03
Here's some perspective.
00:11:04
It takes about eight minutes
00:11:05
for radio waves to get
from here to the sun.
00:11:08
So, these planets?
00:11:10
Yeah, a little farther away.
00:11:12
[Diamond]
Over your shoulder, Graham,
00:11:14
there's a NASA illustration
of the Trappist system,
00:11:16
and there's at least three
rocky, Earth-like planets
00:11:19
where liquid water
can potentially be maintained--
00:11:22
Right.
00:11:23
...and that gives rise
to the possibility
00:11:25
that biology could have formed
in this system.
00:11:27
What's really interesting
00:11:29
about this particular
planetary system,
00:11:31
these planets
are very close together,
00:11:33
much closer than, for example,
Earth to Mars.
00:11:36
That means there could be
communication happening
00:11:39
between these planets,
00:11:41
and what we can
potentially do is listen in.
00:11:45
Not that we can have
a conversation
00:11:47
or understand
what they're, uh...
00:11:49
-[Mackintosh] We don't need to.
-We don't need to.
00:11:50
[Mackintosh] I love
this kind of observation
00:11:53
because it has
as its basic principle
00:11:56
something
that's really important.
00:11:58
It's not all about us.
00:12:00
No one's sending us a signal,
00:12:02
no one's trying
to get our attention.
00:12:04
The whole point
00:12:05
about the search for
extraterrestrial intelligence
00:12:07
is you don't even--
00:12:09
We don't know
what we're looking for.
00:12:10
Right, right.
00:12:11
Instead of looking
for something specific,
00:12:14
you have to look
for the exceptions
00:12:16
from what is normal.
00:12:17
That is where I think
00:12:19
A.I. is gonna just completely
change the game for SETI.
00:12:21
[Mackintosh]
Maybe it's communication,
00:12:23
maybe it's just a byproduct
00:12:26
of some technologically
advanced civilization
00:12:28
going about its business.
00:12:30
All we care about
00:12:31
is it doesn't look like
the rest of nature.
00:12:35
If it's a needle
in a haystack,
00:12:37
it doesn't look like hay.
00:12:39
It's like this, each one
of these little blips
00:12:41
is like a point in time
of radio power,
00:12:44
and we take different
points in time,
00:12:47
different windows into the data,
00:12:49
and we analyze them together
00:12:51
to see if there's
any kind of repetition,
00:12:54
anything at all
00:12:55
that might indicate
that something isn't random,
00:12:58
like this,
right in the middle here,
00:13:00
where the random dots
aren't random.
00:13:03
In a computer,
00:13:05
think of it like
a thousand of these sheets,
00:13:07
and it's moving them
a million times a second.
00:13:10
[Downey] To find order
in the randomness,
00:13:12
the A.I. picks a small area
00:13:14
and studies
its radio frequency data
00:13:16
to learn what normal
sounds like.
00:13:18
Then, it uses this info to
filter out background signals
00:13:21
from all the data
that's been collected.
00:13:23
What's left is any signal,
pattern, or repetition
00:13:27
that is unnatural.
00:13:39
They're coming up
to perfect alignment.
00:13:41
Conjunction now!
00:13:43
[Richards] We're recording.
00:13:45
[Diamond]
Wanna check the audio?
00:13:51
This is good.
00:13:57
This is good,
nice clean data.
00:13:59
Crispy clean.
00:14:01
[Richards] Silence.
00:14:02
Yeah, that's what we want.
I just, well--
00:14:06
I mean, we've been working up to
this for the last month.
00:14:08
It looks like,
it looks like nothing to us,
00:14:09
but that's the point.
00:14:10
[Diamond] That's the point.
00:14:11
That random sound
is music to my ears.
00:14:14
This picture here
is just immediate,
00:14:17
real-time results,
00:14:18
something that your normal
Allen Telescope Array
00:14:21
would discard as nothing.
00:14:23
Our point is, not so fast.
00:14:25
There could well be
more in there than we realize.
00:14:31
We do see
some little blip right here...
00:14:35
That's true,
in and around it.
00:14:37
Yeah.
So here, let's press...
00:14:41
So, now,
this all looks similar.
00:14:43
It's the sort of normal signal,
00:14:45
but that's interesting.
00:14:47
It just seems, I don't know--
00:14:48
It's like it spreads here
for some reason.
00:14:50
Well, I don't know
what that means.
00:14:51
It also is
a higher average power.
00:14:53
It is.
00:14:54
So, yeah, it's...
this is weird, right?
00:14:56
It is.
00:14:57
[Diamond]
There are a couple of things
00:14:59
that we are looking at
in the data
00:15:01
that look interesting.
00:15:02
Now, it's very subtle,
00:15:04
and this is why we'll need
machine-learning to extract
00:15:06
whether what we're seeing
is just something we're seeing,
00:15:08
or it's real,
a real phenomenon.
00:15:11
All right, so we are done
with the Trappist system.
00:15:14
[Mackintosh] This is great.
00:15:16
We've clearly grabbed good data.
00:15:17
It's exactly what we need.
00:15:19
[Downey] It's gonna
take Graham a few days
00:15:21
to analyze the data,
00:15:22
nothing compared
to what it used to take
00:15:24
to do manually.
00:15:25
[Pedro Domingos]
Some people think
00:15:27
that the emergence
of artificial intelligence
00:15:28
is the biggest event
on the planet since life,
00:15:32
because it's going to be
a change that is as big
00:15:34
as the emergence of life.
00:15:36
It will lead
to different kinds of life
00:15:37
that are very different
00:15:39
from the entire set of,
you know, DNA, carbon-based life
00:15:42
that we've had so far.
00:15:43
[Downey] While some
are ramping up the search
00:15:46
in outer space,
00:15:47
others are using A.I.
00:15:49
to further explore inner life.
00:15:59
[Suzanne Gildert]
In 20 to 30 years' time,
00:16:01
you might see a street
like this,
00:16:03
with humans
walking up and down it,
00:16:04
but there might also be
a new thing,
00:16:07
which is human-like robots
00:16:08
might be walking
up and down, too, with us.
00:16:12
Humans and robots
00:16:13
are really gonna be doing
the same kinds of things,
00:16:15
and some of the things
they'll be doing
00:16:17
will be maybe
superior to humans.
00:16:20
[Downey] Suzanne
is one of the founders
00:16:21
of Sanctuary A.I.,
00:16:23
a tech startup that's building
what they call "synths,"
00:16:26
or synthetic humans.
00:16:28
That's right.
00:16:29
Artificial intelligence
wrapped in a body.
00:16:32
[Gildert] Our mission
is to create machines
00:16:34
that are indistinguishable
from humans
00:16:36
physically, cognitively,
and emotionally.
00:16:39
[Downey] Doing so
00:16:40
involves solving problems
of engineering,
00:16:42
computer science,
neuroscience,
00:16:44
biology,
even art and design.
00:16:47
But for her,
00:16:49
the problem of artificially
replicating a person
00:16:51
boils down
to a deeper question...
00:16:53
What does it mean to be human?
00:16:55
[Gildert] Understanding
what it is to be human
00:16:57
is a question that
we've been asking ourselves
00:16:59
for many thousands of years,
00:17:01
so I'd like to turn science
and technology to that question
00:17:05
to try and figure out
who we are.
00:17:06
[Downey] We love stories
and films about clones
00:17:09
and replicants
and humanoid robots.
00:17:12
Why are we so obsessed
00:17:14
with the idea
of recreating ourselves?
00:17:17
Is it biological?
00:17:18
Existential?
00:17:21
[Gildert] To try
and understand something fully,
00:17:23
you have to reverse-engineer it,
00:17:25
you have to
put it back together.
00:17:28
[Downey] The human
that Suzanne knows best
00:17:30
is... Suzanne,
00:17:32
so one of her projects
00:17:33
is to build a synthetic replica
of herself.
00:17:36
[Gildert] There's this thing
called the Turing test,
00:17:38
which is trying to have an A.I.
00:17:40
that you can't tell
is not a human.
00:17:43
So I wanna try and create
a physical Turing test,
00:17:46
where you can't tell
whether or not
00:17:48
the system you're actually
physically interacting with
00:17:51
is a person,
or whether it's a robot.
00:17:53
So here we have 132 cameras...
00:17:57
which are all pointed at me,
00:17:59
and they all take a photograph
simultaneously.
00:18:02
This data is used to create
00:18:05
a full three-dimensional
body scan of me
00:18:07
that we can then use
to create a robot version of me.
00:18:11
[Downey] Suzanne believes
00:18:12
that we experience life
through the senses,
00:18:14
so she's putting as much work
into making the body lifelike
00:18:17
as she is the mind.
00:18:19
[Gildert] We broke down
this very ambitious project
00:18:21
into several
different categories.
00:18:23
The first category is physical.
00:18:25
Can you build a robotic system
that looks like a person?
00:18:34
So the synth
has bones and muscles
00:18:36
that are roughly analogous
to the human body,
00:18:40
but not quite as complex.
00:18:43
These hands are 3D-printed
as an entire piece
00:18:46
on our printers.
00:18:48
[Gildert] We can actually
print in carbon fiber
00:18:50
and Kevlar,
00:18:51
and we can create robot bones
00:18:53
that are stronger
than aluminum machined parts,
00:18:57
with these beautiful
organic biological shapes.
00:19:00
So I'm adding in
a finger sensor.
00:19:04
This, uh, current generation
00:19:05
has a single sensor
on the fingertip.
00:19:09
[Gildert] We build a machine
that perceives like a human
00:19:11
by trying to copy the human
sensorium very accurately.
00:19:17
The most complicated part
of the perception system
00:19:21
is actually
the sense of touch.
00:19:23
Are you monitoring the touch?
00:19:24
Yes.
Touch received.
00:19:28
[Holly Marie Peck]
We've actually embedded
00:19:30
capacitive touch sensors
in the synth's hand,
00:19:32
essentially pressure sensors
00:19:34
allowing it to feel, uh,
its environment,
00:19:36
and interact
and manipulate objects.
00:19:38
Let's just test the pressure.
00:19:40
-Okay.
-This should max it out.
00:19:42
Yep, yep. Maxed out.
00:19:43
Just stretch out her hand.
Okay, go.
00:19:45
[Gildert] The reason
the hand and the arm
00:19:47
is able to move so fluidly
00:19:49
is because of
pneumatic actuators.
00:19:51
They work using compressed air.
00:19:55
You actuate
one of these devices,
00:19:57
and it kind of contracts
00:19:59
and pulls on a tendon,
00:20:01
so the actuation mechanism is
very similar to a human muscle.
00:20:05
It's just not yet
quite as efficient.
00:20:09
[Shannon] I'm adding
the camera into the eyeball.
00:20:12
Now I'm adding
the cosmetic front of the eye.
00:20:15
[Gildert] The eyes are
super important to get right.
00:20:18
Similar
to our own vision system,
00:20:20
they can see
similar color spectrum,
00:20:22
and they can also,
because there's two cameras,
00:20:25
they can have
depth perception too.
00:20:26
[Peck] Restarting
facial detection.
00:20:30
[Gildert] That actually
looks pretty good.
00:20:32
[Peck] Mm-hmm. Do you wanna
come forward a little bit?
00:20:34
Yeah.
00:20:35
-I'm gonna restart
her headboard.
-[Gildert] Okay.
00:20:37
That information
00:20:38
is fed through a series
of different A.I. algorithms.
00:20:42
One algorithm
is a facial detection system.
00:20:45
She's definitely seeing me.
00:20:47
[Peck] Yes, she is.
00:20:49
I can tell
she's looking at me,
00:20:50
'cause she looked
straight at me.
00:20:52
Yeah, gaze tracking
is working.
00:20:53
Okay, cool.
Now, do you wanna just smile?
00:20:54
I'll see if she's actually
capturing your emotion?
00:20:58
[Gildert] If you're smiling,
00:20:59
the corners of your mouth
come up,
00:21:01
your eyes open a little bit,
00:21:02
and the A.I. system
can actually detect
00:21:04
how those landmarks have moved
relative to one another.
00:21:08
[Rana el Kaliouby]
I think the moment in time
00:21:09
we're at right now
00:21:10
is very exciting
00:21:12
because there's this field
that's concerned
00:21:14
about building human-like
generalized intelligence,
00:21:17
and sometimes even kind of
surpassing human intelligence.
00:21:22
[Daphne Koller]
There's people out there
00:21:23
who believe that this is
on our immediate horizon.
00:21:27
I don't.
00:21:28
I think we're a long ways away
00:21:31
from machines
that are truly conscious
00:21:35
and think on their own.
00:21:36
She's responding.
I can see her face changing.
00:21:39
[synth] You look happy.
00:21:41
-Good.
-Mm-hmm.
00:21:42
I'm gonna look sad.
00:21:45
You look sad.
00:21:47
Okay, good.
00:21:48
[Peck]
We have actually configured
00:21:50
a lot of A.I. algorithms
on the back end
00:21:52
that give the robot
00:21:53
the capabilities
of recognizing people,
00:21:55
detecting emotion,
00:21:56
recognizing gestures and poses
that people are making.
00:22:00
It then responds
in various ways
00:22:02
with its environment.
00:22:04
[Gildert]
Bring up her node graph
00:22:05
so you can see
what's running in her brain.
00:22:08
Yeah, let's see
all the online modules.
00:22:10
The chatbot,
emotion detection,
00:22:12
object detection...
00:22:14
Wonderful. Gaze tracking...
00:22:15
[Gildert] The body,
in a way, is the easy part.
00:22:19
Creating the mind
is a lot harder.
00:22:21
[Downey] Creating the mind
is more than hard.
00:22:24
It's basically impossible,
00:22:25
at least for now,
00:22:27
and maybe forever,
00:22:28
because a mind
is not just knowledge,
00:22:30
or skill, or even language,
00:22:32
all of which
a machine can learn.
00:22:34
The part that makes us
really human is consciousness;
00:22:37
an awareness,
a sense of being,
00:22:39
of who we are
00:22:40
and how we fit in time
and space around us.
00:22:43
A human mind has that...
00:22:45
and memory.
00:22:47
"I remember the experience
of buying a new pencil case
00:22:50
and the supplies to go in it,
00:22:51
getting all those
new little things
00:22:53
that smelled nice,
00:22:54
and were all clean
and colorful."
00:22:56
If you think about
how people work,
00:22:59
it's very unusual for you
to meet a person
00:23:01
that doesn't have a backstory.
00:23:02
I can use all the data
that I have about myself
00:23:05
to try and craft something
that has my memories,
00:23:09
it has my same mannerisms,
00:23:10
and it thinks and feels
the way I do.
00:23:18
I would like them
to become their own beings,
00:23:20
and to me,
00:23:22
creating the copy is a way
of pushing the A.I. further
00:23:26
towards making it
a realistic human
00:23:28
by having it be a copy
of a specific human.
00:23:31
I remember going
to Bolton Town Center
00:23:35
quite often.
00:23:36
We just called it "Town."
00:23:39
[Gildert] The basic idea
00:23:41
is you send in
a large amount of text data,
00:23:43
and the system learns
correlations between words,
00:23:47
and the idea
00:23:48
is that the synth could use
one of these models
00:23:50
to kind of blend together
an idea of a memory
00:23:53
that may have happened
or may not have happened,
00:23:55
so it's a little bit
of an artistic way
00:23:58
of recreating memories.
00:23:59
I remember going
into WH Smith.
00:24:02
It had a very distinct smell
that I can still recall.
00:24:06
[Gildert] So by giving them
these backstories now,
00:24:10
we believe that we will be able
to learn in the future
00:24:13
how they can create
their own memories
00:24:15
from their experiences.
00:24:18
[Bran Ferren] I love the idea
00:24:19
that there are
passionate people
00:24:20
who are dedicating
their time and energy
00:24:24
to making these things happen.
00:24:25
Why?
00:24:26
Because if and when
it does happen,
00:24:28
it's going to be because of
those passionate people.
00:24:31
We talk about
the computer revolution
00:24:33
like it's done.
00:24:34
It's barely begun.
00:24:37
We don't understand
00:24:38
where the impact
of these technologies will be
00:24:41
over the next five, ten,
00:24:42
20, 30, 50, 100 years.
00:24:44
If you think it's exciting
and confusing now,
00:24:48
fasten your seatbelts,
00:24:50
because it hasn't begun.
00:24:51
What is your name?
00:24:53
My name is Holly.
00:24:55
What is your name?
00:25:00
Hmm.
00:25:01
[Gildert] Of course
there's that unknown,
00:25:02
like are we gonna
run into a problem
00:25:04
with trying to recreate a mind
00:25:06
that no one's thought of yet?
00:25:07
My name is Nadine.
00:25:10
Interesting.
00:25:11
I am glad to see you.
00:25:13
[Downey] Even if we do
one day figure out
00:25:15
how to create a virtual mind,
00:25:17
it's not just the science.
00:25:18
There's also the ethics.
00:25:20
What kind of rights
will the robots have?
00:25:22
Can we imbue it
with good values,
00:25:24
make sure it's unbiased?
00:25:26
What if breaks the law
or commits a crime?
00:25:28
Are we responsible
for our synths?
00:25:31
[el Kaliouby] There are
big ethical challenges
00:25:33
in the field of A.I.
00:25:35
I believe that as a community
of A.I. innovators
00:25:38
and thought leaders,
00:25:39
we have to really be
at the forefront
00:25:41
of enforcing and designing
00:25:45
these best practices
and guidelines
00:25:47
around how we build
and deploy ethical A.I.
00:25:50
I like to say
that artificial intelligence
00:25:52
should not be
about the artificial,
00:25:54
it should be about the humans.
00:25:56
You look angry.
00:25:57
Landmarks are registering.
00:25:59
[Ferren] I think
it's perfectly reasonable
00:26:02
to have a set of rules
that govern ethical behavior
00:26:05
when you are dealing
with technologies
00:26:08
that can have direct impact
into people's lives
00:26:11
and their families
and the future.
00:26:13
[Gildert] The vision's
very ambitious for this.
00:26:15
We'd like to think that that is
a 10- to 20-year mission.
00:26:19
You might say
we're somewhere like
00:26:20
five to 10% of the way along.
00:26:23
Why is her arm doing that?
00:26:25
It's almost like
it's not clearing the buffer.
00:26:28
Yeah... interesting.
00:26:31
Let's just restart you
so your arm goes--
00:26:33
Oh, wait, it's going
back down again.
00:26:36
Okay, that's good.
00:26:37
Okay.
00:26:38
How do you feel today, Nadine?
00:26:40
It feels good to be a synth.
00:26:43
Nice.
00:26:43
"It feels good to be a synth."
00:26:45
[Gildert] The synths
are not mobile at the moment,
00:26:47
they can't move around,
00:26:48
they can't walk yet.
00:26:49
That's something
we're going to be adding in
00:26:51
within the next couple of years.
00:26:53
The grand goal
00:26:54
is to make these
into their own beings
00:26:56
with their own volition
and their own rights.
00:27:01
There are these moments
you can have
00:27:03
where you really feel
something that's unusual.
00:27:06
It's surprising.
00:27:08
I was adjusting
the synth's hair,
00:27:11
and then she suddenly,
like, smiled,
00:27:13
and opened her mouth
a little bit,
00:27:15
like, you know, like I'd just
tickled her or something.
00:27:18
It was just, like, synchronous
with what I was doing.
00:27:21
[Downey] In some ways,
00:27:23
Suzanne's vision
is already coming alive.
00:27:25
She's making a connection,
albeit small, with a machine.
00:27:28
Isn't that something?
00:27:30
[Domingos] I think A.I.
is part of evolution.
00:27:32
The same evolution
00:27:33
that led from bacteria
to animals,
00:27:34
and has led people
to create technology,
00:27:37
has led them to create A.I.
00:27:38
In some ways, we're still
in the very early infancy
00:27:41
of this new age.
00:27:43
[Downey] Will we ever
create intelligent life
00:27:45
here on Earth...
00:27:47
or maybe we'll find it
out there first?
00:27:57
So I'm on my way
00:27:58
to the SETI Institute
headquarters
00:28:00
in Mountain View,
00:28:01
and, and I'm gonna show, uh,
what the A.I. system found
00:28:06
in the data that we collected.
00:28:07
I'm excited.
I'm a little nervous too.
00:28:18
[Tarter] We need to be able
to follow up in real time...
00:28:20
[Diamond] Mm-hmm.
00:28:21
[Tarter]
...as closely as we can,
00:28:22
so that a signal that's there
00:28:24
is still gonna be there
when we go back to look for it,
00:28:27
and we can then classify it.
00:28:28
Jill Tarter is really a legend
00:28:30
in this whole field
of SETI research.
00:28:34
Also really a pioneer
as a woman astronomer.
00:28:39
The character played
by Jodie Foster in Contact,
00:28:42
is based, at least
in the first half of that movie,
00:28:44
on Jill Tarter.
00:28:45
[Tarter] People often talk
00:28:47
about finding
a needle in a haystack
00:28:49
as being a difficult task,
00:28:51
but the SETI task is far harder.
00:28:54
If I got out of bed
every morning
00:28:57
thinking, "This is the day
we're gonna find the signal,"
00:28:59
I have pretty good odds
00:29:01
I'm gonna go to bed
that night disappointed.
00:29:05
I don't get up in the morning
thinking that.
00:29:08
What I do
get up in the morning thinking
00:29:10
is that today,
I'm going to figure out
00:29:12
how to do this search better,
00:29:14
do new things,
00:29:16
do things you could not do
in the past.
00:29:19
Early on, the technology
just wasn't there...
00:29:21
Mm-hmm.
00:29:22
...and now we're doing something
00:29:23
that we've never
been able to do.
00:29:25
I'm excited.
00:29:30
-Hello?
-Oh, hey!
00:29:31
-Look who's here!
-How are ya?
00:29:33
-Good to see you!
-Hi, Graham.
00:29:34
-Nice to see you.
-Nice to see you.
00:29:36
Likewise.
Good to see you too.
00:29:37
-Hey, Bill.
-It's been a couple
of whole days?
00:29:39
-I know! [laughs]
-Thanks for coming down.
00:29:41
-My pleasure, I'm excited.
-Yeah.
00:29:42
We're thinking maybe
you've got some news.
00:29:44
Well, I wanna
step you through it.
00:29:47
Here you can kinda see
00:29:49
the system is
initially very active.
00:29:50
It's all lit up,
00:29:52
and very quickly,
00:29:53
it starts to get a handle
on what the shape,
00:29:56
you know, what a signal
from the Trappist-1 system
should look like.
00:29:58
Over on the far right
is its areas of interest...
00:30:02
What I'm showing here
00:30:03
is a time-compressed video
of the A.I. system
00:30:06
looking at the signal
we gathered.
00:30:09
...and if you focus in on that,
00:30:11
the A.I. system did indeed
flag this one area,
00:30:14
at that point, saying,
00:30:16
-"Whoa, back up.
Something just happened."
-[Tarter] Ooh, wow.
00:30:18
"That's not right,"
00:30:20
and if you zoom in
on the actual data,
00:30:22
sure enough, there's that spike,
00:30:24
so that is not
from the Trappist system.
00:30:26
That was generated
by the Allen Telescope Array,
00:30:29
but, you know, beyond that,
00:30:30
this is an area
that the A.I. system is saying,
00:30:32
"This isn't quite
what I would have expected."
00:30:36
This is a little
more interesting
00:30:37
'cause there's
more structure to it,
00:30:39
and we should take its hints,
00:30:40
and have a deeper analysis done
of this part of the observation.
00:30:45
We didn't write any code.
00:30:46
We didn't tell it to...
to look for spikes of power
00:30:50
or anything else.
00:30:51
We just said, "You know what,
you figure out what's normal,
00:30:53
and you let us know
00:30:54
when something catches
your attention,"
00:30:56
which is exactly
what it's doing there.
00:30:58
It's encouraging,
00:30:59
because already
with just this one observation,
00:31:02
we started to see
some real progress
00:31:04
in what the A.I. system can do
compared to our own eyes,
00:31:07
and that's just
one observation.
00:31:09
What about the next,
and the next,
00:31:11
and as it gets better
00:31:12
with each new round of data
that we collect?
00:31:14
This is after two hours.
00:31:16
I wonder how good it's gonna get
after a hundred hours.
00:31:20
Yeah.
00:31:21
If we just routinely keep
feeding the data from the A.T.A.
00:31:25
into this model,
00:31:26
it's gonna get better
and better and better.
00:31:28
We can just scale this out.
00:31:29
-Right.
-Absolutely.
00:31:30
We just got smarter.
Thank you, machine.
00:31:32
Yes, exactly.
00:31:33
[Tarter]
I'm absolutely so excited.
00:31:35
I'm really blown away.
00:31:36
I can see
the tools that are being built
00:31:41
give us a new way
of looking for things
00:31:43
that we hadn't thought of,
00:31:44
and things that we don't have
to define up front,
00:31:47
anomalies that
the machines will find
00:31:51
simply because
they've looked at so much data.
00:31:55
[Mackintosh] I do think
we're going to find ET.
00:31:58
I do think we are gonna find
signs of civilization
00:32:02
beyond Earth,
00:32:04
and I do think that it's going
to be A.I. that finds it.
00:32:12
[Downey] Is there
intelligent life out there?
00:32:14
Can we create
human-like machines?
00:32:18
[Domingos]
The odds are overwhelming
00:32:20
that we will eventually be able
to build an artificial brain
00:32:24
that is at the level
of the human brain.
00:32:26
The big question
is how long will it take?
00:32:30
[Downey] Outer space,
00:32:32
inner life...
00:32:34
Age-old mysteries
now seem more solvable.
00:32:37
[Chris Botham]
If we wanna go to Mars,
00:32:38
if we wanna populate
other planets,
00:32:40
these types of things require
these advanced technologies.
00:32:42
[Downey] Moonshots, yeah,
00:32:45
but also
other pressing problems,
00:32:47
like...
00:32:48
-[gasps of shock]
-All five! Whoa!
00:32:51
[Downey]
...the mind and body.
00:32:52
[Tim Shaw]
Are you working today?
00:32:55
[beeping]
00:32:56
It's wonderful.
00:32:59
[Downey] Adaptation...
00:33:00
[Jim Ewing]
I'm thinking and doing
00:33:01
and getting instant response.
00:33:03
It makes it feel like
it's part of me.
00:33:05
[Downey] Work...
00:33:06
Action!
00:33:07
[Downey]
...and creativity...
00:33:09
These types of technologies
can help us do our tasks better.
00:33:11
Three, two, one.
00:33:13
[computer voice]
Autonomous driving started.
00:33:16
[el Kaliouby] I believe
if we do this right,
00:33:18
these A.I. systems can truly,
truly compliment
00:33:21
what we do as humans.
00:33:22
[Eric Warren]
We use the A.I. tools
00:33:24
to predict what the future
not only is,
00:33:27
but what it should be.
00:33:28
Yo, what's up?
This is will.i.am.
00:33:30
[laughing]
00:33:31
[Mark Sagar] This is
the new version of you.
00:33:32
The way it's looking so far
is mind-blowing.
00:33:36
[firefighter]
Stay close, I'll lead.
00:33:38
[Downey] Survival...
00:33:39
[firefighter]
Over here, I see him!
Three yards at 2:00!
00:33:41
[Martin Ford] I believe
that artificial intelligence
00:33:43
is really going to be
00:33:45
the most important tool
in our toolbox
00:33:47
for solving the big problems
that we face.
00:33:49
[firefighter] I got him!
00:33:51
[crowd chanting]
00:33:53
[Downey] Conservation...
00:33:54
The fact that we can
look across the world
00:33:56
and find where famine
might happen
00:33:58
four months from now,
00:34:00
it's mind-blowing.
00:34:01
[Downey] All out of the realm
of sci-fi and magic,
00:34:05
and now just science.
00:34:07
Still hard problems,
but now possible,
00:34:11
with innovation,
00:34:13
computing power,
will, and passion...
00:34:15
-[cheering] Yay!
-Yes!
00:34:17
There it is.
00:34:18
[Downey] ...and yet,
despite all that,
00:34:20
a vestige of unknown endures.
00:34:23
Who are we?
00:34:25
What are we becoming?
00:34:27
Every major
technological change
00:34:30
leads to a new kind of society,
00:34:32
with new moral principles,
00:34:34
and the same thing will happen
with A.I.
00:34:36
[Downey] Technology's
changing us, for sure.
00:34:40
The whole idea
of what it means to be human
00:34:42
is getting rewired.
00:34:44
A.I. might be humanity's
most valuable tool...
00:34:49
...but it's also just that.
00:34:51
A tool.
00:34:52
[clattering]
00:34:55
[Downey] What we choose
to do with it...
00:34:59
that's up to you and I.
00:35:08
[Seth Shostak]
If you could project yourself
00:35:10
into the next millennium,
00:35:11
a thousand years from now,
00:35:13
would we look back
on this generation and say,
00:35:15
"Well, they were the last
generation of Homo sapiens
00:35:18
that actually ran the planet"?
00:35:20
[James Parr]
There's a lot of paranoia.
00:35:22
The media's done
a really good job
00:35:24
of making people frightened,
00:35:25
but A.I. is just
a portrait of reality,
00:35:29
a very close portrait,
but it isn't reality.
00:35:31
It's just a bucket
of probabilities.
00:35:33
Where I think human beings
will always have the edge
00:35:37
are understanding other humans.
00:35:39
It's going to take a long time
00:35:41
before we have an A.I.
00:35:42
that can understand
all of the nuances
00:35:46
and various layers
of the human experience
00:35:48
at a societal level.
00:35:50
[Shostak] James Parr, thanks
so very much for being with us.
00:35:52
Great, thank you.
