Artificial intelligence and algorithms_ pros and cons _ DW Documentary (AI documentary)

FlickFrenzy

Developments in artificial intelligence (AI) are leading to fundamental changes in the way we live. Algorithms can already detect Parkinson's disease and cancer, and control both cars and aircraft. How will AI change our society in the future?  This documentary journeys to the hot spots of AI research in Europe, the USA and China, and looks at the revolutionary developments which are currently taking place. The rapid growth of AI offers many opportunities, but also many dangers. AI can be used to create sound and video recordings which will make it more and more difficult to distinguish between fact and fiction. It will make the world of work more efficient and many professions superfluous. Algorithms can decide whether to grant loans, who is an insurance risk, and how good employees are. But there is a huge problem: humans can no longer comprehend how algorithms arrive at their decisions. And another big problem is AI’s capacity for widespread surveillance. The Chinese city of Rongcheng is already using an AI-supported 'social credit system' to monitor and assess its citizens. Does AI pose a danger to our personal freedoms or democracy? Which decisions can we leave to the algorithms - and which do we want to? And what are AI’s social implications?  A documentary by Tilman Wolff und Ranga Yogeshwar  #Ranga_Yogeshwar  --------------------------------------------------------------------  DW Documentary gives you knowledge beyond the headlines. Watch high-class documentaries from German broadcasters and international production companies. Meet intriguing people, travel to distant lands, get a look behind the complexities of daily life and build a deeper understanding of current affairs and global events. Subscribe and explore the world around you with DW Documentary.

Transcript

00:00Artificial intelligence is making rapid strides.

00:07There's talk of a new evolution that could fundamentally change life on our planet.

00:13Artificial intelligence has the potential to revolutionize every aspect of daily life.

00:18Work, mobility, medicine, the economy, and communication.

00:23But will AI really make medicine better and doctors superfluous?

00:27When will self-driving cars hit our roads?

00:31Will intelligent robots usurp our jobs?

00:34And are we heading for a dystopia with no privacy and total surveillance?

00:40What exactly is artificial intelligence and how much can it really do?

00:46What will change and what will remain pure fantasy?

00:54To answer these questions, we embarked on an exciting journey

00:57to meet the scientists working on our future in the U.S., Britain, Germany, and China.

01:04Our first stop, Silicon Valley in California.

01:26Apple, Google, and Facebook all have their headquarters here.

01:30It's the epicenter of the digital revolution.

01:36The tech industry has changed the face of the San Francisco Bay Area.

01:40New startup companies launch every day.

01:43Rents have exploded and artificial intelligence is the buzzword.

01:48A new type of supermarket recently opened its doors here.

01:52Amazon Go.

01:55All you need here is an app.

01:59Hold your mobile phone to the scanner and you're in.

02:12As Leonardo shows me Amazon's new menus

02:15and explains that the language assistant Alexa can help with the preparation at home,

02:19I'm under constant surveillance.

02:25What shelf do I stop at?

02:29Which products am I interested in?

02:35On the ceiling, sensors, and cameras.

02:39Intelligent image recognition captures my every move.

02:42What do I take off the shelf?

02:44What do I put back?

02:45And what do I take with me?

02:48This branch is still in its test phase,

02:51but Amazon plans to open 50 such grocery stores this year alone.

02:56The end of the sales assistant.

02:58Just walk out.

03:00No more standing in line.

03:01No cashiers.

03:04I feel a bit like a shoplifter as I leave.

03:09Comfort at the cost of privacy.

03:14My receipt.

03:20One block away, a robot cafe.

03:24Another test lab for the future.

03:29Order by app and touch screen.

03:33The increasingly ubiquitous tools of trade.

03:41My first ever cup of coffee served by a robot.

03:48So this is the taste of the future.

03:54AI will change our shopping experience,

03:57but what will happen to employees?

04:01Stanford University is at the forefront of global AI research

04:04with an annual budget of $6.5 billion.

04:13I want to know, how will artificial intelligence change medicine?

04:22Researchers here have developed an artificial intelligence algorithm

04:25that can screen x-rays for certain diseases.

04:28Computer scientist Pranav Rajpurkar shows me how easy it is to use.

04:33Take a picture of an x-ray with your mobile phone,

04:36upload the image, and a few seconds later, you get the diagnosis.

04:41It's a mass, and it's saying this thing over here is a possibly cancerous lesion.

04:47And I can see that right over here.

04:49Okay, so it gives you, if I may have a look at it,

04:52now probabilities for pneumonia, nodule, edema, effusion, and that goes bang.

05:00Yeah.

05:01Now, how does this work? I mean, how did you get to the point?

05:04We started with a large data set of chest x-rays,

05:08which were released by the NIH, and these contained x-rays

05:12and then also labels of different pathologies and whether they existed in those x-rays.

05:17So it might say, okay, here's an image, and on this image, I have pathology 1, 2, and 3.

05:22And we had 100,000 of these images.

05:27So we trained a model that can take it as an input, an x-ray,

05:32and then output the probability of several different pathologies on this x-ray.

05:39Artificial intelligence is modeled on the human brain,

05:43a gigantic network of almost 100 billion interconnected neurons.

05:54Put in very simple terms, this is how a brain cell works.

05:58First, incoming impulses are passed in a domino effect from one neuron to the next.

06:05The resulting circuit connects the neurons,

06:08and it is this circuit that artificial intelligence tries to simulate

06:12as a digital neural network.

06:17Like our brains, the network can learn how to identify tuberculosis, for instance.

06:24First, the network needs to be trained or taught.

06:27X-rays of tuberculosis patients are fed to the system.

06:31Initially, it struggles to correctly identify the condition.

06:37But every time an x-ray is fed in, the network's structure is adapted,

06:42and its diagnostic ability improves.

06:45It takes thousands and thousands of clinical data sets to train the machine.

06:55Only after the network is optimized in this way

06:59can it correctly identify an unknown x-ray.

07:09But how accurate is artificial intelligence compared to a doctor's expertise?

07:14We have actually done this test twice at this point,

07:18once with a set of studies from the NIH data set

07:22that we had a group of radiologists label,

07:25and then we compared the accuracy of the model to the radiologists,

07:29and we found that they were very similar in terms of accuracy on most pathologies.

07:34On one of them, the model was outperforming the radiologists,

07:37and on three of them, the radiologists were outperforming the model.

07:40And then we repeated the experiment, this time using a data set from Stanford,

07:44which we recently released, which is 200,000 chest x-rays.

07:48And then we had a similar setup where we had three subspecialty radiologists.

07:53These are very uncommon, very trained radiologists

07:56to decide what the ground truth for a particular set of images was,

08:00and then we compared general radiologists to the algorithm at the task

08:04and found that they had similar levels of performance.

08:07These are all Stanford radiologists, so they're trained.

08:11They should be good.

08:14Reading x-rays accurately is a complicated process,

08:18but artificial intelligence is making fast progress.

08:22When it comes to identifying or recognizing simple images,

08:26computers have surpassed human accuracy.

08:32If I look at your picture, it's always probability,

08:36so there are cases where the machine is not really sure

08:40what would be sort of a clear decision to say,

08:43okay, this is, I don't know, pneumonia or something else.

08:47Yeah, I think it's good to talk in terms of probabilities

08:51because probabilities also give a sense of the algorithm,

08:56the model's uncertainty on that particular problem.

08:59I think one difficulty with probabilities is that

09:02it does make it hard for humans to interpret.

09:07Like, what does a probability of 88% versus 92% mean

09:11in terms of the decision I should make in the clinic?

09:14And so I think in that sense,

09:16one of the things that we could experiment with doing in the future

09:20is rather than showing probabilities that are so fine-grained,

09:23maybe we can show things like unlikely,

09:26or this pathology is likely, or this pathology is possible.

09:31In health care, artificial intelligence is powering a revolution.

09:35Scientists are using artificial intelligence algorithms

09:39to sift through seemingly banal data,

09:41such as the up-and-down motion of the steps we take every day.

09:45They're looking for conspicuous patterns

09:48that could serve as early warning signs of disease.

09:51Scientists in the English city of Birmingham

09:54are working on a revolutionary diagnostic method.

10:01To date, there are no specific tests to detect Parkinson's disease,

10:05making diagnosis difficult.

10:08A.I. could change that.

10:10Max Little is a mathematician at Aston University.

10:26Voice changes can be an early indicator of Parkinson's.

10:29Max and his team collected thousands of vocal recordings

10:32and fed them to an algorithm they developed,

10:34which learned to detect differences in voice patterns

10:37between people with and without the condition.

10:40In a lab-based study of the recordings,

10:42the algorithm was able to correctly identify a Parkinson's diagnosis

10:46nearly 99% of the time.

10:48Max Little's work is an example of the far-reaching changes

10:52A.I. is bringing to the field of medicine.

10:55It's no longer just doctors who are using artificial intelligence

10:58to develop new diagnostic methods,

11:00but data scientists, programmers, and mathematicians like Max.

11:10One example?

11:11When a person walks, sensors in their smartphone

11:14register the up-and-down motion of their gait.

11:19But what information can be gleaned from such data?

11:23If we measured a pattern of someone's walking behavior,

11:27then someone who's healthy might have,

11:29we might measure the accelerometer to look like that.

11:32Okay, so it's just the sort of movement he would have here.

11:35Yeah, their hips going up and down regularly,

11:37that kind of thing, along with their pace.

11:46But if you looked at somebody with Parkinson's disease,

11:49they may have these small steps like this,

11:51and they may be irregular, or they may have patterns like that,

11:56or they may even freeze and stop like that.

11:58So you can see that there's...

12:00There's a difference.

12:01There's a difference.

12:02So you can also now train an algorithm, for instance,

12:04to pick out features like what is the distance between,

12:08the time distance between these peaks.

12:11And it could also do the same with this,

12:12and it would be able to do that very precisely.

12:15And by doing so, we may be able to measure, for instance,

12:17that here there's large variability between these.

12:21The advantage of the algorithm really comes when the,

12:25for instance, you might have somebody who is, say,

12:29who measures a pattern which looks like this,

12:32and there might just be one small change, perhaps,

12:37that occurs very, very, maybe not like that,

12:41but sort of, you know, some small variation.

12:45That's right, in the sequence of these,

12:48in the timing of these events.

12:50Even to a professional eye,

12:52because they don't have the level of precision,

12:54they may not be able to detect that this is outside of,

12:57say, the normal range of variation.

13:00But, of course, an algorithm connected

13:02to a high-precision sensor will, you know,

13:06will be able to determine that difference.

13:09And in this case, this person here may, in fact,

13:11have a precursor symptom of the disease.

13:15So this would mean that this person,

13:19with the help of an algorithm,

13:21could be diagnosed as having Parkinson's,

13:24whereas the doctor himself would miss him out.

13:29That could, for the first time,

13:31make it possible to detect precursor symptoms of Parkinson's

13:35and enable early intervention.

13:39But what else does the data on our smartphones reveal?

13:44Right now, you have already apps

13:47tracking your so-called activity.

13:50So, in fact, the data might be already there.

13:55Well, the data potentially could be there, that's right.

13:58But there are ethics about whether we collect that kind of data

14:02and use it for these sorts of purposes.

14:05Now, clearly, we can't just collect this data

14:08and start diagnosing people.

14:12We should not.

14:13We should not, no, absolutely not.

14:15But we could.

14:16We could, but we really wouldn't want to.

14:19There are very good reasons not to do it.

14:21And there may be good reasons for doing it as well,

14:25but that's the kind of thing that needs to be worked through

14:29in a proper, you know, regulated setting.

14:37After our interview,

14:39Max Little tells me he's received several lucrative offers

14:42to join tech giants who smell new business opportunities.

14:48He turned them down.

14:53Artificial intelligence will undoubtedly improve

14:56doctors' abilities to detect and diagnose disease.

15:00But amid all the opportunities AI offers,

15:03there's an urgent need for regulation.

15:12We're on our way to China,

15:14a country that has experienced breathtaking change

15:17in recent years.

15:23Its capital, Beijing, is buzzing.

15:26The whole country is hungry for progress

15:29and is on a fast track to the future.

15:34Time seems to move faster here.

15:37By the year 2030,

15:38China aims to be the global leader

15:40in the field of artificial intelligence.

15:44And there's a lot to indicate it will meet that goal,

15:47because the government has bankrolled subsidy programs

15:50worth billions of euros.

16:03These robots aren't assembling cars.

16:05They're the big attraction in Beijing's latest smart restaurant.

16:15AI in the kitchen,

16:17and automated waiters.

16:25I have a meeting here with the design researcher Giesche Juulst.

16:30A former Internet ambassador for the German government,

16:33she's currently spending a research semester

16:35at Tongji University in Shanghai.

16:38I ask her about her impressions of China.

16:45There's this real hunger in the city,

16:48and it's super fun to talk to young people

16:50because they want to be the motors of change.

16:53They work day and night.

16:55They have a new work-life balance model,

16:57it's called 996.

16:59I thought, what do you mean 996?

17:02And they said, we work from 9 a.m. to 9 p.m. six days a week.

17:06That's the better model now,

17:08because they used to just work non-stop.

17:12But no one's stopping, no one's hitting the brakes.

17:14They work like crazy because they want to bring about change.

17:18This restaurant costs $20 million, just the one restaurant.

17:23They've invested this huge sum to digitize the entire operation.

17:28They aren't just robots serving the food,

17:30the whole kitchen is digitized.

17:32Refrigeration is monitored, supply chains are monitored,

17:35there are dashboards for everything.

17:37Everything is connected here.

17:39They're testing what works and which aspects can be implemented

17:42in other restaurants of this chain.

17:45That's the idea here, to just try things and to think big.

18:03So I'll just help myself, if I may.

18:14But what about privacy?

18:18There's seen to be a trade-off between security and privacy.

18:22You often hear how AI has increased public safety,

18:25for instance that the ubiquitous surveillance cameras

18:27have dramatically increased the crime-solving rate.

18:31It's hard for us to relate to,

18:33because privacy and personal rights are so important for Germans.

18:38Here there's a different tradition and take on the issue.

18:53I'm fascinated by China, but it also puzzles me.

18:59How can they be reconciled,

19:01the high civilization of ancient China and the modern industrial state,

19:06with surveillance cameras everywhere?

19:20The Longgan District in Shenzhen.

19:24In the heart of China's booming economic region, north of Hong Kong,

19:28we visit the Smart City Control Center.

19:31A giant monitor displays the data of the entire district in real time.

19:37Numbers of new residents by neighborhood to plan schools,

19:41water supply levels, power outages.

19:44All this information is collected, compiled and evaluated

19:48using artificial intelligence.

19:52The showcase project was developed with Chinese tech giant Huawei.

19:56Chief Engineer Chen Bangtai tells me

19:59the city now operates more efficiently.

20:05So what you're doing here is urban planning?

20:09Yes. The systems are a big help.

20:14These are hospital beds.

20:17Right now there are 15,000 doctors and nurses,

20:23and 7,600 beds.

20:25So is Shenzhen currently healthy or sick?

20:31A smart surveillance system scans the entire city.

20:34Illegal structures, like this one on a roof,

20:37are quickly identified and demolished.

20:41To me, some of this feels like the backdrop to a science fiction movie.

20:46Employees with live streaming body cams inspect side streets.

20:51This is total surveillance.

20:56Chen shows me how cameras installed in restaurant kitchens

21:00even keep tabs on cleanliness.

21:03But doesn't the chef mind being monitored all the time?

21:09The system logs all the people who view the images.

21:13Anyone who looks at them without permission is punished.

21:16Total transparency for the purpose of progress.

21:19Chen says residents of Longgang District approve.

21:24Jaywalking is not allowed,

21:26and offenders are immediately identified.

21:31Look here. You jaywalk once,

21:33and right away your social credit score is up.

21:37You can't do it again.

21:40You can't do it again.

21:43You can't do it again.

21:46Your social credit score drops.

21:52This degree of surveillance is unthinkable in the West.

21:56But here in China, they take a different view

21:58and say it's driven a drop in crime.

22:02What does it say here?

22:04Male, youth, without glasses.

22:07Youth?

22:10Yes, suddenly you're a youth.

22:13I love Chinese facial recognition.

22:16Youth.

22:24A transparent society in the interest of efficiency.

22:28Some of this appears useful.

22:30But do we really want to measure, control, and analyze everything

22:34just because it's technically possible?

22:42Won't that inevitably lead us down a road to data dictatorship?

22:58Maybe trust is better than smart control.

23:02Silicon Valley.

23:04A synonym for innovation and unlimited freedom.

23:08The biggest players in the field of AI are based here.

23:12But their headquarters are hidden behind inconspicuous, low-rise buildings.

23:19Facebook.

23:21We use their services, entrust them with our data,

23:24but the company is impervious to the public.

23:27A selfie at the entrance gate is just about tolerated.

23:33Next door at Apple, the Visitor Center's 3D model of the campus

23:37is as close as non-employees can get to the new building.

23:42What's going on inside?

23:46It's all confidential.

23:48We want to visit Google here in California,

23:50and requested an interview weeks before our arrival.

23:54But all we get are stalling tactics.

23:57Like these visitors, Google leaves us out in the cold.

24:02Apart from a small store,

24:04this is the only visitor highlight accessible to the public.

24:08It's the only place where you can find Google.

24:12It's the only place where you can find Google.

24:15It's the only visitor highlight accessible to the public.

24:20These Android lawn statues are even a designated location on Google Maps.

24:28Welcome to Google.

24:35Facebook have refused to answer questions.

24:39The European Union handing Google a $2.7 billion antitrust fine.

24:46These companies command growing power over our daily lives

24:50and growing political influence.

24:53Google spends more than 6 million euros a year

24:56lobbying Brussels alone.

24:58The EU's Transparency Register

25:00lists more than 200 meetings with Google representatives since 2014.

25:05Google is the busiest lobbyist in Brussels.

25:09We finally get our interview,

25:11not in California, but in Munich, Germany,

25:15with one of the longest-serving employees, Jens Redme.

25:20How important is AI for Google?

25:24AI is so important to us

25:26that two years ago we rebranded our entire research division to Google AI.

25:32AI drives a significant part of our product development.

25:35AI, above all, drives a significant part of our efforts

25:38to improve the quality of our products.

25:41Take machine translation.

25:43Through the use of machine algorithms

25:45we've seen faster progress over the last two years

25:48than we did over the entire previous decade.

25:51Society will undoubtedly be propelled forward

25:53by the implementation of these services

25:55and the use of AI in the years to come.

25:58What's key is that it's done responsibly,

26:00under the principles of AI.

26:02What's key is that it's done responsibly,

26:04under the principles of transparency.

26:07We need to explain how things work,

26:09why they are needed, where people's data goes,

26:11how they can control it, how they can delete it,

26:14if they want to delete it or forward it.

26:16The user must have control.

26:21But what about technologies like Google Home,

26:23the smart microphone sitting in people's living rooms?

26:27Google Home isn't eavesdropping.

26:29There's a small chip on the device

26:31that listens out for the so-called hot word.

26:33It's waiting for the command,

26:35OK Google or Hey Google,

26:37and only then is the microphone switched on

26:39to send a voice command or search request

26:41to the Internet, the Google server.

26:44It then presents the result.

26:50So, as a science reporter,

26:52I'm naturally curious about the future.

26:54There's this patent application from September 2016.

26:58Google's application gives a detailed account

27:01of what can be deduced from household noises.

27:04How long we brush our teeth,

27:06whether we argue, or whether a housemate is ill.

27:13It's much more about capturing atmosphere and habits than words,

27:16and it's a Google patent application that anyone can look up.

27:20I don't know anything about this particular patent application.

27:23We have a whole series of patent applications every year.

27:26Most of them are imaginary, fictitious services,

27:29which, like in many other companies,

27:31are never translated into real services.

27:34So I can't say anything about this particular patent at this point.

27:42Patents for imaginary, fictitious services?

27:45Google's EU lobbying activities, at least,

27:48are unarguably real.

27:51How much does Google intervene in the EU?

27:55I think the more important question is the one that lies between the lines.

27:59Namely, how ethically does a company deal with product development?

28:03And we have set our own rules according to principles

28:06that guide our own actions.

28:08Research and product development

28:10that also guides our business decisions.

28:14On its home turf in the United States,

28:17Google is facing mounting political pressure.

28:20In Washington, we meet Barry Lynn,

28:22head of the think tank Open Market Institute.

28:25He warns of the dangers posed by tech giants' influence.

28:29We need to know in our society

28:31that the people who bring information to the public sphere,

28:35who talk to the press,

28:37who talk to the media,

28:39who talk to the public sphere,

28:41who talk to the press,

28:43who talk to the representatives in Congress,

28:46that they represent their own selves,

28:49that they're speaking in their own name,

28:51and not for someone else.

28:53That they're not stooges, that they're not puppets.

28:56And the fact is, today, our society,

28:59this is true here in Washington, it's true in Europe,

29:02our society is filled with puppets, with stooges,

29:05who represent the interests of Google and Facebook and Amazon.

29:10Given big tech's monopoly power,

29:12calls for regulation are growing louder in Washington.

29:17When you have a monopoly,

29:19whether it's over retail, whether it's over search,

29:22then it means that the public doesn't actually have the ability

29:26to understand how that information is being used,

29:28how that power is being used.

29:30Monopoly per se, unless it is regulated closely by the public,

29:34is a danger.

29:35Google means to take over the world.

29:40They mean to direct our thoughts between person and person,

29:46our communications between person and person,

29:49our dealings in business between person and corporation.

29:55They mean to direct everything that they can.

29:58They want to know what's going on in our thermostats,

30:03in our houses, they want to know what we're watching on TV.

30:06They are at a level of hubris

30:10that even the Stalinists could never have imagined pushing for.

30:19Google, Facebook, Amazon.

30:22Will the influence of tech giants continue to grow?

30:25What can be done to rein in their monopolies?

30:30One thing is clear.

30:32Artificial intelligence is consolidating their grip on power.

30:36There's an urgent need to rethink antitrust policy.

30:43Mobility is another area in which AI is powering the march of innovation.

30:48In the near future, it could put self-driving cars on city roads.

30:55But how realistic is this vision?

30:59We've come to Boston,

31:01to the prestigious Massachusetts Institute of Technology.

31:07Sertac Karaman is a leading expert in the field of self-driving cars.

31:12He and his team are working on prototype autonomous vehicles.

31:17I think that we've nailed a couple items with computers and machines.

31:22One is all of this mapping and localization,

31:25all the technology works super well.

31:28Computers can know where they are within centimeter,

31:31maybe sometimes even millimeter precision,

31:33way more than what is required to drive.

31:35Computers are now able to look around

31:37and understand where everybody else is.

31:39But that's not the only thing.

31:42Computers are now able to look around

31:44and understand where everybody else is.

31:46But that's not what's required for driving.

31:49What's really required is to understand what's going to happen next,

31:53in the next three seconds, five seconds, next minute,

31:55maybe even sometimes next hour.

31:57And so that's the key missing piece.

32:00And I think the problem there is that right now,

32:03it's very hard for you to describe to me

32:07how you understand whether or not a person is going to use the sidewalk

32:11or is going to use the crosswalk and cross the street.

32:14Sometimes you look at the face of a person

32:16and that facial impression gives them away,

32:19and you will slow down.

32:20Sometimes not.

32:22They may be looking at the same direction.

32:24They may be standing in the same location.

32:26It's just a little face impression,

32:28maybe just the way they stand.

32:30And unfortunately, that kind of intuition, gut feeling, and so on,

32:34is very hard for us to program into the computers.

32:39It works in a simple lab environment,

32:41but in real-life settings,

32:43the algorithms are still totally out of their depth.

32:46Not that that bothers advertisers.

32:59Our test drives were nothing but a series of glitches.

33:05So...

33:09An inexplicable emergency stop,

33:12and another one on the second attempt.

33:20The sensors on this vehicle were overtaxed

33:23by a car parked on the curb.

33:27And here, this smart car overlooked a car veering into our lane.

33:36Well, that didn't work.

33:39Talking to the MIT engineers,

33:41it becomes clear that to build a self-driving car,

33:44developers need to meet a massive scale of technical requirements.

33:50What I think about fully autonomous cars is that I think

33:53I would be very surprised if it happens in less than 10 years.

33:57Also, I would be very surprised, I'm a big believer,

34:00I'd be very surprised if it doesn't happen in 20, 30 years.

34:03I think it will happen at some point.

34:05But I do think that people really underestimate

34:08the kind of technology required to be built

34:11to make your car fully autonomous under every condition,

34:14every circumstance, every whatsoever.

34:16That's the very hard part.

34:24Driving is not as trivial a process as you might think.

34:28And that's because you constantly have to watch what's going on around you.

34:34Cyclists, pedestrians, sometimes you have to second-guess.

34:38Does this guy want to cross the road or not?

34:41It's hard to imagine all that being calculated automatically.

34:54A self-driving vehicle would have to be able to deal with all of this, too.

34:58Here we have a truck doing a three-point turn.

35:02I may have to back up now if he doesn't make it.

35:06Does she want to cross the road or not?

35:08Some people don't even wait.

35:29The first sign of an autonomous vehicle is that it's waiting.

35:33The fully autonomous car is a distant dream, but driver assistance systems are already

35:43making our roads safer.

35:55An accident filmed from a car equipped with emergency brake assist.

36:00The sequence of events can be assessed in slow motion.

36:04The red vehicle ahead overlooks the upcoming traffic jam.

36:08No brake lights appear.

36:10But the distance sensor in this car registers the jam, brakes, and prevents a further collision.

36:17But which principles should guide decisions made by technology in an accident situation?

36:23In recent years, MIT's Media Lab has been addressing the ethical questions raised by

36:28artificial intelligence.

36:33What moral compass should future smart devices refer to?

36:44Iyad Ravan is one of the world's leading experts on such issues.

36:49He and his team developed a survey called the Moral Machine to explore ethics for programming

36:53autonomous vehicles, like in the event of an accident.

36:59Most of the time, people don't remember anything, and people have no time to react.

37:05Everything happens very quickly, so they just are surprised.

37:10Maybe they see something in front of them, and they just swerve in some random direction.

37:15Or maybe they just freak out and press the brakes.

37:20You cannot expect a human being to do the right thing in such a small time scale.

37:27Unless they made a decision beforehand, like did they drink and drive, or did they know

37:33that they were going to cross a red light.

37:36Then you blame them.

37:37But otherwise, you can't really blame the human.

37:40But with a machine, because of the speed of the electronics, because the autonomous car

37:44is evaluating the environment millions of times per second, then time goes much slower

37:52for the machine.

37:53And it's able to recalculate the situation and maybe recalculate the strategy.

37:58And this is where we can make a potentially better judgment than whatever random choice

38:05the human used to make in this situation.

38:08Now what is better, though, is a very interesting question, and it's not obvious.

38:13Let's see a case where we have people versus people.

38:15So now we have the vehicle has two people inside of it, and it's going to either swerve

38:23and hit the barrier, so the people will die in the car, or the car will go straight and

38:29kill the pedestrians, who are crossing illegally, but they're also women, and the people in

38:37the car are males.

38:38So now it gets very complicated very quickly.

38:41Should you prioritize women over men, or should everybody be the same?

38:45Should you prioritize pedestrian over passenger or not?

38:49Should you take into account the fact that people are crossing illegally in this case?

38:55So as you can see, once you have multiple dimensions, it becomes not obvious what the

39:01right thing to do is.

39:03A or B, who should die?

39:07The elderly lady crossing on red, or the child in the self-driving car?

39:13What choice should the algorithm make?

39:16EADD's online survey presents respondents with various scenarios, each with its own

39:20unique dilemma.

39:22Respondents are then asked to choose how they would want an algorithm to decide.

39:26So as a result, we have 40 million decisions, and they're still counting, from people all

39:32over the world, and it enables us to start analyzing what do people agree on, but also

39:38how do they differ.

39:47So does our culture influence our moral judgments?

39:52People always agree on saving more lives, saving children, saving people who cross legally

39:59over people who don't cross legally, and so on.

40:02The most interesting part is you could pick a country, like Germany, and you could see

40:08how they compare to the global average.

40:10You could see...

40:11Ah, okay.

40:12So the status is not really important, but what you can see in Germany is preferring

40:16inaction.

40:17Yeah, so if you don't have to, if you prefer to just go straight, which is the default.

40:23Don't take a decision.

40:24Exactly.

40:25So this means Germans don't like to take a decision?

40:27Yes.

40:28Germans don't like, also Augen zu und durch, you say, close your eyes and go.

40:38And just go.

40:39So this means, in other words, you can see a bit the acceptance of technology taking

40:46a decision, and the more you say inaction means, well...

40:51Means the car goes straight, yeah.

40:52Okay.

40:53Fate.

40:54A comparison of Germany and France reveals cultural differences.

40:59The French tend to favor sparing women, and there's a stronger focus on children.

41:09And contrary to Germans, the French don't want to leave things to fate.

41:13They want the machine to make the decision.

41:18The machine is kind of a mirror.

41:20For the first time, something that you did subconsciously or maybe instinctively in the

41:27case of an accident, you know, you just act randomly, now you have to make a conscious

41:32choice and the machine is forcing you to make a choice, right?

41:37You cannot hand wave it, because in the end you have to program something.

41:42Driverless cars aren't yet ready for the road, and ethical questions still abound.

41:48Artificial intelligence harbors immense potential to benefit daily life, medicine, or mobility.

41:54But we also need to look beyond the technical possibilities.

41:58What aim does such progress serve?

42:00It's a question artificial intelligence algorithms can't answer.

42:04Only humankind can do that.

Category

Transcript

Recommended