and I make work about technology.
I do things like draw life-size outlines of military drones
in city streets around the world,
so that people can start to think and get their heads around
these really quite hard-to-see and hard-to-think-about technologies.
I make things like neural networks that predict the results of elections
what the actual possibilities of these weird new technologies are.
Last year, I built my own self-driving car.
But because I don't really trust technology,
I also designed a trap for it.
(Laughter)
And I do these things mostly because I find them completely fascinating,
but also because I think when we talk about technology,
we're largely talking about ourselves
and the way that we understand the world.
So here's a story about technology.
This is a "surprise egg" video.
It's basically a video of someone opening up loads of chocolate eggs
and showing the toys inside to the viewer.
That's it. That's all it does for seven long minutes.
And I want you to notice two things about this.
First of all, this video has 30 million views.
(Laughter)
it comes from a channel that has 6.3 million subscribers,
that has a total of eight billion views,
and it's all just more videos like this --
30 million people watching a guy opening up these eggs.
It sounds pretty weird, but if you search for "surprise eggs" on YouTube,
it'll tell you there's 10 million of these videos,
and I think that's an undercount.
I think there's way, way more of these.
If you keep searching, they're endless.
There's millions and millions of these videos
in increasingly baroque combinations of brands and materials,
and there's more and more of them being uploaded every single day.
Like, this is a strange world. Right?
But the thing is, it's not adults who are watching these videos.
These videos are like crack for little kids.
There's something about the repetition,
the constant little dopamine hit of the reveal,
that completely hooks them in.
And little kids watch these videos over and over and over again,
and they do it for hours and hours and hours.
And if you try and take the screen away from them,
they'll scream and scream and scream.
and I've already seen people in the audience nodding --
if you don't believe me, find someone with small children and ask them,
and they'll know about the surprise egg videos.
It's 2018, and someone, or lots of people,
are using the same mechanism that, like, Facebook and Instagram are using
to get you to keep checking that app,
and they're using it on YouTube to hack the brains of very small children
in return for advertising revenue.
At least, I hope that's what they're doing.
I hope that's what they're doing it for,
because there's easier ways of making ad revenue on YouTube.
You can just make stuff up or steal stuff.
So if you search for really popular kids' cartoons
like "Peppa Pig" or "Paw Patrol,"
you'll find there's millions and millions of these online as well.
Of course, most of them aren't posted by the original content creators.
They come from loads and loads of different random accounts,
and it's impossible to know who's posting them
or what their motives might be.
Does that sound kind of familiar?
Because it's exactly the same mechanism
that's happening across most of our digital services,
where it's impossible to know where this information is coming from.
It's basically fake news for kids,
and we're training them from birth
to click on the very first link that comes along,
regardless of what the source is.
That's doesn't seem like a terribly good idea.
Here's another thing that's really big on kids' YouTube.
This is called the "Finger Family Song."
I just heard someone groan in the audience.
This is the "Finger Family Song."
This is the very first one I could find.
It's from 2007, and it only has 200,000 views,
which is, like, nothing in this game.
But it has this insanely earwormy tune,
which I'm not going to play to you,
because it will sear itself into your brain
in the same way that it seared itself into mine,
and I'm not going to do that to you.
So within a few years, these finger family videos
and you get versions in different languages
with popular kids' cartoons using food
or, frankly, using whatever kind of animation elements
you seem to have lying around.
And once again, there are millions and millions and millions of these videos
available online in all of these kind of insane combinations.
And the more time you start to spend with them,
the crazier and crazier you start to feel that you might be.
And that's where I kind of launched into this,
that feeling of deep strangeness and deep lack of understanding
of how this thing was constructed that seems to be presented around me.
Because it's impossible to know where these things are coming from.
Some of them appear to be made of teams of professional animators.
Some of them are just randomly assembled by software.
Some of them are quite wholesome-looking young kids' entertainers.
And some of them are from people
who really clearly shouldn't be around children at all.
(Laughter)
And once again, this impossibility of figuring out who's making this stuff --
Is this a person? Is this a troll?
What does it mean that we can't tell the difference
And again, doesn't that uncertainty feel kind of familiar right now?
So the main way people get views on their videos --
and remember, views mean money --
is that they stuff the titles of these videos with these popular terms.
So you take, like, "surprise eggs"
and then you add "Paw Patrol," "Easter egg,"
all of these words from other popular videos into your title,
until you end up with this kind of meaningless mash of language
that doesn't make sense to humans at all.
Because of course it's only really tiny kids who are watching your video,
and what the hell do they know?
Your real audience for this stuff is software.
It's the software that YouTube uses
to select which videos are like other videos,
to make them popular, to make them recommended.
And that's why you end up with this kind of completely meaningless mash,
But the thing is, you have to remember,
there really are still people within this algorithmically optimized system,
people who are kind of increasingly forced to act out
these increasingly bizarre combinations of words,
like a desperate improvisation artist responding to the combined screams
of a million toddlers at once.
There are real people trapped within these systems,
and that's the other deeply strange thing about this algorithmically driven culture,
you have to end up behaving like a machine
And also, on the other side of the screen,
there still are these little kids watching this stuff,
stuck, their full attention grabbed by these weird mechanisms.
And most of these kids are too small to even use a website.
They're just kind of hammering on the screen with their little hands.
where it just keeps playing these videos over and over and over in a loop,
endlessly for hours and hours at a time.
And there's so much weirdness in the system now
that autoplay takes you to some pretty strange places.
This is how, within a dozen steps,
you can go from a cute video of a counting train
when all of these different keywords,
all these different pieces of attention,
this desperate generation of content,
all comes together into a single place.
This is where all those deeply weird keywords come home to roost.
You cross-breed the finger family video
with some live-action superhero stuff,
you add in some weird, trollish in-jokes or something,
and suddenly, you come to a very weird place indeed.
The stuff that tends to upset parents
is the stuff that has kind of violent or sexual content, right?
Children's cartoons getting assaulted,
weird pranks that actually genuinely terrify children.
What you have is software pulling in all of these different influences
to automatically generate kids' worst nightmares.
And this stuff really, really does affect small children.
Parents report their children being traumatized,
becoming afraid of their favorite cartoon characters.
If you take one thing away from this, it's that if you have small children,
keep them the hell away from YouTube.
(Applause)
But the other thing, the thing that really gets to me about this,
is that I'm not sure we even really understand how we got to this point.
We've taken all of this influence, all of these things,
and munged them together in a way that no one really intended.
And yet, this is also the way that we're building the entire world.
We're taking all of this data,
a lot of historical data full of prejudice,
full of all of our worst impulses of history,
and we're building that into huge data sets
And we're munging it together into things like credit reports,
into things like predictive policing systems,
This is the way we're actually constructing the world today
And I don't know what's worse,
that we built a system that seems to be entirely optimized
for the absolute worst aspects of human behavior,
or that we seem to have done it by accident,
without even realizing that we were doing it,
because we didn't really understand the systems that we were building,
and we didn't really understand how to do anything differently with it.
There's a couple of things I think that really seem to be driving this
and the first of those is advertising,
which is the monetization of attention
without any real other variables at work,
any care for the people who are actually developing this content,
the centralization of the power, the separation of those things.
And I think however you feel about the use of advertising
the sight of grown men in diapers rolling around in the sand
in the hope that an algorithm that they don't really understand
suggests that this probably isn't the thing
that we should be basing our society and culture upon,
and the way in which we should be funding it.
And the other thing that's kind of the major driver of this is automation,
which is the deployment of all of this technology
as soon as it arrives, without any kind of oversight,
kind of throwing up our hands and going, "Hey, it's not us, it's the technology."
Like, "We're not involved in it."
That's not really good enough,
because this stuff isn't just algorithmically governed,
it's also algorithmically policed.
When YouTube first started to pay attention to this,
the first thing they said they'd do about it
was that they'd deploy better machine learning algorithms
Well, machine learning, as any expert in it will tell you,
is basically what we've started to call
software that we don't really understand how it works.
And I think we have enough of that already.
We shouldn't be leaving this stuff up to AI to decide
It'll start censoring other things.
It'll start censoring queer content.
It'll start censoring legitimate public speech.
What's allowed in these discourses,
it shouldn't be something that's left up to unaccountable systems.
It's part of a discussion all of us should be having.
that the alternative isn't very pleasant, either.
YouTube also announced recently
that they're going to release a version of their kids' app
that would be entirely moderated by humans.
Facebook -- Zuckerberg said much the same thing at Congress,
when pressed about how they were going to moderate their stuff.
He said they'd have humans doing it.
And what that really means is,
instead of having toddlers being the first person to see this stuff,
you're going to have underpaid, precarious contract workers
without proper mental health support
(Laughter)
And I think we can all do quite a lot better than that.
(Applause)
The thought, I think, that brings those two things together, really, for me,
It's like, how much do we really understand -- by agency, I mean:
how we know how to act in our own best interests.
Which -- it's almost impossible to do
in these systems that we don't really fully understand.
Inequality of power always leads to violence.
And we can see inside these systems
that inequality of understanding does the same thing.
If there's one thing that we can do to start to improve these systems,
it's to make them more legible to the people who use them,
so that all of us have a common understanding
of what's actually going on here.
The thing, though, I think most about these systems
is that this isn't, as I hope I've explained, really about YouTube.
These issues of accountability and agency,
of the violence and exploitation that inherently results
from the concentration of power in a few hands --
these are much, much larger issues.
And they're issues not just of YouTube and not just of technology in general,
They've been with us for ages.
But we finally built this system, this global system, the internet,
that's actually showing them to us in this extraordinary way,
Technology has this extraordinary capacity
to both instantiate and continue
all of our most extraordinary, often hidden desires and biases
and encoding them into the world,
but it also writes them down so that we can see them,
so that we can't pretend they don't exist anymore.
We need to stop thinking about technology as a solution to all of our problems,
but think of it as a guide to what those problems actually are,
so we can start thinking about them properly
(Applause)
(Applause)
Helen Walters: James, thank you for coming and giving us that talk.
when you think about the films where the robotic overlords take over,
it's all a bit more glamorous than what you're describing.
But I wonder -- in those films, you have the resistance mounting.
Is there a resistance mounting towards this stuff?
Do you see any positive signs, green shoots of resistance?
James Bridle: I don't know about direct resistance,
because I think this stuff is super long-term.
I think it's baked into culture in really deep ways.
A friend of mine, Eleanor Saitta, always says
that any technological problems of sufficient scale and scope
are political problems first of all.
So all of these things we're working to address within this
are not going to be addressed just by building the technology better,
but actually by changing the society that's producing these technologies.
So no, right now, I think we've got a hell of a long way to go.
But as I said, I think by unpacking them,
by explaining them, by talking about them super honestly,
we can actually start to at least begin that process.
HW: And so when you talk about legibility and digital literacy,
I find it difficult to imagine
that we need to place the burden of digital literacy on users themselves.
But whose responsibility is education in this new world?
JB: Again, I think this responsibility is kind of up to all of us,
that everything we do, everything we build, everything we make,
needs to be made in a consensual discussion
with everyone who's avoiding it;
that we're not building systems intended to trick and surprise people
but that they're actually involved in every step in educating them,
because each of these systems is educational.
That's what I'm hopeful about, about even this really grim stuff,
that if you can take it and look at it properly,
it's actually in itself a piece of education
that allows you to start seeing how complex systems come together and work
and maybe be able to apply that knowledge elsewhere in the world.
HW: James, it's such an important discussion,
and I know many people here are really open and prepared to have it,
so thanks for starting off our morning.
(Applause)