Articles, Blog

Are self-driving cars the future? | VPRO Documentary

Are self-driving cars the future?  | VPRO Documentary


This data is incredibly valuable
so everyone is jumping on it. Cars will be used as Trojan horses
to get their hands on data. If you’re still naive now, you have
learned nothing from the past 25 years. Transport will become
like running tap water. It does cost a bit of money
but we don’t know how much exactly. We don’t think: This is expensive,
I’ll put less in my glass. That’s how cheap transport will become,
like tap water. With a self-driving car you don’t have
to worry about when to start driving. The car knows. It’s not that hard. It’s very useful in the city,
with all the traffic jams. It recognises the turn.
That’s not hard. I will get a bit closer
or I won’t make the light. It’s a very comfortable feature. It’s nice that the car can overtake
in a traffic jam or at a light. I couldn’t go without this feature now,
but it couldn’t go without me either. You have to be able to intervene.
It does silly things sometimes. Like when I overtake a truck which
then goes slightly left, it hits the brake. You have to be ready to prevent that,
or it would be dangerous. The car could become more popular
instead of the other way around. For one, it will be cheaper. The biggest effect of electric cars
is not that they’re cleaner… …but that even without a subsidy,
it will become the cheaper option. There’s no maintenance,
and it will last longer. The only expensive part is the battery,
which will become cheaper too. So cars will be even cheaper
than they already are. Cars will drive part of the journey
themselves. First they’ll eliminate
the most annoying parts… …like driving in heavy traffic
and long distances. If you remove those aspects,
driving will be even more comfortable. So the popularity of cars certainly
won’t decrease in the next decades. You expect there might even be more?
– Yes, I’m afraid so. The remaining problem is that we don’t
really have room for cars in urban areas. So we’ll have to come up
with other ideas to get them out. There’s a trend in many cities
to handle that better. We won’t get rid of them,
but we can be more efficient. Particularly in areas
where they’re unwanted guests. Do you think people will drive more
dangerously once they get an autopilot? Yes, that seems to be the case. People rely too heavily on them,
and this has led to accidents. It can turn left itself if I indicate,
but we’re probably too close. Yes, we’re too close to the car.
Now it’s OK. The wheel is turning left. That’s quite good, isn’t it? Is it addictive, letting go of the wheel?
– Yes, I do it a lot. You can’t let go completely.
It’s not advanced enough. But once you’ve got used to it,
you can’t go without it. This is the heart, or the brain even,
of a self-driving car. What you see here
is lots of computers. They collect all the data of the sensors
in and around the car. Video streams,
data from radars, lidars. It’s all collected and calculated here… …so the car can brake, steer,
and accelerate by itself. Where does all this data go? At the moment,
everything is stored on hard drives. A lot of data,
several gigabytes per second. In total maybe we acquire
over 100 petabytes. It’s too much to send here via Internet. So to gather data worldwide,
DHL delivers an empty hard drive… …puts it in the car,
takes the full one out… …and takes it to Amsterdam to
a big data centre where it’s stored. And all that will be possible
in a tiny box? Not the data collection part, but
calculating the data needed to drive… …will be done by a small device
hidden away neatly in the car. You will still see the sensors, because
self-driving cars need lots of data. Cameras behind the windscreen,
a radar behind the bumper… …maybe something on the roof,
so that will remain. But far more concise than this. It’s telling me that
autonomous driving is available. So I push these two buttons
on the steering wheel… …and the car takes over
the steering and the driving. To teach a network, you first need
to tell it what it’s supposed to learn. So we have to indicate manually
or semi-manually: This is a pedestrian,
this is a mark, this is a sign. We use thousands of images,
so it’s a lot of work. We use this data to teach networks
what pedestrians or signs look like… …and then we can download it
in the car and start driving. To put it bluntly, how ‘stupid’
is the machine right now? The machine is stupid.
It’s not an intelligent machine. But they are capable of driving
automatically in nearly all conditions. But that’s not good enough. To be able to offer this car
to a regular customer…. …we have to make sure
it doesn’t cause accidents. Although humans are good at driving… …25,000 people still die in traffic
every year in Europe. Most people don’t think about that
when they get into a car. It doesn’t have to be your mistake,
it could be someone else’s. Driving carries risks. 25,000 deaths is a high number. It’s equivalent to several
plane crashes a week. Some people have a fear of flying… …but the taxi drive to the airport
is more dangerous than the flight. Safety is a top priority
for self-driving vehicles. It would be unacceptable
for these cars to cause accidents. And you have to factor in
safety margins. You must stick to traffic regulations
and speed limits. After a while, the more data
and knowledge we have… …we might start taking more risks.
But not to begin with. Self-driving cars have so much potential
in terms for road safety… …more effective road systems,
building cities and infrastructures. We have a collective interest
in finding a good way forwards. That’s what we’re doing. I think we can all agree that
there are too many cars in the city. And that there should be
less cars in general. I think people will be prepared to
share cars or drive electric cars. Strangely, the industry has responded
to the need for less cars… …by saying we need these self-driving,
autonomous cars. I don’t understand the logic of that.
But that’s what we are told. They say: Get used to it, everyone,
this is going to happen. But if you consider the values of
public areas and self-determination… …you could say: No, it’s not. Autonomous cars and systems
tend to force us to do things. We need to display certain behaviour
for the system to act autonomously. So how autonomous
are these systems really? If we want to prevent our life
from becoming a computer model… …we shouldn’t be optimising systems… …but accept that people act a certain
way and systems should adapt to that. In conditioned situations, like
motorways, traffic jams, that’s possible. But you don’t want to condition
your whole country, your reality. That’s the fundamental dilemma
we are faced with. All digital systems require conditioning. We are expected to adapt to them,
to that conditioning. That’s a fundamental decision to make. And what are we condition to,
what are we optimising? It’s based on the concept
that people cause many mistakes… …and machines are better at it.
That’s an invalid argument. The argument that self-driving cars
will cause less accidents… …cannot be substantiated. There isn’t a single situation anywhere… …where autonomous cars drive as many
kilometres as cars driven by humans. So they’re all models, suggestions,
that are being presented as facts. Like in the created, safe environment
in Arizona… …where they go around in circles,
that’s going well. If they did that in a city like Amsterdam,
things would go wrong. Very little has been published
about these test results. There are a few figures regarding
how many accidents or interventions… …occurred per driven kilometre. But they don’t tell you what conditions
the cars were driving in. Were these conditioned environments… …with few cyclists, people,
or dogs walking around… …or was it in the city centre
of The Hague or Amsterdam? That cyclist isn’t supposed to be there.
Officially this is a road. Cyclists in Holland only have a few rules
to stick to, which they generally don’t. Yet the system works well,
because it’s based on unwritten rules. I wonder if we’ll ever be able
to programme those into a car. Particularly because those unwritten
rules only apply to the Netherlands… …and even here differ per city
and change over time. These people are walking in the road
and you don’t want to forbid them. But to overtake him, you can’t keep
a safe distance from these people. This is not really enough space
to pass there. I’ll do it, in this case. You could say the truck wasn’t allowed
to be parked there, but that’s reality. These things must be hard to programme,
these unwritten rules. I think autonomous driving
will get quite far. On the motorway, this car could manage
about 80 or even 90% of the drive itself. Maybe 99 percent, they’re working hard
on it in this rat race that’s going on. But that last percentage might be harder
than the other 99. And it’s undesirable,
because you’ll get overly careful cars… …that follow the law completely,
and they won’t be popular… …because they’ll block traffic. On the motorway you need to keep a
two-second distance to the car in front. This means that when merging, you’d
have to wait for a gap of four seconds. You don’t have gaps like that
in the Netherlands during the day. To get through traffic efficiently you need
to be able to bend the law sometimes. That’s something we should
fundamentally never allow robots to do. Those cars would be too careful
by definition. You could argue that we adapt the rules. But these rules are often hard to define,
unwritten rules. How you can you programme that?
They differ per country and over time. By definition,
these cars can’t handle that. Who owns this data? Initially, the data acquired in the car
belongs to the driver, or the owner. They have the data.
And then as a user… …you can make agreements with car
manufacturers to share that data. Just like you do with your phone. In principle the data is yours, but you
accept that it’s shared with other parties. It will be similar with cars. Does that carry risks like
those of Facebook, Twitter… …of sharing data that can subsequently
be used against you? Could that happen in this industry?
– Yes, there is that risk. The data from your car isn’t that different
from the data from your phone. Those risks are there. That’s why companies like Google
are interested in the car trade. It’s a huge source of data. You will have to consider
who you share data with. We have to make sure the driver knows
if and who he’s sharing data with. Is it anonymous or not?
We’ll have to handle that with care. The problems with mobility will be the
same as with internet and smartphones. Without knowing,
you’re paying with your privacy. The car knows exactly
what you’re doing. Cars will have around 15 cameras,
so having access to 1% of all cars… …means you will know everything
that’s happening in the world. If that data becomes worth more
than it costs to drive the car… …you might get cars for free soon.
And pay with your privacy. Like all those free services
Google offers. Driving a car might be like that. This data is so valuable.
Everyone is after it. That’s why Google are into cars now.
Cars themselves aren’t that lucrative. They’ll use their cars as Trojan horses,
to get their hands on that data. If you are still naive at this point… …you’ve learned nothing
from the past 25 years. No citizen, policy maker
or politician can still be naive. It’s a cowboy environment
where a lot is already going on. And later we’ll all think:
This can’t be right. How did we end up in this situation? It’s not the robots that take over jobs,
it’s the companies behind them. This power issue isn’t raised
often enough in technology. So how do we prevent this? By viewing technology as an
expression of culture and power. Not something that simply
happens to us. Something that happens automatically,
like a force of nature. But technology is something we design. It is owned. By keeping this power issue on the table
when it comes to technology… …and allowing people to take part
in this debate… …we can put an end to this naivety. We can’t allow ourselves
to be naive anymore. It’s telling me to switch lanes,
and I confirm that. And it automatically switches lanes. There are still plenty of errors. I recently swapped cars… …but was able to track my old car
for six weeks. I could have picked it up, or open
the roof while this guy was driving. I didn’t of course, but that was a bit of
mistake on the manufacturer’s part. There more examples of that,
some of which are worrying. It hasn’t got out of hand yet. But it would still be relatively easy
to hack 20 or 30 million cars. Maybe something should go
a bit wrong before we wake up. The concept of building a machine
that does the right thing… …is impossible, because ‘the right thing’
is subject to context. There are not just one
but several moral issues. We will end up with various different
ethical artificial intelligence systems. The only way to protect that
is through a rule of law. It needs to be built in the justice system.
Justice is synonymous with moral. We’ve made such a big effort
to get the car out of the city centre. And now they’re driving themselves
back. That’s a weird image. We’ve said we want to give pedestrians
and cyclists more space… …and cars as little as possible. Many cities are optimised for cars. The difficult thing for me is that I’m
discerning, but not against technology. We should let go of this idea
of being in favour or against. I’m against the power
that companies have… …but in favour of what
technology can offer us. As long as it’s in the hands
of a democratic community. The future of the car industry
and the use of cars and roads… …is very uncertain due to the many
changes that are taking place: China starting to manufacture cars,
electric driving, autonomy… …connectivity and data sharing… …they’re all reasons why the car trade
is going to change dramatically. How do you see it?
– I think the different companies… …in the developing market
will make different choices. Some companies will choose
to take more risks than us. We have a safety culture.
We programme cars cautiously. We test them thoroughly
and build in safety margins. The downside is that it’s not the fastest
car in the road and drives cautiously. The courts will decide who is responsible
and if it’s been done well or not. We’ll have to get through that stage. I think that we as companies as well
as governments and other partners… …have to find the best way through it. When China starts to bring self-driving
cars on the European market… …we’ll have to do our best
to compete with them. Will we follow the Silicon Valley route,
where companies dominate the trade? Where we have to trust that they
have our best interest at heart? Or do we follow the Chinese path,
where the state decides… …how to be good citizens? I think Europe has the right attitude. Here, the collective interest
and values are central. I see Europe acting accordingly,
partly because we have no choice. We’ve missed out on what’s going on
in Silicon Valley. We’re empty-handed.
We’re no big tech at the moment. China is an unattractive perspective. So Europe has its own definition
of what to do with digital infrastructure… …and how we want it to work for us. If that’s your point of departure,
it’s a whole different ball game. I think that cities that make
clear political decisions… …will be able to take part
in these platforms. I do worry about cities
where that political climate is absent. There, citizens are at the mercy
of these platforms. 70 or 80 percent of accidents that occur
now could have been prevented… …with the technology that already exists. We need to work hard to get
more of that technology into cars. The Netherlands is somewhat behind
because we drive older, smaller cars… …which don’t have this technology. That argument is used for self-driving
cars but we can already take steps. Then why would you want to drive
completely autonomously? Maybe to drive kids to school. But taking them off the bike and into a
self-driving car seems a step backward. So maybe the fully autonomous car
is a complicated solution… …that creates a problem. I don’t know
why we’d still want it, if cars were safer. The question is what are you
optimising the system for? I think it would be nice
if we focus on people. That sounds a bit like ‘Team Human’. It’s like coming out of the closet,
saying you’re ‘in favour of people’. You want to improve things
to make people happier and get along. I’m not in favour of optimising a system
or Uber’s share ownership. Those are the two choices you have.

Tagged , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

26 thoughts on “Are self-driving cars the future? | VPRO Documentary

  1. Why don't they have subtitles more on speak different languages English is the language speak English German Nazi crap….

  2. 11:00 – 11:50 Why yes, i also agree that the price of Tuna should be lowered. Man, there is a lot of elven language in this video, cannot watch it, sorry.

  3. I actually speak Dutch, but I'm pretty damn sure I'm in the minority. Don't publish a video without translations. That's not a request, that's an instruction.

  4. Like some factors involved with such technology where long distance travel is concerned.. however, in tomorrow's economy… I do not get the impression that "My Kind" is welcome to share the same space in a new economic system that has evolved into a much more technical and needs fewer people in the labour market. Even if the tech is affordable.. hard to pay the bills if one cannot gain employment to pay the bills and save up for an education in what seems to be a new world order.
    Sure that it's just nerves and all will be well in these ventures with technology and buisness.
    Feels like in such a short lifetime, the education received equals in comparison to the value of a new computer being obsolete after walking out of a brick and mortar store upon purchasing it.
    Leaves an individual feeling without worth and value as a human being, and lied to by those institutions we place our trust that one day, after having given our time and labour after receiving an education to provide the system with a much needed service, that one day we can retire and live our lives in peace.
    Much is up in the air with all this, and I think many are afraid of being made redundant and becoming homeless because of such drastic changes to the economy. I certainly don't feel that comfortable myself at this point and time.. but do like the concept of traveling with a autopilot system on the road should I not have the presence of health and mind to do so on long trips, or going to the clinic for an appointment. Would make getting home after a surgery more easier where one may not have a person to drive them home whom they trust avalible, among other areas of concern.

  5. Europe and it's Transport planners are to in bed with motor industry and is falling behind China and the US in autotomy and self-driving shared mobility solutions

  6. People forget the fact that if a car makes a mistake, it can be taught not to make it again and all other cars in the world will have their model updated. Try doing that with humans.

  7. Autonomous vehicles are destined to create a paradigm shift in global surface transport. The most profound change will be the eventual elimination of death and injury from traffic accidents. Traffic congestion and total vehicle emissions will be drastically reduced and eventually eliminated. The day is coming when private vehicle ownership will be a thing of the past. People will no longer need to own a vehicle and they will have the ability to call up a driverless ride at any given moment and in a matter of seconds, using a smart phone app. This scenario, as far fetched as it may seem, is not a matter of if, it is imminent and only a matter of when.

  8. Good doc! Really enjoyed how you left the moral issue open because nobody knows and not many are thinking about, great stuff to ponder!

  9. Around the 19 minutes mark and onwards there are a some sceptical voices that make the same tired fallcies we've heard since Kasparov was beaten by Deep Blue, it was hard to sit through:

    "The idea that self driving will be safer than human drivers has not been substantiated." True, because there are no large scale deployments of self driving vehicles. We can infer, however, that unless there's some unknown law of physics that makes it so that human beings represent the best possible drivers allowed for by the Universe, we might be able to create better ones. Once we have those we will likely be able to gradually improve on them, and improve on the newer model further and so on… In 1992, it could not be substantiated that the internet would revolutionize the world, but most thinking people could infer that it would.

    "Sure, the robots will be able to handle easy tasks like driving down an empty highway, but never complicated ones like safely navigating a city full of pedestrians." Let's look at how that argument has played out historically:

    "A computer can do simple repetitive tasks, but never something computationally demanding like chess." "Oh crap!"
    "Sure, they beat us at chess. but they'll never be able to do uniquely human things like understand spoken language." "Oh crap!"
    "…but they cannot simulate intuition, so they'll never beat us at Go." "Oh crap!"
    "Sure, they beat us at perfect information games with discrete positioning, they'll never handle something like Starcraft though." "Oh crap!"
    "But psycology is much to complicated for a machine to grasp, so the best poker player will always be human." "Oh crap"

Leave a Reply

Your email address will not be published. Required fields are marked *