June 6, 2019 • 37 mins
We travel to Silicon Valley where Astro Teller welcomes us inside X, Google's secretive innovation laboratory, to explain how one of the most powerful companies on earth is building the future. X is working on everything from creating new antibiotics to restoring internet connectivity after natural disasters. And they're not alone. Tech companies are increasingly involved in building infrastructure, and even playing the role of government. But what kind of power does this give them? And what can we do to rein it in?
In this episode: Astro Teller of X, Lina Khan of Columbia Law School, Sal Candido & Pamela Desrochers of Loon, Jack Clark of OpenAI, and Jamie Susskind, author of "Future Politics".
Learn more about your ad-choices at https://www.iheartpodcastnetwork.com
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:04):
Sleepwalkers is a production of I Heart Radio and Unusual
productions in Casa Yeah Familiar. I was already at home,
resting with my family when I suddenly get a phone call.
Had fallen from the sky. Picture of field in rural Mexico.
(00:31):
Surrounding it yellow tape, a police car flashing lights, soldiers
carrying automatic weapons. Inside the line of cordon is a
large transparent globe, a bit like a washed up jellyfish,
but in fact it had fallen from the sky. Here
is state official Juan Carlos Castile, important to corder. It
(00:54):
fell from the sky at n in an erratic way,
making circles. The artifact itself had some sensors and light
that was flashing beeping one Manuel Sanchez. You had the
guts to approach the object, look at it closely, take
pictures and share them. Those pictures quickly turned into a
social media phenomenon. It really cost punic in the people.
(01:18):
Maybe they have never seen something like this. Actually, people
came from other settlements to see the artifact to take
selfies with it. I imagine that it could be an
ESPIONA charity fact that relay the images. Classified Information Officer
Castillo worried about the origin when a beeping wearing jellyfish
falls out of the sky. Sure it could be spycraft,
(01:41):
but if it's something even scarier, something other worldly, Officer
Castillo noticed that the beast was tagged with the phone number,
and if you dial that number, you'd ultimately get to
balloon C s I. This is where we find out
why balloons either last a hundred days or they don't.
That's La des Roche and where at X, a secretive
(02:03):
lab at alphabet, Google's parent company. Their mission is no
less than to invent the future. And it's X who
piloted the program to launch balloons into the stratosphere and
retrieve them when they come down. So, in order to
look at a balloon that's the size of a tennis court,
where a failure smaller than a millimeter is a huge
problem for us UM, we do have to build specialized equipment.
(02:26):
So behind me is the world's largest flatbed scanner. We're
working on models of what does a certain type of
damage look like, what do we think causes that damage?
It kind of makes a fingerprint. But why does X
and balloons into the sky for months at a time
in the first place, I'm Velosen and this is Sleepwalkers.
(02:59):
So carry you first told me about this balloon being
discovered in rural Mexico, and then we tracked down the
first responders. But what grabbed you about the story? First
of all, it's a crazy story. Something fell out of
the sky into a field in rural Mexico, which I love.
But it's also one of those moments when you know,
real people come in contact with technology in a way
that almost feels like not to mention them again, but
(03:21):
like a Steven Spielberg movie, you know. I mean, just
imagine not knowing what this thing is in your backyard,
and it sort of reminds me of other stories that
we've reported on, whether it's Gillian with targeting, Glen with
parole algorithms. You know, the way in which people interact
with technology is changing, and this is just the perfect
(03:43):
example of that. This February, we went to Mountain View
in California to visit x and learn how they're involved
with giant balloons floating in the stratosphere, because, as it
turns out, rural Mexico isn't the only place they've made contact.
We had internet before Maria, we had so many things
(04:04):
that we depended on the Internet for everything. Five days
after Hurrican Maria devastated Puerto Rico, the SEC granted an
experimental license to X to restore cell service to the island.
The government let X step in to provide a service
it couldn't, and Google sent their balloons. So when we
saw what was happening in Puerto Rico, you know, it
(04:24):
was really hard for us not the help right like that.
The whole company in some sense really wanted to get
behind that effort. It's kind of a rare opportunity when um,
there's a problem like that, where you know, problem with connectivity,
everyone's offline and you happen to have a fleet of
stratospheric Internet balloons like UM, I think we're probably the
only ones who can say that. That's Sell Candido, head
(04:48):
of Engineering Fallon And if cell towers in the sky
sound like the stuff of science fiction, they're already here
on A massive magnitude eight earthquake hit Peru, and while
cell towers and cables were down, Loons balloons were able
to restore temporary Internet access within just forty eight hours.
(05:09):
But how does Sell and his team get the balloons
to go where they're needed. So the idea in Act
one was we were going to build this ring or
balloons around the world, covering an entire latitude band. Obviously,
that's kind of a challenging concept to execute. Most of
them will be over the ocean, not connecting anyone. The
idea just wasn't feasible. So to make it work Google
(05:31):
had to figure out how to get the balloons to navigate.
Astro Tylor runs X and he supervised the original development
of Loon. We always hoped that the balloons could be
intelligent in which winds they choose to jump onto. As
the balloons have gotten better and better at predicting what
(05:52):
the winds will be at different altitudes, they can play
these more and more sophisticated chess games. If I go
up by a kilometer, I think I could catch a
wind that's going to the left at ten miles an hour,
and I'll hang out there for about three hours. Then
I'll go down by two kilometers and it plays this
out and makes this plan for how it's going to
(06:14):
get not just kind of to Australia, but right over Perth,
constantly reading the winds and predicting how they might change
in order to stay around the Earth is no small task,
and according to Sell, it takes a lot of computing power.
I don't think that you would be able to do this,
you know, if you had a person navigating each balloon.
(06:37):
The information processing capability and the act that you have
to be constantly watching making adjustments, it's a job that's
really well suited to a computer. It is a huge
volume of computation in our data center. I think that
is an area where Alphabet has a big advantage. How
big is a giant data center? Oh um, they're giant
(06:59):
shy uping center, bigger than shopping centers, powered by often
renewable energy, so they're often built next to a river
just so it can use an entire hydro electric plant.
You know, there's a time of computers being put to
all kinds of problems across Alphabet Luon is one of them. Right.
The reality is that today a company like Alphabet has
(07:20):
more computational power, more data and more engineering expertise than
most countries. So it can restore connectivity to Puerto Rico
or Peru after a natural disaster, and it can help
in daily emergencies as well as Obviously, Christio told us.
There are situations in his underconnected area that are currently
impossible to communicate quickly enough to receive help in time,
(07:42):
and Luon could change that. So the balloon that fell
out of the sky in Mexico, it was a sign
of things to come of a new global infrastructure being
built by technology companies, not governments, and that brings real leverage,
which is something we should all think about, even if
we don't find a giant gen fish in our backyard.
(08:04):
I'm all for development. There are emergency situations in the
Sierra that are impossible to communicate to us in order
to act within reasonable time frames, but I don't dismiss
the possibility of it being a trick by a company
to steal information from us either. Of course, these balloons
(08:25):
are actually very well intentioned fundamentally to bring internet to
places where it doesn't otherwise exist. That said, Kara, Google
aren't the only people trying to do this. Facebook are two,
and they had a program for a while attempting to
use solar powered drones to connect the world. Right, And
when Facebook and Google are both trying to do something,
it's says to me that it's probably not purely a
(08:48):
philanthropic endeavor, there's probably some bottom line and getting all
those people online. So anyway, we just heard from astro Teller,
who runs X, which is historically a very secretive organization.
We were invited inside the building. Do you understand a
bit more about how one of the world's most powerful companies,
Google is thinking about inventing the future. Hi, I'm astro Teller.
(09:12):
I am the captain of Moonshots here at X. It's
the part of alphabet where we try to design things
that can become, if we're lucky, new businesses, hopefully Google scale,
new businesses that can be as good for the world
as Google husband. So Astro has charged with bulk producing
Google scale innovations, as impossible as that sounds, but like
(09:36):
any good factory supervisor, he's got some clear criteria. In
order for something to be a moonshot, we require that
it has three things. A huge problem with the world,
radical proposed solution, and then some kind of breakthrough technology
that gives us at least a fighting chance of making
that science fiction sounding product. If you look around, you'll
see a lot of bear concrete, polished cement floors apply
(10:00):
would on the walls. The building's work in progress because
the projects here are a work in progress. To even
be considered for development at X, ideas have to be spectacular,
balanced at the edge of the impossible. When someone says,
what if we put a band of copper around the
North Pole and let the flux of Earth's magnetic core,
(10:24):
which goes up and down like a reversing, very slow
alternating current, turn it into current in that wire, and
then we could pipe all that current down to Norway
or something and power the Earth that way. That may
not actually work, but that statement that what if definitely
(10:47):
took you outside of the normal, And as Astro points out,
the lab itself is a moon shot. He has charged
with delivering ten X impact on the world's most intractable problems.
But they take think concerns seriously too, which is reportedly
why X abandoned what on an invisibility device. We go
talk to the public, we go talk to experts and
(11:09):
thought leaders, we go talk to regulators, and we put
these half formed ideas and prototypes in front of them
and say, this is the problem we're trying to solve,
here's how we're currently trying to solve it. What do
you think And we get feedback and that helps educate
us and repoint us in various ways. Internally, we also
(11:30):
play a lot of what if games, So an example
is what we call design fiction, where we either make
pictures or literally write stories about our technology and how
it might play out. But if you can't imagine it
working out in really good ways for society, we definitely
shouldn't be making it. And if you can imagine it
(11:53):
working out, but you can describe some bad things that
this moonshot might cause in society, being able to describe
those things means maybe we can change how we're working
on the moon shot. It's good to hear that there's
a strong ethical framework governing innovation at X, but it's
still a private company and they're making ethical decisions that
(12:14):
impact all of us. Astro has been talking about infrastructure,
but remember in our first episode the program to deter
potential terrorists using targeted ads that was also sponsored by Google.
So providing Internet and deterring terrorists are both good ideas,
but when the same company does both and so much more,
(12:34):
the concentration of power is concerning. Then again, in a
political environment of gridlock and proposed cuts to science funding,
maybe we should be grateful that the big technology companies
are stepping into tackle urgent problems. You've heard the bacteria
around the world are becoming more resistant to antibiotics. There's
(12:56):
no way for some drug company to chase after all
of these increasingly antibiotic resistant bacteria. And yet if we
don't chase after them, we're going to look back at
this time in the world as the golden age of
antibiotics and sure miss it. Even if you could just
simulate a bacteria inside a computer, that is asked, what
(13:21):
would happen to this bacteria if I knocked out this gene,
if I changed the pH in the solution that it's
sitting in, if I subjected it to a lot of
UV light. If you could just ask those kinds of
basic questions, you could start to design new antibacterial agents
at a thousand times the rate because you don't have
(13:43):
to go around pipetting and waiting for these things to
grow or die. If you could simulate life, that is,
model any part of biology inside the computer, it would
be a do over for the life sciences. That's an
example of something we're in the very early days of exploring.
(14:03):
Alphabet Google x loon Jigsaw. This constellation of interlocking companies
is working on everything from bringing the Internet remote parts
of the world, to deterring terrorists to creating new kinds
of antibiotics. And if that sounds a lot like the
work of government, well that's because it is. When we
come back, we ask what does this concentration of power
(14:25):
mean for us and what should we do about it.
We've been talking a lot about Google Cara, but of
course there's not just Google. We mentioned Facebook earlier, and
then there's Amazon. Amazon makes up almost half of all
online sales in the US. Just think about everything, you know.
(14:46):
We've talked about Alexa gathering intimate data and more than
a hundred million American homes and Jeff Be's His ambitions
also include redefining healthcare in the US and also recently
colonizing space where nicely so humble. Um, but it is
getting harder and harder to define what these massive businesses
actually are. Donald Trump calls the Washington Post the Amazon
(15:09):
Washington Post, which it is since Bezo sported. I mean,
I guess, not exactly the Amazon Washington Post. But it
does identify a real issue, which is that it's getting
harder and harder to say what these multi industry companies are. Well, yes,
because they all do so much, So how do we
regulate them. One person who's been out in front of
this issue is Lena Can While at Yale Law School
(15:30):
and age eight, she wrote a landmark article called Amazon's
antitrust Paradox. And we're now starting to have conversations as
a society about using monopoly law to limit the power
of the big tech companies. But Lena really kicks out
of the conversation. So I was very excited when she
agreed to join us on the show. These technologies are complicated.
(15:51):
There are oftentimes involved in multiple lines of business, and
so for many everyday people, including lawmakers, they don't actually
understand and how these firms operate. According to Lena, the
big three Amazon, Facebook, and Google have one thing in common.
They've emerged as gatekeepers of the digital economy. Amazon, Google
(16:13):
and Facebook are in a position where they can really
pick winners and losers, especially among the merchants or the
content producers or the app developers that are now reliant
on their platform to get to market. So if you're
a consumer, you're primarily thinking about price, about convenience, about quality, um,
you know, if the fact, if you're a new parent
(16:33):
and you can just order diapers and they'll be reliably
at your doorstep the next morning, there's no doubt that
Amazon has provided important benefits to consumers. But if you're
thinking about the company as a citizen, and you're thinking
about the market power that it has, if you're thinking
about the way that it was able to avoid paying
(16:54):
sales taxes for the first years that it was in business,
if you're looking at the way in which Amazon or
distorted its searched for its second headquarters, where it was
pretty ruthlessly pitting city against city and showing that it
was really willing to extort municipalities try and get the
biggest subsidy possible, and then ended up playing a bit
of a bait and switch where it collected all of
(17:15):
this information from all of these different cities, which now
inevitably will give it a competitive advantage. I mean, there's
just so many different dimensions of Amazon's dominance that are troubling.
If there was any doubt about Amazon's power to compel politicians,
the search for the second headquarters should have cleared it up.
And now Amazon is beginning to build leverage over the
(17:36):
federal government. There have been reports from the Institute for
Local Self Reliance about how Amazon is now on the
cusp of receiving potentially a big contract with the Pentagon Um.
So it's just entrenching itself deeper and deeper into our
daily lives in ways that if you just look at
an isolation, you will miss the bigger picture. But I
(17:57):
think as a citizen and as somebody who's looking at
the social and political implications of this, it's quite troubling.
I'm very struck Carab by Lena's distinction between being a
consumer and being a citizen, because of course, as consumers,
we always want the best price and the most convenience.
But it's fascinating to think that that can be directly
at odds with our responsibilities and even our interests as citizens. Yeah,
(18:20):
you know, I think a lot of readers who love
independent bookstores have this problem. You know, do I order
on Amazon because I want the book tomorrow and because
it's cheap and because I'm trying to save money, or
do I buy it at the independent bookstore where you know,
I've gone and loved for years. So yeah, I always
think about that. But you know, I think we also
have to think about shareholders and politicians when we talk
about this. Amazon shareholders are trying to keep Amazon from
(18:42):
selling Amazon's recognition technology to the US government, But those
same shareholders are profiting off of Amazon success, So we
really can't always rely on shareholders to do that. Yeah.
I mean, those shareholders who don't want Amazon to sell
facial recognition technology to the government are actually acting against
them interests, and that's an anomaly, which is why we
(19:03):
have government theoretically. But remember the Facebook Senate hearings. I mean,
the senators did a terrible job of holding Zuckerberg to account.
Mr Zuckerberg, I remember well your first visit to Capitol
Hill back in two thousand and ten, you spoke to
the Senate Republican high Tech tasks for us, which I
chare You said back then that Facebook would always be free.
(19:25):
How do you sustain a bitteress model in which users
don't play for your service? Senator, we run ads? Nice, see, Senator,
we run ads. That was Mark Zuckerberg's response to Senator
orn Hatch, which was probably the low point in a
series of low points in those Senate hearings, and also
(19:47):
the moment that really maybe want to make this podcast
because clearly the people who have the power and duty
to question the big technology companies are either being deredicted
their responsibilities or they simply don't understand what's going on.
And I'm not sure which is worse. Part of the
issue with the Senate is that collectively they are a
million years old, um. But part of it is also
(20:08):
that they all rely on Facebook ads and pages for reelection,
and they want Facebook money in their states. That's what
we call a conflict of interest. Conflict of interest and
a friend of mine from university literally wrote the book
on this issue. It's called Future Politics. He's a lawyer
called Jamie Suskind. It's an ancient idea in human civilization
(20:31):
that we don't allow great forms of power to be
erected over us without some degree of transparency, so we
know what's being done with technology. It's very early days,
but we need to look at the stuff as citizens
like we would at any form of power that a
cruise over our head, whether it's corporate power or political
(20:51):
power or great economic power. One of the big problems,
as Jamie sees it, is that as these big technology
companies get involved with things like creating new drug or
fighting terrorism. We start to talk about them as nation states,
and we missed the mark. You know, you'll see journalists
and commentators saying our tech firms are the new states,
and I just think that's that's sloppy thinking they might
(21:13):
have some stuff in common with states. But tech firms
are commercial entities operating in a market system for the
pursuit of profit and are answerable to their shareholders, which
is obviously just a profoundly different social institution to a state.
Kara mentioned that Amazon shareholders were pressuring the company to
ban facial recognition software sales, but in late May of
(21:34):
that measure was roundly defeated by a vote at the
company's annual general meeting, so Amazon will continue selling facial
recognition technology. And this highlights what Jamie is saying. These
firms are motivated by profit, unlike states, which are motivated
by protecting their citizens. Senator Elizabeth Warren has suggested one
solution might be to treat the big technology companies like
(21:56):
utility companies, things like regulating price and access. But Jamie
thinks this actually underplays just how powerful and consequential these
companies really are. They're not like utility firms. The water
company doesn't get you to do things you wouldn't otherwise do.
It doesn't affect the democratic process, it doesn't set the
limits of your liberty, and it doesn't distribute things of
(22:18):
importance throughout society according to principles of justice, which are
all things that I would say that tech firms now do.
So rowing back, we don't really have the words to
describe what a tech firm is conceptually and politically, and
so it's no wonder that we're not coming up with
policies and regulations and laws because we don't even have
the words to describe the future, even if we could
see it. When we come back, we investigate that future
(22:43):
and ask what we can do to regain some control. So, Kara,
everyone is now thinking a lot about the role of
technology in our lives, and one of the popular things
recently has been for reporters to try and live without technology.
I feel like phone detox is the new weight watchers,
(23:04):
and I'm kind of sick of it. But Kashmir Hill
did this really interesting story for Gizmoto where she tried
not to use any products from the big five tech companies.
You know, she blocked Amazon, Facebook, Google, Microsoft, and Apple,
and she said it was hell, you know, I think
about my own life, Google and Amazon are absolutely indispensable
to me. And then there are times where I'm like, oh,
(23:24):
I don't really use Facebook that much, you know, and
I'm sitting in a public restroom at two pm on
Instagram and sort of check on my there's Facebook, right,
And I think what Kashmir Hills story also revealed is
that even when we think we're not using one of
those companies products, we may well be using a website
or an app that's powered by them. So it's basically
(23:45):
impossible to opt out his Lina again, if what you
mean by opt out is you know, not have an
Amazon Prime account or not have Gmail. You know, there
may be ways in which you can stop using these
services in a day to day sense, But Amazon also
owns Amazon Web Services UM, which you know much of
the Internet now relies on. Google is providing the back
(24:08):
end infrastructure for so many other services. So I think,
you know, if you're if you're trying to UM, if
you're trying to delete these firms from all aspects of
your life, it actually becomes very difficult to live in
modern day society and Like we said, this applies to
all of the Big five, but let's play it out
with Amazon and we can start with the easy stuff.
(24:29):
So Kara, no more things for the kitchen from Amazon
dot Com well as Walmart. Um, no more groceries from
Whole Foods. Can we afford it anyway? Um? And no
more Amazon Prime Video. Sad to miss Mrs Mazel But
how the Netflix you know you like mazel us Um.
(24:49):
I think the consumer facing stuff is obviously interesting, but
the more consequential piece of this is the back end
of the Internet. You know, Lena was talking about Amazon
Web Services, which is like the cloud computing, the powers
so many of the services we use every day, and
Amazon basically sells this to other companies. So actually, without
(25:09):
this back end from Amazon, there is no Netflix, Unilever, Visor,
General Electric. These companies all rely on Amazon Web Services.
So does NASA, where I went to space camp sort of.
And even Apple, who compete with Amazon in areas like
streaming and devices, are also on track to spend more
than three hundred million dollars with Amazon this year. So
(25:31):
Apple are paying literally one of their competitors to host
their data. Yes, so if you're using the Internet, you're
in Amazon's territory. Territory is a good word. One of
the more interesting analogies I've heard is, well, I'll just
let Jack Clark say it. I think it's kind of
analogous to feudalism. Like you and I get to live
(25:52):
on some sort of a state like Google or Facebook
or Twitter. You know, via state is owned by a
few you door lord, which is the owners of these
companies and their boards, and the estate is able to
extract my labor benefit from it, and what it gives
me is stability. Jack is the policy director of Open Ai,
(26:14):
an AI research company founded by Elon Musk and Sam Altman.
When Jack says labor, he's referring to use a data Now,
part of open eyes mission is to compete with the
AI labs at Google, Facebook, and Amazon. So Jack might
not be fully impartial, But let's think about that comparison
to feudalism. If you're a peasant in a village owned
(26:34):
by a feudal lord, you had some guarantee of stability
as long as you stayed on that platform, but you
couldn't leave the platform. And actually, as early as the
thirteenth century, English peasants were being required to carry identification
cards because people would get really really grump you affirm
if they walked off the estate there on and tried
to go somewhere else. And I think that's actually very
(26:56):
similar to where we are today. We don't have possible
data our our ability to economically benefit ourselves with our
data is actually very very limited, and we live in
this kind of neo feudal system where you get to
pick your platform, they get all of the data of
the benefit, and you get some free service in exchange.
(27:17):
So we perceive what we're getting from these companies as
free car and all we have to do is tolerate
a few advertisements. But as we talked about, we're not
just giving them our eyeballs to look at ads. No,
we're allowing them to understand our patterns, behaviors, and desires
better than we understand them ourselves. And then they're using
that knowledge about us to hold our attention, influence our behavior,
(27:38):
and monetize us and sell us stuff that we don't need.
When you think about it, historically leaders have killed for
this sort of power influence, and we're handing over now
freely to Bazos and Zuckerberg and Larry page. We're eroding
our power as citizens because it's easier and more comfortable
to be consumers. I think it's important to remember that
the same place where a person can by a twelve
(28:01):
pack of toilet paper is also the same company that
is selling Amazon Web services to the Pentagon. Yeah, I
mean it's it's sobering. It is sobering. We focus a
lot on how the big technology companies are powered by
our data, but there's something else this as well. Jack
Clark of open ai argues that it's distracting us this
(28:21):
focus on data from something even more fundamental. Who owns
the computing power? Now. A bet that we have is
that the value of those large amounts of data is
going to reduce over time as you develop algorithms that
are better able to extract structure from smaller and smaller
amounts of data. But you're always going to need compute
(28:42):
to allow you to run more experiments and train bigger systems.
So we think in the long term compute might be
the key determiner of AI progress. Just to clarify, compute
is short for computing power, the ability to process large
amounts of data. It's what's needed to help loons loons
model and catch the winds, and a sale mentioned Google
(29:04):
used so much of it that they build their massive
data centers next to rivers so they can use hydro
electric energy to power them. And compute has grown at
an extraordinary rate as we've become able to make smaller
and smaller circuit boards and pack more power into smaller cases.
Now openly, I recently did an analysis where we looked
at the amount of compute that had been used in
(29:26):
breakthrough AI systems in recent years. And when we did
this analysis, we found out that the the amount of
compute has grown by three hundred thousand times in six years.
Three hundred thousand times in six years. What does that
mean in practical terms? What does that growth actually look like? Well,
I think a better measure here is to think about
your phone battery. That's equivalent to your phone going from
(29:49):
having a battery that lasted for one day six years
ago to a battery that lasted for eight hundred years today.
That's what that growth looks like, and it means that
more power for capabilities are coming into view faster than
we expect. It's that kind of growth that's allowed a
fleet of loon balloons to model the winds, adjust their
(30:09):
altitude and sail to precise locations. But the computational power
is in the hands of the private sector, and that
is dramatically increasing their power. We have brilliant people come
and work with us from places like you know, m
I T or Stanford, and one of the things that
attracts them to work here, or attracts them to work
at a Google or a Facebook, is we can give
(30:32):
them more computers and they can get their home institution.
I think if we don't solve this disparity, we're going
to really wreck the public benefits of scientific research because
you're going to have a whole class of research which
only occurs in the private sector, and therefore there are
very few guarantees but that research will always be public.
(30:54):
Earlier this year, President Trump announced an executive order called
the American AI Initially IF that laid out a plan
to address some of the concerns we've raised in this episode,
but it didn't come with any funding, So innovation and
ethical decisions will remain in private hands for the time being,
and this means that AI technology isn't guaranteed to serve
(31:15):
the public interest and may even get into the wrong hands.
I think for the first step we as a community
need to take is to acknowledge potential harms. And once
we have that mindset, I think it becomes easier to
sell the scientific community on Okay, we have a sense
of what harm looks like, what do we commit ourselves
to minimize that? So, yeah, that conversation has to happen.
(31:38):
I think if that conversation does not happen, then you're
going to have this arms race absent the conscious creation
of norms here. I think that that's for default, and
that default world terrifies me because I can see AI
research today that in two or three years is going
to give us, say, unprecedented capabilities in drone autonomy, that
(32:00):
to drone navigate to a target in a small urban area. Now,
that's obviously an amazingly good thing. If we want to
create rapid response drones that can deliver, say, tools for
dealing with cardiac arrest someone undergoing that on a city street,
I don't want to think about the version of this
where the drone has some explosive strap to it and
it is being used to assassinate somewhat. And I want
(32:21):
to have the AI community confront this problem. But the
AI community won't be able to do this alone. We
need our politicians to step up and create meaningful laws
and regulations. Ultimately, rely on tech companies to regulate themselves
is an abdication of responsibility. Here's Lena Can again. The
(32:42):
dominance of these tech companies is not inevitable, and none
of the economic outcomes that we're seeing in these markets
are inevitable, and they're deeply shaped by laws and policy
and the political choices that we're making about how we
allow these firms to expand and grown kinds of practices
they're allowed to engage in. So I think it's really
important to push back against determinism and inevitability narratives and
(33:07):
reassert the role of law and policy in shaping economic outcomes.
There's an important balance between public and private institutions in America,
and letting either side grow too powerful creates problems. But
right now, the big technology companies are becoming so large
and powerful as to be ungovernable, and no matter how
(33:28):
ethical or what intention they may be, they're not motivated
by fairness or protecting the weakest in society. They're motivated
by shareholders and profit that may seem like an insurmountable problem.
But we have a history in this country of making
laws that disrupt special interests and raise the quality of
life for everyone. Think about the formation of the e
(33:48):
p A, the New Deal, even the Civil Rights Act.
And actually, just this week, as we've been preparing to
release this episode, the House Judiciary Committee has launched a
major bipartisan and trust investigation into the big tech companies.
This is the first time in decades that Congress has
investigated a specific industry. And who is advising the investigation, Well,
(34:10):
that would be Lena con Yeah. We got in touch
with Lena to hear about our thoughts on the probe
and what we might be able to expect from the investigation,
but she's not able to comment publicly yet. Still, she's
said a couple of interesting things in our original interview.
The goal of anti trust is to keep markets competitive.
And I think it's important because it affects us as consumers,
it affects us as workers, it affects us an entrepreneurs,
(34:31):
and it affects us a citizens. Right. I mean, I
think the structure of our economy has huge consequences for
our day to day lives, and anti trust is a
key component of that. If we saw more robust anti
trust enforcement, then these markets should become more open to competition,
and so in ten years you would not necessarily see Google, Amazon, Facebook, Apple,
(34:54):
Microsoft continue to be as dominant as they are, because
we would have seen innovators. We would have seen the
next wave of of the breakthrough disruptors. I think it's
also worth acknowledging that the dominant these firms enjoy is
certain markets maybe something that we regulate rather than addressed
(35:14):
through breaking up. So since we've started working on Sleepwalkers,
there has been tangible progress, both in the US and
around the world. But this is only the beginning and
we can't relax just yet. In the next episode, we
bring things down to earth and look at how AI
may help us feed the world, and along the way,
we use machine learning to invent a brand new Seltzer
(35:36):
flavor just for us. I'm a veloshen see you next time.
Sleepwalkers is a production of our heart radio and unusual
productions for the latest AI news, live interviews, and behind
(35:59):
the scene in footage. Find us on Instagram, at Sleepwalker's
Podcast or at Sleepwalker's podcast dot com. Special thanks this
episode to the whole team at Weird Moved West, an
incredible production company in El Paso, Texas, who helped us
track down an interview Officer Castillo about his encounter with
the Loon Balloon. Special shout out to j W. Rogers,
(36:22):
Jorge carry On, asail Anya and Leonel Portillo, who voiced
Officer Castillo's translated lines. Sleepwalkers is hosted by me Ozveloshin
and co hosted by me Kara Price, with produced by
Julian Weller with help from Jacopo Penzo and Taylor Chacogne,
mixing by Tristan McNeil and Julian Weller. Our story editor
is Matthew Riddle. Recording assistance this episode from Miguel Paris
(36:44):
and Chris Handbrake. Sleepwalkers is executive produced by me Ozveloshin
and Mangesh Hattikiler. For more podcasts from My Heart Radio,
visit the I Heart Radio app, Apple Podcasts, or wherever
you listen to your favorite shows.
Sleepwalkers is a production of I Heart Radio and Unusual
productions in Casa Yeah Familiar. I was already at home,
resting with my family when I suddenly get a phone call.
Had fallen from the sky. Picture of field in rural Mexico.
(00:31):
Surrounding it yellow tape, a police car flashing lights, soldiers
carrying automatic weapons. Inside the line of cordon is a
large transparent globe, a bit like a washed up jellyfish,
but in fact it had fallen from the sky. Here
is state official Juan Carlos Castile, important to corder. It
(00:54):
fell from the sky at n in an erratic way,
making circles. The artifact itself had some sensors and light
that was flashing beeping one Manuel Sanchez. You had the
guts to approach the object, look at it closely, take
pictures and share them. Those pictures quickly turned into a
social media phenomenon. It really cost punic in the people.
(01:18):
Maybe they have never seen something like this. Actually, people
came from other settlements to see the artifact to take
selfies with it. I imagine that it could be an
ESPIONA charity fact that relay the images. Classified Information Officer
Castillo worried about the origin when a beeping wearing jellyfish
falls out of the sky. Sure it could be spycraft,
(01:41):
but if it's something even scarier, something other worldly, Officer
Castillo noticed that the beast was tagged with the phone number,
and if you dial that number, you'd ultimately get to
balloon C s I. This is where we find out
why balloons either last a hundred days or they don't.
That's La des Roche and where at X, a secretive
(02:03):
lab at alphabet, Google's parent company. Their mission is no
less than to invent the future. And it's X who
piloted the program to launch balloons into the stratosphere and
retrieve them when they come down. So, in order to
look at a balloon that's the size of a tennis court,
where a failure smaller than a millimeter is a huge
problem for us UM, we do have to build specialized equipment.
(02:26):
So behind me is the world's largest flatbed scanner. We're
working on models of what does a certain type of
damage look like, what do we think causes that damage?
It kind of makes a fingerprint. But why does X
and balloons into the sky for months at a time
in the first place, I'm Velosen and this is Sleepwalkers.
(02:59):
So carry you first told me about this balloon being
discovered in rural Mexico, and then we tracked down the
first responders. But what grabbed you about the story? First
of all, it's a crazy story. Something fell out of
the sky into a field in rural Mexico, which I love.
But it's also one of those moments when you know,
real people come in contact with technology in a way
that almost feels like not to mention them again, but
(03:21):
like a Steven Spielberg movie, you know. I mean, just
imagine not knowing what this thing is in your backyard,
and it sort of reminds me of other stories that
we've reported on, whether it's Gillian with targeting, Glen with
parole algorithms. You know, the way in which people interact
with technology is changing, and this is just the perfect
(03:43):
example of that. This February, we went to Mountain View
in California to visit x and learn how they're involved
with giant balloons floating in the stratosphere, because, as it
turns out, rural Mexico isn't the only place they've made contact.
We had internet before Maria, we had so many things
(04:04):
that we depended on the Internet for everything. Five days
after Hurrican Maria devastated Puerto Rico, the SEC granted an
experimental license to X to restore cell service to the island.
The government let X step in to provide a service
it couldn't, and Google sent their balloons. So when we
saw what was happening in Puerto Rico, you know, it
(04:24):
was really hard for us not the help right like that.
The whole company in some sense really wanted to get
behind that effort. It's kind of a rare opportunity when um,
there's a problem like that, where you know, problem with connectivity,
everyone's offline and you happen to have a fleet of
stratospheric Internet balloons like UM, I think we're probably the
only ones who can say that. That's Sell Candido, head
(04:48):
of Engineering Fallon And if cell towers in the sky
sound like the stuff of science fiction, they're already here
on A massive magnitude eight earthquake hit Peru, and while
cell towers and cables were down, Loons balloons were able
to restore temporary Internet access within just forty eight hours.
(05:09):
But how does Sell and his team get the balloons
to go where they're needed. So the idea in Act
one was we were going to build this ring or
balloons around the world, covering an entire latitude band. Obviously,
that's kind of a challenging concept to execute. Most of
them will be over the ocean, not connecting anyone. The
idea just wasn't feasible. So to make it work Google
(05:31):
had to figure out how to get the balloons to navigate.
Astro Tylor runs X and he supervised the original development
of Loon. We always hoped that the balloons could be
intelligent in which winds they choose to jump onto. As
the balloons have gotten better and better at predicting what
(05:52):
the winds will be at different altitudes, they can play
these more and more sophisticated chess games. If I go
up by a kilometer, I think I could catch a
wind that's going to the left at ten miles an hour,
and I'll hang out there for about three hours. Then
I'll go down by two kilometers and it plays this
out and makes this plan for how it's going to
(06:14):
get not just kind of to Australia, but right over Perth,
constantly reading the winds and predicting how they might change
in order to stay around the Earth is no small task,
and according to Sell, it takes a lot of computing power.
I don't think that you would be able to do this,
you know, if you had a person navigating each balloon.
(06:37):
The information processing capability and the act that you have
to be constantly watching making adjustments, it's a job that's
really well suited to a computer. It is a huge
volume of computation in our data center. I think that
is an area where Alphabet has a big advantage. How
big is a giant data center? Oh um, they're giant
(06:59):
shy uping center, bigger than shopping centers, powered by often
renewable energy, so they're often built next to a river
just so it can use an entire hydro electric plant.
You know, there's a time of computers being put to
all kinds of problems across Alphabet Luon is one of them. Right.
The reality is that today a company like Alphabet has
(07:20):
more computational power, more data and more engineering expertise than
most countries. So it can restore connectivity to Puerto Rico
or Peru after a natural disaster, and it can help
in daily emergencies as well as Obviously, Christio told us.
There are situations in his underconnected area that are currently
impossible to communicate quickly enough to receive help in time,
(07:42):
and Luon could change that. So the balloon that fell
out of the sky in Mexico, it was a sign
of things to come of a new global infrastructure being
built by technology companies, not governments, and that brings real leverage,
which is something we should all think about, even if
we don't find a giant gen fish in our backyard.
(08:04):
I'm all for development. There are emergency situations in the
Sierra that are impossible to communicate to us in order
to act within reasonable time frames, but I don't dismiss
the possibility of it being a trick by a company
to steal information from us either. Of course, these balloons
(08:25):
are actually very well intentioned fundamentally to bring internet to
places where it doesn't otherwise exist. That said, Kara, Google
aren't the only people trying to do this. Facebook are two,
and they had a program for a while attempting to
use solar powered drones to connect the world. Right, And
when Facebook and Google are both trying to do something,
it's says to me that it's probably not purely a
(08:48):
philanthropic endeavor, there's probably some bottom line and getting all
those people online. So anyway, we just heard from astro Teller,
who runs X, which is historically a very secretive organization.
We were invited inside the building. Do you understand a
bit more about how one of the world's most powerful companies,
Google is thinking about inventing the future. Hi, I'm astro Teller.
(09:12):
I am the captain of Moonshots here at X. It's
the part of alphabet where we try to design things
that can become, if we're lucky, new businesses, hopefully Google scale,
new businesses that can be as good for the world
as Google husband. So Astro has charged with bulk producing
Google scale innovations, as impossible as that sounds, but like
(09:36):
any good factory supervisor, he's got some clear criteria. In
order for something to be a moonshot, we require that
it has three things. A huge problem with the world,
radical proposed solution, and then some kind of breakthrough technology
that gives us at least a fighting chance of making
that science fiction sounding product. If you look around, you'll
see a lot of bear concrete, polished cement floors apply
(10:00):
would on the walls. The building's work in progress because
the projects here are a work in progress. To even
be considered for development at X, ideas have to be spectacular,
balanced at the edge of the impossible. When someone says,
what if we put a band of copper around the
North Pole and let the flux of Earth's magnetic core,
(10:24):
which goes up and down like a reversing, very slow
alternating current, turn it into current in that wire, and
then we could pipe all that current down to Norway
or something and power the Earth that way. That may
not actually work, but that statement that what if definitely
(10:47):
took you outside of the normal, And as Astro points out,
the lab itself is a moon shot. He has charged
with delivering ten X impact on the world's most intractable problems.
But they take think concerns seriously too, which is reportedly
why X abandoned what on an invisibility device. We go
talk to the public, we go talk to experts and
(11:09):
thought leaders, we go talk to regulators, and we put
these half formed ideas and prototypes in front of them
and say, this is the problem we're trying to solve,
here's how we're currently trying to solve it. What do
you think And we get feedback and that helps educate
us and repoint us in various ways. Internally, we also
(11:30):
play a lot of what if games, So an example
is what we call design fiction, where we either make
pictures or literally write stories about our technology and how
it might play out. But if you can't imagine it
working out in really good ways for society, we definitely
shouldn't be making it. And if you can imagine it
(11:53):
working out, but you can describe some bad things that
this moonshot might cause in society, being able to describe
those things means maybe we can change how we're working
on the moon shot. It's good to hear that there's
a strong ethical framework governing innovation at X, but it's
still a private company and they're making ethical decisions that
(12:14):
impact all of us. Astro has been talking about infrastructure,
but remember in our first episode the program to deter
potential terrorists using targeted ads that was also sponsored by Google.
So providing Internet and deterring terrorists are both good ideas,
but when the same company does both and so much more,
(12:34):
the concentration of power is concerning. Then again, in a
political environment of gridlock and proposed cuts to science funding,
maybe we should be grateful that the big technology companies
are stepping into tackle urgent problems. You've heard the bacteria
around the world are becoming more resistant to antibiotics. There's
(12:56):
no way for some drug company to chase after all
of these increasingly antibiotic resistant bacteria. And yet if we
don't chase after them, we're going to look back at
this time in the world as the golden age of
antibiotics and sure miss it. Even if you could just
simulate a bacteria inside a computer, that is asked, what
(13:21):
would happen to this bacteria if I knocked out this gene,
if I changed the pH in the solution that it's
sitting in, if I subjected it to a lot of
UV light. If you could just ask those kinds of
basic questions, you could start to design new antibacterial agents
at a thousand times the rate because you don't have
(13:43):
to go around pipetting and waiting for these things to
grow or die. If you could simulate life, that is,
model any part of biology inside the computer, it would
be a do over for the life sciences. That's an
example of something we're in the very early days of exploring.
(14:03):
Alphabet Google x loon Jigsaw. This constellation of interlocking companies
is working on everything from bringing the Internet remote parts
of the world, to deterring terrorists to creating new kinds
of antibiotics. And if that sounds a lot like the
work of government, well that's because it is. When we
come back, we ask what does this concentration of power
(14:25):
mean for us and what should we do about it.
We've been talking a lot about Google Cara, but of
course there's not just Google. We mentioned Facebook earlier, and
then there's Amazon. Amazon makes up almost half of all
online sales in the US. Just think about everything, you know.
(14:46):
We've talked about Alexa gathering intimate data and more than
a hundred million American homes and Jeff Be's His ambitions
also include redefining healthcare in the US and also recently
colonizing space where nicely so humble. Um, but it is
getting harder and harder to define what these massive businesses
actually are. Donald Trump calls the Washington Post the Amazon
(15:09):
Washington Post, which it is since Bezo sported. I mean,
I guess, not exactly the Amazon Washington Post. But it
does identify a real issue, which is that it's getting
harder and harder to say what these multi industry companies are. Well, yes,
because they all do so much, So how do we
regulate them. One person who's been out in front of
this issue is Lena Can While at Yale Law School
(15:30):
and age eight, she wrote a landmark article called Amazon's
antitrust Paradox. And we're now starting to have conversations as
a society about using monopoly law to limit the power
of the big tech companies. But Lena really kicks out
of the conversation. So I was very excited when she
agreed to join us on the show. These technologies are complicated.
(15:51):
There are oftentimes involved in multiple lines of business, and
so for many everyday people, including lawmakers, they don't actually
understand and how these firms operate. According to Lena, the
big three Amazon, Facebook, and Google have one thing in common.
They've emerged as gatekeepers of the digital economy. Amazon, Google
(16:13):
and Facebook are in a position where they can really
pick winners and losers, especially among the merchants or the
content producers or the app developers that are now reliant
on their platform to get to market. So if you're
a consumer, you're primarily thinking about price, about convenience, about quality, um,
you know, if the fact, if you're a new parent
(16:33):
and you can just order diapers and they'll be reliably
at your doorstep the next morning, there's no doubt that
Amazon has provided important benefits to consumers. But if you're
thinking about the company as a citizen, and you're thinking
about the market power that it has, if you're thinking
about the way that it was able to avoid paying
(16:54):
sales taxes for the first years that it was in business,
if you're looking at the way in which Amazon or
distorted its searched for its second headquarters, where it was
pretty ruthlessly pitting city against city and showing that it
was really willing to extort municipalities try and get the
biggest subsidy possible, and then ended up playing a bit
of a bait and switch where it collected all of
(17:15):
this information from all of these different cities, which now
inevitably will give it a competitive advantage. I mean, there's
just so many different dimensions of Amazon's dominance that are troubling.
If there was any doubt about Amazon's power to compel politicians,
the search for the second headquarters should have cleared it up.
And now Amazon is beginning to build leverage over the
(17:36):
federal government. There have been reports from the Institute for
Local Self Reliance about how Amazon is now on the
cusp of receiving potentially a big contract with the Pentagon Um.
So it's just entrenching itself deeper and deeper into our
daily lives in ways that if you just look at
an isolation, you will miss the bigger picture. But I
(17:57):
think as a citizen and as somebody who's looking at
the social and political implications of this, it's quite troubling.
I'm very struck Carab by Lena's distinction between being a
consumer and being a citizen, because of course, as consumers,
we always want the best price and the most convenience.
But it's fascinating to think that that can be directly
at odds with our responsibilities and even our interests as citizens. Yeah,
(18:20):
you know, I think a lot of readers who love
independent bookstores have this problem. You know, do I order
on Amazon because I want the book tomorrow and because
it's cheap and because I'm trying to save money, or
do I buy it at the independent bookstore where you know,
I've gone and loved for years. So yeah, I always
think about that. But you know, I think we also
have to think about shareholders and politicians when we talk
about this. Amazon shareholders are trying to keep Amazon from
(18:42):
selling Amazon's recognition technology to the US government, But those
same shareholders are profiting off of Amazon success, So we
really can't always rely on shareholders to do that. Yeah.
I mean, those shareholders who don't want Amazon to sell
facial recognition technology to the government are actually acting against
them interests, and that's an anomaly, which is why we
(19:03):
have government theoretically. But remember the Facebook Senate hearings. I mean,
the senators did a terrible job of holding Zuckerberg to account.
Mr Zuckerberg, I remember well your first visit to Capitol
Hill back in two thousand and ten, you spoke to
the Senate Republican high Tech tasks for us, which I
chare You said back then that Facebook would always be free.
(19:25):
How do you sustain a bitteress model in which users
don't play for your service? Senator, we run ads? Nice, see, Senator,
we run ads. That was Mark Zuckerberg's response to Senator
orn Hatch, which was probably the low point in a
series of low points in those Senate hearings, and also
(19:47):
the moment that really maybe want to make this podcast
because clearly the people who have the power and duty
to question the big technology companies are either being deredicted
their responsibilities or they simply don't understand what's going on.
And I'm not sure which is worse. Part of the
issue with the Senate is that collectively they are a
million years old, um. But part of it is also
(20:08):
that they all rely on Facebook ads and pages for reelection,
and they want Facebook money in their states. That's what
we call a conflict of interest. Conflict of interest and
a friend of mine from university literally wrote the book
on this issue. It's called Future Politics. He's a lawyer
called Jamie Suskind. It's an ancient idea in human civilization
(20:31):
that we don't allow great forms of power to be
erected over us without some degree of transparency, so we
know what's being done with technology. It's very early days,
but we need to look at the stuff as citizens
like we would at any form of power that a
cruise over our head, whether it's corporate power or political
(20:51):
power or great economic power. One of the big problems,
as Jamie sees it, is that as these big technology
companies get involved with things like creating new drug or
fighting terrorism. We start to talk about them as nation states,
and we missed the mark. You know, you'll see journalists
and commentators saying our tech firms are the new states,
and I just think that's that's sloppy thinking they might
(21:13):
have some stuff in common with states. But tech firms
are commercial entities operating in a market system for the
pursuit of profit and are answerable to their shareholders, which
is obviously just a profoundly different social institution to a state.
Kara mentioned that Amazon shareholders were pressuring the company to
ban facial recognition software sales, but in late May of
(21:34):
that measure was roundly defeated by a vote at the
company's annual general meeting, so Amazon will continue selling facial
recognition technology. And this highlights what Jamie is saying. These
firms are motivated by profit, unlike states, which are motivated
by protecting their citizens. Senator Elizabeth Warren has suggested one
solution might be to treat the big technology companies like
(21:56):
utility companies, things like regulating price and access. But Jamie
thinks this actually underplays just how powerful and consequential these
companies really are. They're not like utility firms. The water
company doesn't get you to do things you wouldn't otherwise do.
It doesn't affect the democratic process, it doesn't set the
limits of your liberty, and it doesn't distribute things of
(22:18):
importance throughout society according to principles of justice, which are
all things that I would say that tech firms now do.
So rowing back, we don't really have the words to
describe what a tech firm is conceptually and politically, and
so it's no wonder that we're not coming up with
policies and regulations and laws because we don't even have
the words to describe the future, even if we could
see it. When we come back, we investigate that future
(22:43):
and ask what we can do to regain some control. So, Kara,
everyone is now thinking a lot about the role of
technology in our lives, and one of the popular things
recently has been for reporters to try and live without technology.
I feel like phone detox is the new weight watchers,
(23:04):
and I'm kind of sick of it. But Kashmir Hill
did this really interesting story for Gizmoto where she tried
not to use any products from the big five tech companies.
You know, she blocked Amazon, Facebook, Google, Microsoft, and Apple,
and she said it was hell, you know, I think
about my own life, Google and Amazon are absolutely indispensable
to me. And then there are times where I'm like, oh,
(23:24):
I don't really use Facebook that much, you know, and
I'm sitting in a public restroom at two pm on
Instagram and sort of check on my there's Facebook, right,
And I think what Kashmir Hills story also revealed is
that even when we think we're not using one of
those companies products, we may well be using a website
or an app that's powered by them. So it's basically
(23:45):
impossible to opt out his Lina again, if what you
mean by opt out is you know, not have an
Amazon Prime account or not have Gmail. You know, there
may be ways in which you can stop using these
services in a day to day sense, But Amazon also
owns Amazon Web Services UM, which you know much of
the Internet now relies on. Google is providing the back
(24:08):
end infrastructure for so many other services. So I think,
you know, if you're if you're trying to UM, if
you're trying to delete these firms from all aspects of
your life, it actually becomes very difficult to live in
modern day society and Like we said, this applies to
all of the Big five, but let's play it out
with Amazon and we can start with the easy stuff.
(24:29):
So Kara, no more things for the kitchen from Amazon
dot Com well as Walmart. Um, no more groceries from
Whole Foods. Can we afford it anyway? Um? And no
more Amazon Prime Video. Sad to miss Mrs Mazel But
how the Netflix you know you like mazel us Um.
(24:49):
I think the consumer facing stuff is obviously interesting, but
the more consequential piece of this is the back end
of the Internet. You know, Lena was talking about Amazon
Web Services, which is like the cloud computing, the powers
so many of the services we use every day, and
Amazon basically sells this to other companies. So actually, without
(25:09):
this back end from Amazon, there is no Netflix, Unilever, Visor,
General Electric. These companies all rely on Amazon Web Services.
So does NASA, where I went to space camp sort of.
And even Apple, who compete with Amazon in areas like
streaming and devices, are also on track to spend more
than three hundred million dollars with Amazon this year. So
(25:31):
Apple are paying literally one of their competitors to host
their data. Yes, so if you're using the Internet, you're
in Amazon's territory. Territory is a good word. One of
the more interesting analogies I've heard is, well, I'll just
let Jack Clark say it. I think it's kind of
analogous to feudalism. Like you and I get to live
(25:52):
on some sort of a state like Google or Facebook
or Twitter. You know, via state is owned by a
few you door lord, which is the owners of these
companies and their boards, and the estate is able to
extract my labor benefit from it, and what it gives
me is stability. Jack is the policy director of Open Ai,
(26:14):
an AI research company founded by Elon Musk and Sam Altman.
When Jack says labor, he's referring to use a data Now,
part of open eyes mission is to compete with the
AI labs at Google, Facebook, and Amazon. So Jack might
not be fully impartial, But let's think about that comparison
to feudalism. If you're a peasant in a village owned
(26:34):
by a feudal lord, you had some guarantee of stability
as long as you stayed on that platform, but you
couldn't leave the platform. And actually, as early as the
thirteenth century, English peasants were being required to carry identification
cards because people would get really really grump you affirm
if they walked off the estate there on and tried
to go somewhere else. And I think that's actually very
(26:56):
similar to where we are today. We don't have possible
data our our ability to economically benefit ourselves with our
data is actually very very limited, and we live in
this kind of neo feudal system where you get to
pick your platform, they get all of the data of
the benefit, and you get some free service in exchange.
(27:17):
So we perceive what we're getting from these companies as
free car and all we have to do is tolerate
a few advertisements. But as we talked about, we're not
just giving them our eyeballs to look at ads. No,
we're allowing them to understand our patterns, behaviors, and desires
better than we understand them ourselves. And then they're using
that knowledge about us to hold our attention, influence our behavior,
(27:38):
and monetize us and sell us stuff that we don't need.
When you think about it, historically leaders have killed for
this sort of power influence, and we're handing over now
freely to Bazos and Zuckerberg and Larry page. We're eroding
our power as citizens because it's easier and more comfortable
to be consumers. I think it's important to remember that
the same place where a person can by a twelve
(28:01):
pack of toilet paper is also the same company that
is selling Amazon Web services to the Pentagon. Yeah, I
mean it's it's sobering. It is sobering. We focus a
lot on how the big technology companies are powered by
our data, but there's something else this as well. Jack
Clark of open ai argues that it's distracting us this
(28:21):
focus on data from something even more fundamental. Who owns
the computing power? Now. A bet that we have is
that the value of those large amounts of data is
going to reduce over time as you develop algorithms that
are better able to extract structure from smaller and smaller
amounts of data. But you're always going to need compute
(28:42):
to allow you to run more experiments and train bigger systems.
So we think in the long term compute might be
the key determiner of AI progress. Just to clarify, compute
is short for computing power, the ability to process large
amounts of data. It's what's needed to help loons loons
model and catch the winds, and a sale mentioned Google
(29:04):
used so much of it that they build their massive
data centers next to rivers so they can use hydro
electric energy to power them. And compute has grown at
an extraordinary rate as we've become able to make smaller
and smaller circuit boards and pack more power into smaller cases.
Now openly, I recently did an analysis where we looked
at the amount of compute that had been used in
(29:26):
breakthrough AI systems in recent years. And when we did
this analysis, we found out that the the amount of
compute has grown by three hundred thousand times in six years.
Three hundred thousand times in six years. What does that
mean in practical terms? What does that growth actually look like? Well,
I think a better measure here is to think about
your phone battery. That's equivalent to your phone going from
(29:49):
having a battery that lasted for one day six years
ago to a battery that lasted for eight hundred years today.
That's what that growth looks like, and it means that
more power for capabilities are coming into view faster than
we expect. It's that kind of growth that's allowed a
fleet of loon balloons to model the winds, adjust their
(30:09):
altitude and sail to precise locations. But the computational power
is in the hands of the private sector, and that
is dramatically increasing their power. We have brilliant people come
and work with us from places like you know, m
I T or Stanford, and one of the things that
attracts them to work here, or attracts them to work
at a Google or a Facebook, is we can give
(30:32):
them more computers and they can get their home institution.
I think if we don't solve this disparity, we're going
to really wreck the public benefits of scientific research because
you're going to have a whole class of research which
only occurs in the private sector, and therefore there are
very few guarantees but that research will always be public.
(30:54):
Earlier this year, President Trump announced an executive order called
the American AI Initially IF that laid out a plan
to address some of the concerns we've raised in this episode,
but it didn't come with any funding, So innovation and
ethical decisions will remain in private hands for the time being,
and this means that AI technology isn't guaranteed to serve
(31:15):
the public interest and may even get into the wrong hands.
I think for the first step we as a community
need to take is to acknowledge potential harms. And once
we have that mindset, I think it becomes easier to
sell the scientific community on Okay, we have a sense
of what harm looks like, what do we commit ourselves
to minimize that? So, yeah, that conversation has to happen.
(31:38):
I think if that conversation does not happen, then you're
going to have this arms race absent the conscious creation
of norms here. I think that that's for default, and
that default world terrifies me because I can see AI
research today that in two or three years is going
to give us, say, unprecedented capabilities in drone autonomy, that
(32:00):
to drone navigate to a target in a small urban area. Now,
that's obviously an amazingly good thing. If we want to
create rapid response drones that can deliver, say, tools for
dealing with cardiac arrest someone undergoing that on a city street,
I don't want to think about the version of this
where the drone has some explosive strap to it and
it is being used to assassinate somewhat. And I want
(32:21):
to have the AI community confront this problem. But the
AI community won't be able to do this alone. We
need our politicians to step up and create meaningful laws
and regulations. Ultimately, rely on tech companies to regulate themselves
is an abdication of responsibility. Here's Lena Can again. The
(32:42):
dominance of these tech companies is not inevitable, and none
of the economic outcomes that we're seeing in these markets
are inevitable, and they're deeply shaped by laws and policy
and the political choices that we're making about how we
allow these firms to expand and grown kinds of practices
they're allowed to engage in. So I think it's really
important to push back against determinism and inevitability narratives and
(33:07):
reassert the role of law and policy in shaping economic outcomes.
There's an important balance between public and private institutions in America,
and letting either side grow too powerful creates problems. But
right now, the big technology companies are becoming so large
and powerful as to be ungovernable, and no matter how
(33:28):
ethical or what intention they may be, they're not motivated
by fairness or protecting the weakest in society. They're motivated
by shareholders and profit that may seem like an insurmountable problem.
But we have a history in this country of making
laws that disrupt special interests and raise the quality of
life for everyone. Think about the formation of the e
(33:48):
p A, the New Deal, even the Civil Rights Act.
And actually, just this week, as we've been preparing to
release this episode, the House Judiciary Committee has launched a
major bipartisan and trust investigation into the big tech companies.
This is the first time in decades that Congress has
investigated a specific industry. And who is advising the investigation, Well,
(34:10):
that would be Lena con Yeah. We got in touch
with Lena to hear about our thoughts on the probe
and what we might be able to expect from the investigation,
but she's not able to comment publicly yet. Still, she's
said a couple of interesting things in our original interview.
The goal of anti trust is to keep markets competitive.
And I think it's important because it affects us as consumers,
it affects us as workers, it affects us an entrepreneurs,
(34:31):
and it affects us a citizens. Right. I mean, I
think the structure of our economy has huge consequences for
our day to day lives, and anti trust is a
key component of that. If we saw more robust anti
trust enforcement, then these markets should become more open to competition,
and so in ten years you would not necessarily see Google, Amazon, Facebook, Apple,
(34:54):
Microsoft continue to be as dominant as they are, because
we would have seen innovators. We would have seen the
next wave of of the breakthrough disruptors. I think it's
also worth acknowledging that the dominant these firms enjoy is
certain markets maybe something that we regulate rather than addressed
(35:14):
through breaking up. So since we've started working on Sleepwalkers,
there has been tangible progress, both in the US and
around the world. But this is only the beginning and
we can't relax just yet. In the next episode, we
bring things down to earth and look at how AI
may help us feed the world, and along the way,
we use machine learning to invent a brand new Seltzer
(35:36):
flavor just for us. I'm a veloshen see you next time.
Sleepwalkers is a production of our heart radio and unusual
productions for the latest AI news, live interviews, and behind
(35:59):
the scene in footage. Find us on Instagram, at Sleepwalker's
Podcast or at Sleepwalker's podcast dot com. Special thanks this
episode to the whole team at Weird Moved West, an
incredible production company in El Paso, Texas, who helped us
track down an interview Officer Castillo about his encounter with
the Loon Balloon. Special shout out to j W. Rogers,
(36:22):
Jorge carry On, asail Anya and Leonel Portillo, who voiced
Officer Castillo's translated lines. Sleepwalkers is hosted by me Ozveloshin
and co hosted by me Kara Price, with produced by
Julian Weller with help from Jacopo Penzo and Taylor Chacogne,
mixing by Tristan McNeil and Julian Weller. Our story editor
is Matthew Riddle. Recording assistance this episode from Miguel Paris
(36:44):
and Chris Handbrake. Sleepwalkers is executive produced by me Ozveloshin
and Mangesh Hattikiler. For more podcasts from My Heart Radio,
visit the I Heart Radio app, Apple Podcasts, or wherever
you listen to your favorite shows.
Sleepwalkers News
Hosts And Creators
Oz Woloshyn
Karah Preiss
Show Links
AboutPopular Podcasts
1. Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
2. Stuff You Missed in History Class
Join Holly and Tracy as they bring you the greatest and strangest Stuff You Missed In History Class in this podcast by iHeartRadio.
3. Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations.