Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Washington Post Live's futurist summit, The
0:03
New Age of Tech, is presented
0:05
by Mozilla. You're
0:07
listening to a podcast from Washington Post
0:10
Live, bringing the newsroom to you
0:12
live. Good morning
0:14
and thank you for joining us at
0:16
Washington Post. I am Ela Hezadi, a
0:18
media reporter and the co-host of the
0:20
Post Reports podcast. And I'm
0:23
thrilled to be here, joined with my
0:25
colleagues. We're joined on stage
0:27
by Sarah Ellison, to my left,
0:29
a national enterprise reporter. Pranshu
0:32
Verma, he is an innovations reporter,
0:34
and Kat Sierkreski, who is a
0:36
national technology policy reporter. Thank
0:38
you three for joining us. Thank
0:41
you. As we saw, this
0:43
is a very important year. And Kat, I
0:45
wanted to start with you. Maybe
0:47
it's easy for those of us in the United States to think
0:49
about 2024 being a pivotal year. Actually,
0:52
this is a year in which nearly
0:55
half of the global population will be
0:57
going to the polls. And actually, just
0:59
on Monday, Secretary of State Blinken warned
1:02
of a, quote, flood of
1:04
falsehoods that could suffocate serious
1:06
civil debate during this pivotal
1:08
year. So I wanted to first
1:10
start and ask you to dig in a little
1:12
deeper into what makes this year so different
1:14
than prior election years. Not
1:16
just the fact so many people are going to the
1:19
polls, but what is the landscape that
1:21
people are going to the polls within? Well,
1:23
so I think there's three things that
1:25
make this year really different. The first
1:28
is scale. The tech companies have cut
1:30
their trust and safety teams that are
1:32
responsible for combating disinformation. But at the
1:34
same time, they're dealing with more elections
1:36
around the world than ever before. And
1:39
so that creates a real tension within the
1:41
companies about where do you place
1:43
your resources in order to protect
1:45
democracies. The second is
1:48
conservative backlash. We've seen a series
1:50
of lawsuits from Republicans who are
1:52
concerned that efforts to fight misinformation
1:54
within the government amount to censorship
1:56
of their views. And that
1:58
has really had a... chilling effect on
2:01
a number of government and
2:03
civil society efforts to
2:06
address misinformation. And
2:08
the third I would say is the tech.
2:10
We've been talking all day about artificial intelligence.
2:13
And that creates a host of new challenges.
2:15
Yeah, it's like the tech is evolving so
2:17
quickly at the same time that
2:19
all these other social and political dynamics are
2:22
taking place. And you mentioned misinformation and disinformation.
2:24
And Pranshu, I wanted to ask you about
2:27
the idea that misinformation during election
2:29
years, that's not new. I mean, that's something we
2:31
saw in 2016, 2020. And
2:35
I'm curious what makes it distinct this
2:37
year. And also disinformation,
2:39
which is, I don't know if you would call it
2:41
this, like I think of it like a cousin to
2:44
misinformation. It's not the same. So
2:46
what examples have you seen lately of this?
2:49
Yeah, I think we're still figuring
2:51
out what the broad scale impact
2:53
is. For example, artificial intelligence,
2:55
like Kat mentioned. But
2:58
we are already seeing some very clear
3:00
examples that, yes, AI is better. And
3:02
it's better able to clone your voice,
3:04
your image, your entire likeness. But
3:07
what's happening is that there are really
3:09
cheap and really easy tools that
3:11
anyone can now create a deep fake. And
3:14
so we're seeing specific examples like we
3:16
just talked about. There was the robocall
3:19
of President Biden urging voters not to vote
3:21
ahead of the New Hampshire primary. And
3:24
then if you look abroad, there were
3:26
examples in Moldova. There
3:28
was a liberal party politician.
3:31
And she was deep faked to say that
3:34
she supports a pro-Putin party ahead of an
3:36
election. Now if you go to
3:38
Bangladesh, a conservative Muslim country, there was an
3:40
opposition lawmaker that was deep faked to be
3:42
in a bikini, right? Obviously
3:45
trying to get at some cultural issues there.
3:48
And then of course in Slovakia, right?
3:50
There were people that were on a
3:52
fake audio, a politician, talking about raising
3:54
the price of beer, right? Raising the
3:56
price of beer. Clearly something that voters
3:58
would give a lot. Yeah, very controversial. Yeah,
4:00
well, if you want to take somebody out of a career, tell
4:02
them they're going to raise the
4:05
price of beer. Oh, it's a wrap. That's
4:07
freaking really genius. But none
4:09
of these are like, this isn't some
4:11
massive campaign that is orchestrated that we
4:13
know of by some nation state actor
4:15
to sow discord, where it could be.
4:18
Oftentimes it's a lone political operative. It
4:21
could be a rogue individual or we're
4:24
not sure, it could be a teenager that's just
4:26
bored and then can post this stuff. And
4:28
if it's savvy enough and it's timed well
4:30
enough, we can see that actually it's hard
4:33
to fact check it in the moment where
4:35
systems are not built to fact check it in the moment.
4:38
And are the companies able to hold
4:40
to their promises and stop the
4:42
ability for political propaganda to be
4:44
made through the tools that they
4:46
create? Yeah, and I want to
4:48
get back to that point of whether the tech
4:50
companies are equipped this year to deal with this.
4:54
But as you were talking, it made me
4:56
think, and Sarah, I wanted to ask you
4:58
this, it made me think about the promise
5:00
of social media and a lot of this
5:02
technology was to democratize information and actually help
5:05
democracies around the world. And I mean, we
5:07
can talk about the Arab Spring and how
5:09
lasting that effect that was, but it
5:12
feels like the promise of social media was
5:14
to be in support of democracy and create
5:17
democracies. And I know that you and
5:19
our colleague, Naomi Nix, reported recently about
5:22
how big tech is actually surrendering to
5:24
disinformation. So it raises this question of
5:27
what impact actually these companies and these
5:29
platforms are having to undermine democracy. And
5:32
can you explain how we got there
5:34
from the promise that was there to
5:36
the current conversation we're having
5:38
today? Sure. I
5:40
mean, Kat touched on two big points, which was that
5:43
there were these mass layoffs earlier
5:45
where people really gutted their trust and safety
5:47
teams. And so the people whose job it
5:49
was not to accumulate
5:51
new users, I mean, there are lots of these
5:53
are big companies and there are lots of different
5:56
teams. The people who were responsible for policing this
5:58
kind of thing were really elitible. eliminated from
6:00
the company. And then
6:02
there was this massive pressure campaign where
6:05
people, where right wing
6:07
and conservative politicians were saying, this
6:11
isn't misinformation, this is my belief.
6:13
This is something that I legitimately
6:15
believe, so why is it that
6:17
you're, like stifling
6:19
that. Why is it you're censoring me? And
6:22
then the last thing that we really found in
6:24
the peg for our story was actually when Elon
6:26
Musk bought Twitter, he sort of opened the Overton
6:28
window in social media to
6:31
say we're going to allow all
6:33
these things that had been previously
6:35
disallowed, come back in, you can,
6:37
I mean, specifically, you can misgender
6:39
people, you can do all of
6:41
these things that previously would have gotten you suspended
6:43
from the platform. And
6:45
what that just means is that at this
6:47
point, these social media companies all decided to
6:50
kind of retreat from this effort because they didn't
6:52
want to be the quote unquote arbiters of truth.
6:54
That was sort of a famous quote from Mark
6:57
Zuckerberg. So now what we have is
7:00
the idea that misinformation isn't new, it's really
7:02
in the bloodstream, people really believe certain things,
7:05
and so as a social media company, they
7:07
don't want to be in the position of
7:09
policing those beliefs. And they've really backed away
7:11
from some of these efforts
7:15
that will in fact let democracy
7:17
can thrive on social media and
7:19
anti-democracy can thrive on social media.
7:21
Right, right. That
7:24
makes me wonder about Kat, I'm
7:26
wondering about the incentives too for
7:28
these tech companies to tamp down
7:30
on misinformation and disinformation. If
7:33
the incentive is to have a profitable business or
7:36
what incentives are in place to
7:39
make sure that these sorts of
7:41
pieces of data content aren't spreading,
7:44
like the deepfakes that Pranshu was
7:46
mentioning, and if actually the incentives
7:48
are misaligned there. There
7:50
have been many criticisms that the
7:53
incentives are horribly misaligned with social
7:55
media and misinformation because these types
7:58
of videos that go viral. role
8:00
on the platform that might be false can
8:02
cause people to spend more time looking
8:05
at Facebook, looking at TikTok, and at the
8:07
end of the day they are trying to
8:09
get that type of engagement in order to
8:11
boost their own profits. One of
8:14
the interesting things to watch as we head
8:16
into this election season is what role advertisers
8:18
might play in putting pressure on the companies.
8:20
We have seen in the past at heated
8:22
moments, if you think back to the 2020
8:24
election or when Elon
8:27
Musk bought Twitter, where the advertiser
8:29
said, hey we don't want our
8:31
ad to appear next to these lies
8:33
about the election or hateful content and
8:35
so I'm closely watching this year what
8:37
role they might play again. That's interesting
8:40
because in some ways reflects some of
8:42
the things we've seen in traditional media
8:44
and especially broadcast media in the wake
8:46
of the 2020 election of advertisers
8:49
pulling back but that's a totally different
8:51
business model because they rely on fees
8:54
from cable companies and that sort of thing. Pranshu,
8:57
going back to and picking
9:00
up on this conversation around deepfakes and
9:02
AI generated fake
9:04
content, it's
9:07
creating this environment in which politicians
9:10
are more easily able to dismiss something
9:12
that's real as fake. It's sort of
9:14
altering our perception of reality. This has
9:16
come up with former President Donald Trump.
9:18
I believe there was a recent example
9:21
of that with him. So have
9:24
you seen other politicians around the world take
9:26
that similar approach? Is this the same sort
9:28
of, you know, the fake news 2.0 of
9:30
being able to dismiss things that are actually
9:33
true like a politician going up and giving
9:35
speech and saying this has been manipulated
9:37
and fake? Yeah, this is kind of like
9:39
my favorite rabbit hole to go down and Drew
9:42
Harwell, a wonderful tech reporter told me
9:44
the term is called the Liars Dividend.
9:46
The Liars Dividend. I love it. As
9:50
a concept, but like you mentioned, I do not
9:52
love it in actuality. But as a concept, it's
9:54
really fascinating because you
10:00
mentioned former President Trump, there was
10:02
an ad generated of him by the LinkedIn
10:04
Project of real gaffes that he had made
10:06
on the campaign trail. And the LinkedIn Project
10:09
is an anti-Trump organization. Exactly.
10:11
And these were real events that happened.
10:13
Reporters were there covering it. This got
10:15
mainstream coverage. And his response was, this
10:18
is AI generated. So imagine
10:21
in 2020, the Access Hollywood
10:24
tape breaks, Trump's grabbing a woman
10:26
by a general's, and
10:28
he has the excuse to say, that wasn't real.
10:30
That was AI generated. And so
10:32
what happens is it offers a
10:35
very potentially potent excuse
10:37
to dismiss away real damning information
10:39
that comes out about you. Because
10:41
like we were saying, you used
10:44
to be able to trust that you could hear something
10:46
or you could see something and that something happened. But
10:49
now with AI getting better, there's this kind of convenient
10:52
and cozy excuse here. And
10:54
we've seen it abroad too. In Taiwan, ahead of
10:56
their elections, there was an opposition
10:58
party lawmaker that was seen
11:01
to be depicted going into a hotel with a
11:03
woman that wasn't his wife. And
11:05
so the opposition lawmakers kind of rallied around him to
11:07
say, no, that's an AI deep fake. That actually wasn't.
11:10
We don't know if it's true. We don't know if
11:12
it's an AI deep fake. But what we know now
11:14
is that there's a question about the truth of it.
11:18
And then same thing in India happened
11:20
where there was an audio deep fake,
11:22
potentially of a
11:24
politician of the ruling party saying that there was
11:26
a billion dollar scam that he was a part
11:28
of. Again, the
11:30
excuse was, well, no, that's an AI generated piece
11:32
of audio. And yet none of that,
11:35
it dismisses away the
11:37
scrutiny enough that you're
11:40
now finding a repetition
11:42
of this type of excuse coming through. And
11:44
so it kind of gets to that point
11:46
of like muddying the waters of reality and
11:48
kind of dismissing and easily allowing you to
11:50
kind of play with the concept of truth.
11:52
And I think that that's actually probably going to
11:54
be in this election cycle, one of the biggest
11:58
one of the biggest things that doesn't really. I mean, yes,
12:00
it's going to matter how good the deepfakes are. Once
12:03
people start to believe or
12:05
not believe what they're seeing, or
12:08
just to create the doubt. Say, well, we don't know
12:10
if it's true or it's not true. People
12:12
can throw up their hands and say, well, they're
12:14
all lying. Or there's this sense of the way
12:16
people are
12:19
going to receive this is just to
12:21
kind of zone out and say, you
12:23
can't believe anything. Right. It almost mirrors
12:25
the way that the traditional press or
12:27
the more mainstream press was villainized in
12:29
previous election cycles as well. That
12:31
you can just sort of throw your hands
12:33
up or like, well, I don't know what to
12:35
believe. Or I certainly don't believe these reputable
12:37
people, because they're disreputable now. And I've been told
12:39
not to believe them. And so if
12:42
you're looking for a touchstone as a voter, it's
12:45
much harder to find one. Well, it's
12:47
interesting because in the introduction,
12:49
we saw Metas,
12:52
Nick Clay, Googles, Sundar, Pichai, and OpenAI,
12:54
Sam Altman. They were all saying that
12:56
they want to be part of the
12:58
solution, not part of the problem, when
13:00
it becomes to election deepfakes specifically. But
13:02
given what we were talking about, a
13:04
lot of these tech companies have gutted
13:06
these teams that dealt with a lot
13:08
of the misinformation and making
13:10
sure that at least what's on their platform
13:12
isn't totally false. That
13:14
those safety rails aren't there the way they were before.
13:18
And it's now the technology is advancing so
13:20
fast. So is the cat almost out of
13:23
the bag? Where can these tech companies do
13:25
anything to control this problem at this point?
13:28
I mean, I would be interested to what you guys
13:30
think. But it seems like
13:32
theoretically, the technology
13:34
would be there to identify,
13:36
in most cases,
13:38
a deep fake. But the
13:41
will to do that, also what exactly
13:43
constitutes a deep fake that needs to
13:45
be policed. That's like
13:47
a philosophical conversation. Yeah, and these
13:49
companies have lots of different priorities.
13:53
If you want to show a deep
13:56
fake of Joe Biden drooling or of Donald Trump,
13:58
I mean, some of that You could say
14:00
that's political satire. We're just, it's like a cartoon.
14:03
And then you have a whole debate about whether
14:05
or not that's an appropriate use of AI. And
14:09
you would have, I'm sure that Meta and all
14:11
those companies would be debating that at the policy
14:13
level. And we've seen them sort of fall down
14:15
and be really susceptible to political
14:17
pressure when it comes to those questions. They are
14:20
companies. They want to make a profit. And they
14:22
want to stay out of controversy. Yeah. There's
14:24
one other company we haven't talked about,
14:27
or platform, and that's TikTok. Yeah.
14:30
And I want to ask you this, Kat, that
14:32
there are roughly, give
14:34
or take, 150 million Americans on TikTok.
14:36
It's a huge number. Maybe you're not
14:38
on it. Maybe your mom is on
14:40
it. My mom is on
14:42
it. I think moms and teenagers are on TikTok.
14:46
Big populations of people are on TikTok.
14:48
And then, you know, Congress, the House
14:51
of Representatives, just passed a bill that
14:53
could lead to a ban on TikTok.
14:55
And I'm curious about the power of
14:57
TikTok within politics to shape our politics,
15:00
because there's also been reporting
15:02
that Trump's allies
15:04
have told him and have counseled him
15:06
that going against TikTok would hurt him
15:08
politically. Meanwhile, we do have the White
15:11
House and President Biden backing this effort.
15:13
So what is the power of TikTok
15:15
right now in our politics? So TikTok's
15:17
power has only grown in our politics
15:19
with every election cycle. It's become more
15:22
and more popular. And as you just
15:24
mentioned, this initially was looked at as
15:26
an app that was just for teen
15:28
dance crazes. But now, you have moms.
15:31
You have significant voting blocs that are
15:33
spending time on TikTok. So that makes
15:35
it more important for campaigns. And I
15:37
think it's putting politicians in a really
15:39
tricky spot. We're seeing this with the
15:42
Biden administration right now. President
15:44
Biden's campaign has joined TikTok. And
15:46
they've made videos there. They've also
15:49
developed partnerships with TikTok influencers in
15:51
order to promote their policies. But
15:54
at the same time, we see
15:56
this growing concern about the app's
15:58
ties to trade. China and
16:00
politically a need to show we're tough on
16:03
China, we're going to take a stand against
16:05
this app. And so you see these politicians
16:07
kind of weighing both sides of
16:09
that coin. Yeah,
16:12
it's interesting because at the same time, I
16:14
don't know if you were
16:16
tracking this, but when this bill was
16:18
being debated in Congress, the app had
16:20
these pop-up messages where users were being
16:23
told to contact their representatives and members
16:25
of Congress, their offices were being flooded
16:27
with all these calls. And it's also
16:29
putting them in this, I wonder if
16:31
this is a year where that more
16:34
than meta, more than X
16:36
is going to be the topic of conversation
16:38
among politicians and to be put on the
16:40
spot of where they stand. I don't know,
16:42
Kat, if you're already seeing indications of how
16:44
this is going to play out. We are
16:46
seeing that. I mean, I was talking to
16:49
congressional aides that day that the pop-ups were
16:51
going out, and they were saying, we have
16:53
to shut down our phone lines because we're
16:55
getting so many calls about TikTok. And so
16:57
we know that this is an issue that
16:59
motivates young voters. I think even
17:01
dating back to more than a year ago
17:04
when the House of Representatives was having hearings
17:06
on this, the Secretary of Commerce came out
17:08
and said, this is a sure way to
17:10
lose voters under the age of 35 if
17:13
we support this legislation. And so I
17:15
think the Trump campaign is realizing that, and
17:18
that's why we've seen them back away from
17:20
some of the rhetoric we had in 2020.
17:23
Right. Pranshu, I wanted
17:25
to also bring up a piece of
17:27
reporting that you've done as well, a
17:29
series called Rising India Toxic Tech. And
17:32
that detailed the vast campaign
17:34
online, the digital campaign by
17:36
Hindu nationalists to stoke conversation
17:39
and discourse in India. If
17:41
you were to sum up what your big takeaway is
17:43
walking away from that reporting and what you found,
17:46
and also given the Indian
17:48
election is upon us as well, how
17:50
do you expect AI to be involved
17:52
in that country's Yeah,
17:55
the Rising India series, which was led by
17:57
our beer chief Jerry and Joe Mann and.
18:00
and Karishma Marhotra, Anand Gupta, and
18:02
Smurfs, it was a pretty
18:04
broad scale finding that the
18:08
Bharti Jantapati, which is the ruling party in
18:10
India, which Modi heads, is uniquely
18:13
adept at using social media
18:15
and artificial intelligence to sow
18:18
discord. There is
18:20
content that is hateful. There is content
18:22
that targets minorities. There is content that
18:25
blatantly lies in propaganda. It
18:27
comes in vast amounts, and there's an orchestration
18:29
within the government to spread this on
18:32
various platforms. But the
18:34
market of India itself is so lucrative
18:37
to tech companies that you often
18:39
find the policing of this content
18:43
is either turning. There's a blind
18:45
eye that's turned. It's willful ignorance.
18:47
Or there are actually demands that
18:49
some tech companies have made, or
18:51
the Indian government has made of tech companies that they
18:53
cat-tat, too. And so you see
18:56
India kind of charting this course of when
18:58
you are a powerful market with hundreds of
19:00
millions of users for social media tools, how
19:03
to use that to get your way to stifle
19:05
some sorts of messaging but let others kind of
19:08
grow is kind of the
19:10
big takeaway. And whether it's WhatsApp messages
19:12
or whether it's the type of Bollywood
19:14
movies that get made by Netflix and
19:16
Amazon, if they're Bollywood movies that are
19:18
critical of the Modi administration, we've found
19:20
that those scripts just kind of go away. So
19:23
it's in various elements. But
19:26
now you said April 9 to I think April 19 to June 1, we
19:30
have the Indian elections coming up. And
19:32
it's a very interesting hotbed for AI
19:35
because India has many languages. And
19:39
so you're seeing AI audio kind
19:41
of allow Modi's voice to be translated into
19:43
multiple languages that he doesn't speak. So that's
19:45
a utility that they've acknowledged. You're
19:47
also finding that there's a full scale
19:49
acceptance of AI amongst some political parties
19:52
to create memes, to create really crazy
19:54
and outrageous memes that maybe might
19:56
not be created, but it's a way to cut
19:58
through the political messaging. Old school. means don't work,
20:00
but these AI generated means kind of do.
20:02
And so you're seeing now, you're going to
20:04
see a flood of content, I think, that
20:06
shows like the full born acceptance of it.
20:09
And then some of these other things that we might
20:11
see, these rapid deep fakes that come up and how
20:14
are these things in the moment fact checked, is it
20:16
even possible? Yeah. So
20:18
before we leave, because we only have a couple
20:20
of minutes left and thinking about, you know, coming
20:22
up these few months that I'm imagining is something
20:24
you're going to be monitoring Pranshu. And I wanted
20:26
to ask Kat and Sarah as well, looking into
20:29
the coming months and in this coming year, what
20:31
is one big unanswered question that you have or
20:33
something that you're going to be paying very close
20:35
attention to in this space? Kat, I'm wondering if
20:37
you can start. I would say
20:39
the Supreme Court. One thing we're
20:42
seeing right now is there is
20:44
this litigation challenging the communication that
20:46
governments have with the social media companies.
20:49
And that could really disrupt a lot
20:51
of the work and planning we've seen
20:53
around elections and past cycles. And so
20:55
I'm watching to see where the Supreme
20:58
Court draws a line to see if
21:00
those types of efforts to fight misinformation
21:02
can continue. Yeah, Sarah. I
21:06
think it's proper that we're all talking
21:08
about the supply of misinformation and disinformation
21:10
and how it gets made, who's making
21:12
money off of it. But I've been
21:14
thinking a lot about the demand for
21:16
misinformation, why it hits and who, why
21:20
people are susceptible to it and
21:22
why they want to consume it. And
21:24
I think that it's not just because people
21:26
are fully disaffected and sitting in diners in
21:28
middle America, which is sort of with our
21:30
2016 answer to that kind of
21:32
question. I think that there's something that we can
21:34
really think about, like what really
21:37
gets people to believe a piece
21:40
of misinformation. And you and I worked together
21:42
for a long time in the media group
21:44
looking at traditional media. And I've
21:46
been talking to misinformation researchers that
21:48
social media is clearly very important.
21:51
But the way that the messages that
21:53
surface on social media get translated and
21:56
amplified and used in traditional
21:58
media is in. some ways
22:01
as if not more important. And so I'm
22:03
really also looking at the way these two
22:05
different worlds kind of communicate with one another.
22:07
Yeah, that's fascinating. Well, Sarah, Pranshu, and Kat,
22:10
thank you so much. We'll have to leave
22:12
it there, but thanks for joining us today.
22:23
Good afternoon. I'm Jonathan Capehart, Associate
22:25
Editor here at the Washington Post
22:28
and here with me Adam Brie,
22:30
CEO of, we're gonna talk about
22:32
this in a minute, of
22:34
Skydio, the leading joint manufacturer
22:36
in the United States. Adam, welcome to the
22:39
Washington Post. Thank you, great to be here.
22:41
So I sort of giggled in saying the
22:43
company name because I asked you how do
22:45
you pronounce it? So how do you pronounce
22:47
the company name? Skydio. Skydio. I
22:49
speak Italian, so when I saw it I said, oh, Skydio.
22:53
And in Italian, Dio means God. So
22:55
they're, oh, how clever. Sky God.
22:58
Was that on purpose? That was
23:00
unintentional. So Skydio comes from Sky
23:02
Studio. We were thinking
23:04
Sky Studio and you Google Sky Studio and you get a
23:06
million hits and when we were a three-person startup there was
23:08
no way we were gonna be at the top of that.
23:10
But Skydio was more unique and we found
23:13
out three months later that it also meant Sky God.
23:16
And so are you in the
23:18
Italian market? Not yet. Take
23:21
it from me. If you go to the Italian market,
23:24
Skydio. Yeah, we already, we know what our launch campaign
23:26
is gonna be there. So you co-founded
23:29
Skydio ten years ago, first producing
23:32
consumer drones and then designing
23:34
drones for military and corporate use. Talk briefly
23:36
about this evolution and how quickly was that
23:39
evolution? So the big bet that we made
23:41
when we started the company was that drones
23:43
can be useful for a really wide range
23:45
of tasks but a fundamental
23:48
limiter at the time and largely still
23:50
today is needing to have an expert pilot there
23:52
flying the drone. So when we
23:54
started we went all in on computer vision
23:56
and AI and this was in 2014 before
23:59
most people we thought these things were cool. We thought they were
24:01
cool. We thought that was gonna be a really big deal. So
24:03
the big bet that we made is make the thing
24:05
smart enough to fly itself, build in the skills of
24:08
an expert pilot, which just makes it more useful to
24:10
more people in more ways. And
24:12
we felt like the consumer market was probably gonna
24:14
develop first. That turned out
24:16
to be true. So our first products were consumer
24:18
oriented, but we always felt like the Consumer Product
24:21
Foundation was kind of the right
24:23
starting point for other kinds of applications. The
24:25
fundamental thing that our products do is put
24:28
sensors in important places to capture
24:30
useful information. You can think of it like a flying
24:32
camera. So the consumer platform did
24:34
in fact turn out to be a really useful starting
24:36
point for all the things that we do now across
24:38
a really wide range of industries. Construction,
24:41
public safety, energy utilities, all
24:44
the way up to national security and defense
24:46
and having soldiers use these things on the
24:48
battlefield. So since we have this here, so
24:52
this is a Skydio drone. Yeah. And
24:55
this is the closest I've ever come to actually
24:57
being near a drone. Is it heavy? Pick
24:59
it up. It's about four and a half
25:02
pound. Huh. And
25:04
how many different types of drones do
25:06
you have? So we have a
25:08
few different models, but this is our flagship drone. This
25:11
is the Skydio X10. We launched this at
25:13
the end of last year. And
25:16
this brings together a lot of really amazing
25:18
stuff. I sometimes say that drones are like
25:20
the Mount Everest of technology, but they have
25:22
every piece of wireless cameras,
25:25
optics, vibration, thermals, aerodynamics. And
25:28
then for our drones, a bunch of artificial intelligence
25:30
and machine learning built in as well. So
25:32
DJI, a Chinese company, it controls
25:35
more than 70% of
25:37
the global market, drone market, but in 2021, the
25:40
US government put the company on
25:42
an investment blacklist. How
25:44
has that impacted Skydio's growth? So
25:46
I would say that there's two big themes
25:48
in the market that we think are really
25:50
important. The first is the transition to AI
25:53
and autonomy. And with a lot of organizations,
25:55
drones have reached a point where they've proven
25:57
that they can be really valuable. Energy.
26:00
utility might have set up a drone program. They've
26:02
got somebody who's an expert drone pilot who started
26:04
bringing their drone to work. They're proving that they
26:06
can capture useful data. But then the question is,
26:08
how do you scale that? And
26:11
training everybody to be an expert pilot or
26:13
hiring expert pilots is very difficult. And so
26:15
AI and autonomy makes these things more useful
26:17
to more people. The other big
26:19
trend, I think, is cyber and national security.
26:22
And most of the drones in the market today are
26:24
made by Chinese
26:27
companies. And I think it's just become clear over
26:30
time that that's fundamentally
26:32
untenable for a lot of critical
26:34
industry applications to have drones that,
26:37
at the end of the day, are going to be
26:39
beholden to Chinese national security policy, which is mostly
26:42
in opposition to our national security. And
26:46
so there's a very strong
26:48
need in the market for secure
26:50
alternatives. And that
26:52
benefits us. But
26:55
I would say more than that, we
26:57
feel a responsibility to deliver because
27:00
it's important for our customers.
27:02
And I think it's important for our country. So
27:05
right now, your drones are being
27:07
used in real world situations. In
27:09
fact, there are hundreds of your
27:11
drones in Ukraine right now. And
27:14
we have some photos that were taken from a
27:16
Skydio drone in Ukraine.
27:19
I don't know which monitor, right behind
27:21
us. Can you explain
27:23
what we're looking at? So this
27:25
is work that we've done
27:28
with USAID in collaboration with
27:30
Ukraine's Office of Prosecutor General. And
27:33
it turns out, over the last few months,
27:35
Skydio drones have really become a critical tool
27:37
for documenting Russian war crimes. So we have
27:40
a product called 3DScan, where you
27:42
basically just tell the drone, here's the thing that I
27:44
want to create a digital copy of. And it will go
27:46
off and do it. And
27:48
at this point, they've scanned over 100 civilian structures
27:50
that have been attacked by the Russians.
27:52
And they're using it to basically gather
27:54
evidence and make a case. And
27:57
within the last couple of weeks, they've gotten a couple of
28:00
individuals. indictments in the International Criminal Court based
28:03
on evidence collected with our drones. So
28:05
this is just one application, but I think that I
28:08
was actually in Ukraine two weeks ago. It's
28:12
really a window into the future when it
28:15
comes to drone use. And
28:18
I think it's very clearly the
28:20
future of conflict and war, but I
28:22
think that it's also the future for
28:24
other industries because the stakes are so
28:26
high there, they have to use drones.
28:29
I mean, on the front lines, they don't do anything without putting
28:31
a drone in the air. They're using them to find targets. They're
28:33
using them to keep their own troops safe. They're
28:35
using them at just massive, massive scale, consuming 10,000
28:38
plus drones a month. So
28:42
as we were showing the pictures there, you
28:44
notice, everyone noticed that we stopped on this
28:46
one particular picture. And this one looks very
28:48
different from all the other ones. Can you
28:50
explain what we're looking at here? Yeah,
28:53
I mean, so this was a civilian structure that
28:55
was attacked by the Russians. And
28:57
basically, they're using... This is a 3D
28:59
model that was created. So the drone
29:01
went out and captured hundreds
29:03
of photos of the structure. And
29:06
then there's a software process called photogrammetry where
29:08
those photos are stitched together to create a
29:11
metric 3D model that you can use to
29:13
take measurements, you can use to capture different
29:15
angles. And so in their
29:17
case, they're using it to figure out what
29:19
kind of munitions were used, which angle those
29:21
munitions came in from as part of
29:23
building a case. And it's actually worth noting that
29:25
this kind of thing is a major use case
29:28
for public safety agencies in the US. They use
29:30
these things for crime and accident scene documentation. This
29:33
is just a very, very high stakes and really
29:35
tragic example. So as you just mentioned a moment
29:37
ago, you were in Ukraine. What's
29:40
your takeaway after having seen the front
29:43
lines and how your drones are being used in
29:45
person? So as
29:47
I said, I think it's really a
29:50
window into the future. And I think
29:52
it unfortunately highlights this tension
29:55
and the necessity of having secure
29:58
alternatives to Chinese. drones. So the
30:00
Ukrainians are still heavily dependent on
30:03
Chinese drones they're using and Chinese drone
30:05
technology. They're still using them at massive
30:07
massive scale but it's a super fragile
30:10
dependency because these drones and the companies
30:12
that make them are ultimately hostile to
30:15
their interests. So you
30:17
know China has put in place export controls to
30:19
try to make it harder for the Ukrainians to
30:21
get these drones. The companies themselves
30:23
have put in place firmware things that make them
30:25
less useful to the Ukrainians so they constantly have
30:27
to hack the firmware to get them to do
30:29
what they need and
30:31
I you know
30:33
I see this as just a significant
30:36
challenge to us and the rest of
30:38
the US industry to
30:40
step up our game and
30:43
deliver products that are useful
30:45
for them and you know the impact isn't
30:47
zero today but it's not
30:49
as much as it should be and this
30:51
is becoming a major major focus for myself
30:55
and Scott you know we've delivered thousands
30:57
of our drones to the US and
30:59
Western allies but where it matters
31:01
most on the battlefield in Ukraine I
31:03
think that that's you know that's the proving ground and
31:06
and that's really what we need to be focused on
31:09
both for their sake but I think ultimately for ours as
31:11
well. So this drone here
31:13
and the drones are being used in
31:15
Ukraine are as you mentioned before being
31:17
used to in addition to
31:19
all the other things you talked about document war
31:21
crimes but there are also there are other drones
31:24
that are being used as weapons of
31:27
war and America has a storied history
31:29
with the use of armed drones to
31:31
kill enemies in foreign countries how do
31:33
you reconcile the products you're developing with
31:36
the moral and ethical concerns of drone
31:39
warfare? Yeah it's a great question and something that
31:41
we think a lot about so the
31:44
Ukrainians are using all different kinds of
31:46
drones I mean they have like wing
31:48
drones, quadcopter drones, multirotor drones, surveillance drones,
31:50
decoy drones and then
31:52
they also have a category of they
31:55
call them FPV kamikaze drones so FPV means
31:57
first-person view it means you fly the drone
31:59
by wearing goggles that show you what
32:01
the drone sees. And this
32:03
is one of the primary methods that they're using
32:05
to deliver strikes now, where you fly an FPV
32:08
drone, you put a munition on it, and
32:10
you use it to go and find a target. So
32:12
that is not something that we make. We are very
32:14
focused on surveillance
32:17
drones. So this is in the military, they
32:19
call it ISR, intelligence surveillance reconnaissance. So our
32:21
drones are really designed to make it really
32:23
easy to see what's happening, get better information,
32:26
and make better decisions. And
32:28
I fundamentally believe
32:31
in the mission of the US military and
32:33
our allies, they have very difficult jobs to
32:35
do. And I think that giving them tools
32:37
that give them better information to
32:39
make better decisions is really important, good work.
32:42
I think there are tricky
32:44
questions as you start getting more and
32:46
more automated systems that
32:49
can deliver strikes, lethal strikes.
32:52
And this is something that the world is going
32:54
to have to grapple with. It's
32:57
not something that we're focused
33:00
on, because we're focused on the information gathering drones.
33:02
But that is still part of the kill chain.
33:04
I mean, that's part of the equation of having these
33:06
automated systems. And it's something that
33:08
we're getting increasingly thoughtful about as a company. But we're
33:11
not alone in it. The US military thinks about this
33:13
stuff a lot. And
33:15
they're increasingly refining their policies around
33:17
weapon systems and weapon systems with
33:19
AI. And I think it's
33:22
going to be a project for the world,
33:24
really, to grapple with. So Adam, let's talk
33:26
about police departments. You work
33:28
with police departments across the United States.
33:30
Explain how drones are being used in
33:32
those departments. So they're used for all
33:35
kinds of things. I
33:37
mentioned crime and accident scene documentation. If
33:40
you have a crash on a highway, it could
33:42
take three or four hours using
33:44
conventional methods for somebody on the ground to walk
33:46
around and perform all the measurements that they need
33:49
to gather the evidence that they care about. With
33:52
a drone, you can do that in 20 minutes.
33:55
You can clear the crash scene faster. It's just
33:57
a phenomenal thing for everybody. They're used for search
33:59
and rescue missions. I think we're seeing a search
34:02
and rescue mission sort of training exercise
34:04
here with Oklahoma City PD using
34:06
one of our drones. Hold on. Are
34:08
those voices coming from the video or is
34:11
something happening? Okay, it's coming from the video. We
34:13
could probably turn off the audio on the video.
34:16
Go ahead. Sorry, Adam. But
34:20
they're also used in the highest stakes
34:22
scenario. You might have a
34:24
situation where you have an armed suspect loose
34:26
in a neighborhood and conventionally
34:28
you'd have officers on an
34:30
eight hour manhunt going door to door, guns
34:33
drawn, looking for something that somebody that could
34:35
be dangerous. And we have
34:37
examples in our customer base increasingly on a daily basis
34:39
where you put a drone in the air. You
34:42
can find the person. You can see if they're armed or
34:44
not. You can guide the officers in. And
34:47
I think that's fundamentally better for everybody involved. You
34:49
know, it keeps the officer safer. It keeps the
34:52
community safer. It even keeps the suspect safer because
34:54
the officers don't have to guess about
34:57
what they're walking into. So
34:59
the notion of drones in policing and especially
35:02
AI enabled drones in policing, I think, can
35:05
get folks rightfully concerned because there's
35:07
certainly potential for misuse,
35:10
abuse, privacy invasion. You
35:13
know, what I would say having worked with a
35:15
lot of agencies over the last few years is
35:17
that our most sophisticated best
35:19
customers take transparency incredibly seriously.
35:21
You know, a lot of them release
35:23
the footage from their drone flights quickly
35:26
after them. And
35:29
in a lot of ways, drones are like
35:31
a flying body camera. Like it creates this
35:33
sort of objective picture of whatever happened. But
35:37
you know, body camera typically can only document
35:39
what happened. Drones can actually affect the outcome
35:41
because you can put a drone into a
35:43
dangerous situation that you wouldn't want to put
35:45
a person to get that information proactively rather
35:47
than reactively and make better decisions. Hmm.
35:53
Yeah, let's talk about it.
35:55
Well, we only have
35:57
four minutes and 30 seconds, so I can't even
35:59
go to... down the rabbit hole I wanted to
36:01
go down. But what do you say to critics
36:04
who believe drones enable police departments to behave like
36:06
big brother? So I don't
36:09
want to live in a world where we have drones
36:11
randomly flying overhead surveilling in our houses. I don't want
36:13
to live in that world. I don't
36:15
want to build that world. And
36:17
we have been very proactive on
36:19
this front. We've engaged
36:21
with public safety community organizations to release
36:24
a set of principles that
36:26
we think lead to the best outcomes
36:28
for their communities, things like
36:31
transparency, protecting privacy. And
36:34
there's a lot of things you can do from a
36:36
technology perspective to help with this. I mean, the
36:38
drone knows where it is. It knows what it's filming.
36:41
It can not point its camera at things that it
36:43
shouldn't be pointing its camera at. It can focus on
36:45
whatever task the officer cares
36:48
about. And
36:50
so I think
36:52
there's some technology things that can be done. I
36:55
think there's practices and best
36:58
practices within public safety
37:00
agencies that can be set up
37:02
and followed. And ultimately,
37:04
these agencies are accountable to
37:06
the communities that they serve.
37:09
There's always some democratic chain up
37:12
through that. And I think
37:14
that that's a really important mechanism.
37:16
And the
37:20
best agencies, I think, really lean into the transparency
37:22
element of it, which I think is really the
37:24
best answer. And what we've seen is that the
37:26
more that a community understands what their officers are
37:28
doing with drones, in general, the more
37:30
supportive of it they become because they can see the
37:32
impact on community safety. When
37:34
you say democratic chain, you mean small
37:36
d, democratic change, meaning people have a
37:38
say. Yeah, I mean like the sheriff.
37:42
We sell to a lot of sheriffs, county sheriffs,
37:44
who are elected directly. So
37:47
you lengthened it earlier to sort of
37:51
like a body cam. And we've seen
37:53
instances where we thought body cams were
37:56
going to be the thing that solved
37:58
everything until we got there. We had
38:00
police involved shootings where, oh my
38:02
god, the body cam wasn't working,
38:04
or the body cam wasn't turned
38:06
on. So in the end,
38:09
who controls the footage that
38:12
the drone picks up? And
38:14
is there any possible way
38:16
that we could see
38:18
the, oh, the drone wasn't recording
38:20
in a high profile situation? So
38:24
this is a newer product, but
38:26
I think it's going to become the
38:28
dominant operating paradigm. We have a capability
38:30
with X10 called Remote Flight Deck, where
38:32
the drone has a LTE modem in
38:34
it. It can be flown remotely. So
38:36
you can have an officer in a
38:38
remote flight center who is, through
38:40
a web browser, controlling the drone. And one
38:42
of the things that that does is
38:45
it gives some time and distance between
38:47
the person who's in the heat of the moment
38:49
on the ground. You've got somebody who's removed from
38:53
that who can be more objective
38:55
and be a guide to the people who are actually on
38:57
the ground. So increasingly, I think
39:00
you'll see the center of control shift
39:02
from the officers who are actually on
39:04
the scene in the moment to folks
39:07
who are helping them and on their
39:09
team. They're police officers themselves, usually, but
39:11
who are operating remotely. And I think
39:14
that separation can be quite valuable in
39:18
terms of the outcomes. And it's all digital stuff.
39:20
I mean, the drone is recording the video. We
39:24
have a partnership with a company called Axon that
39:26
has a digital evidence management system where
39:28
the video gets uploaded. It tracks chain of
39:30
custody. So there's good
39:32
controls in place. And I
39:34
think that the availability of this objective
39:36
aerial video is fundamentally a
39:39
good thing for transparency and
39:41
accountability of police officers.
39:43
Don't be fooled by the clock, because it says we
39:45
have 45 seconds. But I've got two questions.
39:47
So we're going to go over time. And
39:50
you've touched on AI technology. And
39:52
I want you to expand a
39:54
little bit in
39:56
reaction to something that retired Admiral
39:58
James Cerritos recently wrote. in
40:01
the Wall Street Journal. And I'm quoting here,
40:03
the winning side will
40:05
be the one that's developed the
40:08
AI-based decision making that can outpace
40:10
their adversary. This is using
40:13
AI technology in drones. My question
40:15
is, what happens when the country
40:17
that's developing those AI drones en
40:21
masse is Iran or
40:24
China? Do you believe the US is
40:26
currently outpacing other countries
40:28
in that space? So
40:32
I basically wholeheartedly agree with the
40:34
sentiment there. And
40:39
I look at things through a drone lens. We're a
40:41
drone company. If you look
40:43
at the last decade of drones, there's no
40:45
question it has been dominated by Chinese companies.
40:48
But I think the shift to
40:50
AI and autonomy presents an opportunity for the
40:53
US to really stage a comeback. I do
40:55
believe that by and large,
40:57
the US is leading the way in
40:59
AI today, like the most cutting edge
41:02
stuff is happening here. And
41:04
this is fundamentally why we founded the company, because
41:06
we believe that AI and autonomy were the future
41:08
of the industry. And we
41:10
have a technology lead there over our Chinese counterparts.
41:13
And I think that that stuff is going to
41:15
become more and more important. So we
41:18
can't stop other countries from developing
41:20
it. But what we can do is move
41:22
as fast as we possibly can ourselves. And
41:25
I think that that really matters. The stakes are very
41:27
high. So we've talked
41:29
about the power and possibilities of
41:31
drones in warfare and public safety.
41:33
But is there anything that gives
41:36
you pause? Pause
41:40
in terms of how it's? Yeah,
41:42
I mean, yeah. So we watch the Terminator
41:44
movies. That's what I
41:46
mean. Well, the drones, there's a humanity.
41:48
Yeah, there might be some software classes
41:51
inside Skydio called Skynet. Look,
41:55
I think that the I'm
41:57
a technology optimist. I think that if you look at the
42:00
at what our products are doing today. The
42:03
most exciting thing to me, we're building this
42:06
incredibly cutting edge sci-fi level technology. But
42:08
we're really deploying it to the core industries
42:10
that our civilization runs on. We're giving it
42:12
to energy utilities and construction workers and departments
42:14
of transportation. And they're using it
42:17
to do things that would be, where the alternative
42:19
is like a person hanging
42:21
off a rope to go inspect something,
42:23
or not knowing the state
42:26
of an insulator in a substation that could lead to
42:28
an explosion. And so there
42:31
are certainly things that you could be concerned about. And
42:33
we try to think through this stuff. We try to
42:35
be proactive as a company. We try to engage with
42:37
external experts to help us think about our product roadmap
42:39
and whatnot. But I think that
42:42
the impact today is overwhelmingly
42:45
positive. And I still think we're just scratching
42:47
the surface of what's possible. So
42:50
I'm generally very excited and optimistic about
42:52
the impact that technology is having now and where
42:55
it's going. OK, so I lied. I have one
42:57
more question because I got to ask it before
42:59
the last question. And you've mentioned this a couple
43:01
of times. And for those of you who know
43:04
drones and the technology, I
43:06
apologize for the simple-minded question. But a
43:08
couple of times, you talked about expert
43:10
pilots. Yeah. How do
43:12
you become an expert pilot? I
43:15
mean, if I flew airplanes, would I be
43:17
a good candidate to become a drone pilot? This
43:20
is a great question to end on. So
43:22
I'm an expert pilot. An
43:24
expert drone pilot? I am, yeah. So I agree. Just to
43:26
be clear. Yeah. Which
43:28
is actually harder than being a real airplane
43:30
pilot. We could talk about that. Because
43:32
when you're flying an airplane, you're in it. You
43:34
see exactly what it sees. You feel what it
43:36
feels. Whereas when you're on the ground trying to
43:38
fly a drone, you have to sort of project
43:41
yourself into it. So I
43:43
grew up flying radio-controlled airplanes. I took it
43:45
way too seriously as a kid. I actually
43:48
won a couple of national championships. There's competitions
43:50
for drone flying. Most
43:52
people don't know about that. I did not. It is
43:54
as nerdy as you would expect. Absolutely. So
43:59
it takes place. years of practice, like anything else,
44:01
like you can spend as much time as you want on it,
44:03
and the more time you spend, the better you get. And actually,
44:05
so I grew up flying this stuff, and
44:07
then I was starting grad school at MIT around the
44:09
time when you could first take radio control airplanes and
44:11
put computers and sensors on them, and write
44:13
software to get them to do smart stuff. So
44:15
my origin into this was basically trying to take
44:18
the things that I could do as an
44:20
expert pilot and codify them into
44:22
algorithms to get the drone to
44:24
do it on its own. And
44:27
that's sort of the genesis of the technology
44:29
stack that's in our Skydio drones
44:31
now. Hmm.
44:35
After hearing that last answer, I really do
44:37
think you should change the pronunciation to Skydio.
44:41
Adam Brie, thank you so much for joining us
44:43
here today. Thank you. Thank you. Hold
44:46
on one second. Yeah,
44:48
this is the closest we'll get to
44:51
TV at a newspaper. That concludes today's
44:53
Futurist Summit. Whether you joined us here
44:55
at the Washington Post or online, thank
44:57
you for being part of a fascinating
45:00
day of conversation. For those in
45:02
the audience, please join us in
45:04
the lounge for lunch. I'll
45:06
be out there with a number of my colleagues from The
45:08
Post across the newsroom. So
45:10
look for the sign Meet The Post. We
45:12
look forward to meeting you. Thank you again.
45:18
Thanks for listening. For more information
45:20
on our upcoming programs, go to
45:22
washingtonpostlive.com. Is
45:25
there anything more satisfying than finding something that
45:27
perfectly lines up with your taste and checks
45:29
all the boxes? Like getting
45:31
the perfect fit with a suit from Indochina.
45:34
They're. Seats are made to measure and
45:36
totally customizable with endless auctions. From.
45:38
Timeless classic to bold statement. you
45:40
can express your style exactly how
45:42
you what. Choose. Your own cut.
45:45
Fabric. Lining buttons, lapels,
45:47
and more. all
45:50
at a surprisingly affordable price. They
45:53
also offer fully customizable blazers,
45:55
pants, outerwear, women's wear, and
45:57
more. Every Indochina piece.
46:00
is made to your exact measurements, and they make
46:02
getting measured easy. Simply set
46:04
up your measurement profile in less than 10 minutes.
46:07
You can send your measurements online from the comfort of
46:09
your home, or make an appointment at one of our
46:11
showrooms. Find the perfect
46:13
fit with Indochino. Go to
46:15
indochino.com and use code PODCAST to
46:17
get 10% off any purchase of
46:19
$3.99 or more. That's
46:22
10% off with code PODCAST.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More