RTS Dataland

 

00:00:10:07 Voiceover

Right now, at this very moment, Artificial Intelligence is trying to figure out exactly who you are. AI mimics our thoughts and emotions. It analyzes our cells to decode our genetic material. No task is beyond its capabilities. AI is even learning to drive our cars for us. AI controls robots that work more efficiently than we do. So where does this leave us in this new world of artificial minds? In the last ten years, we have developed an inseparable bond with our cell phones. Smart phone capabilities are bolstered by artificial intelligence, which serves our needs, whilst keeping a watchful eye. Network operators record our movements to the nearest meter, as some applications Geo-tag users every 60 seconds. Surveillance cameras scrutinize our every move. Unknown entities siphon our data from public WiFi networks. Sensors in the roads we drive on record our journeys. AI regulates traffic, predicts the weather, and records pollution levels.

00:01:57:01 Voiceover continued…

Our frenzied online activity generates an avalanche of data. Each minute, we make 3.3 million posts on Facebook, 3.8 million searches through Google, and send 29 million messages WhatsApp. By 2020, each human being will generate 1.7 MB of information each second. These torrents of data are stored in banks of servers all over the world. In the era of Big Data, your digital footprint is worth more by the day.

00:02:50:06 Voiceover continued…

In Paris last March, 15,000 participants spent thousands of euros to access a trade show on Big Data. Events like these are big business, and you don't want to be the one missing out. Modern AI systems fed by colossal data sets are opening doors to lucrative opportunities. Machine Learning revolutionized AI development, that is, machines capable of teaching themselves without human input. Over the last 5 years, this technology has transformed AI development. These algorithms are able to instantly identify objects placed in front of a camera. But this ability to recognise has many potential applications. A Paris-based start-up has developed new facial coding technology.

 

00:03:48:20 Music begins, Voiceover continues…

This algorithm is capable of reading facial expressions, which are so nuanced, they may pass unnoticed to the human eye. The endpoint is commercial - reactions are measured against different adverts to see which images are the most effective.

 

00:04:10:11 – Anne-Marie Gaultier – Datakalab President

Our software allows us to get a direct read on someone's emotions. 90% of things that happen to us over the course of a day are processed subconsciously. So our AI algorithm allows us to gain insights into what the customer is feeling at any given second.

 

 

00:04:34:09 – Xavier Fischer – Product Development Manager Datakalab

We track around 90 points on a person's face and we analyze the movement of about 42 muscles. From there we quantify emotions; joy, surprise, fear, anger, sadness, disgust and contempt. So we see a dip in emotion between the 5th and 6th second, conversely there's a spike at the 12th second to 100%, which is very strong we can then see what works and what doesn't, because the point of all of this is to create better content that is more engaging.

00:05:08:01 Voiceover

This algorithm is not limited to commercial applications. It can also measure people’s reactions to politicians. It was tested during the last French presidential election cycle during a televised debate.

00:05:37:02 Anne-Marie Gaultier – Datakalab President

Thanks to facial coding we were able to measure which candidate generated the greatest emotion Al response - Mélenchon. Then the candidate who generated the most negative emotions - Marine Le Pen. And then the candidate who's 'presidential manner' was most engaging was Macron, so from March we were able to identify these trends.

00:06:03:01 – Voiceover

To make such an algorithm effective, it must be fed millions of data points. The Gods of the mega data world are the Big Four – Google, Amazon, Facebook and Apple. We headed to Zurich to see it in action. Here, Google has built its largest research center dedicated to AI outside of the US, with 2000 contributors draw from all over the world.

00:06:35:00 Anna Ukhanova – Head of Technical Programmes, Google A.T.

Google has always been excited about artificial intelligence and the use of machine learning has increased rapidly in recent years, and it makes possible what was impossible just a few years ago. Machines are learning in a different way from humans. They need to see a lot of examples to understand what they seeing on a photograph. For example, if you are teaching a computer to recognise a cat in the image, you need to show it a lot of examples. And then the system would try to take a guess for the new pictures and every time it makes a mistake, for example, mislabel a dog for a cat, it would adjust the parameters a tiny bit. And the after seeing probably a million examples, the system would learn to differentiate between cats and dogs as well as people do.

00:07:45:15 Behshad Behzadi – Chief Technical Officer, Google Research Centre

When is my next meeting?

- The next thing on your calendar is today at 1pm. It's called context.

00:07:54:13 Voiceover

Behshad Behzadi leads a team developing a next-gen voice-command assistant, a project very much in keeping with the ambitions of big tech companies. Thanks to progress made with image and voice recognition, the assistant can answer any questions a user could ask.

00:08:13:15 Behshad Behzadi - Chief Technical Officer, Google Research Centre

-Who painted this?

- Mona Lisa was created by Leonardo da Vinci.

-Translate this.

-This might be right:

check out the translation in the image above.

-How many calories does it have?

- There are 95 calories in one medium apple.

00:08:42:02 Voiceover

This virtual assistant slips into the heart of your home as a smart speaker.

00:08:54:01 Behshad Behzadi - Chief Technical Officer, Google Research Centre

The assistant gets better the more you use it. It both learns from the interactions to be more personalized towards my needs but also I can teach things to the assistant. For example, I can just say, "OK Google, my favorite team is Barcelona."

- OK, I'll remember that you said: "My favorite team is Barcelona."

And now the assistant remembers this and then from now on, I can simply say: "My team, when is my team's next game," or things like that. People do not need to learn how to interact with machines, they can just say it. You can just say: "Hey, Google, order a taxi for me." It makes the services and benefits available to all people. OK, Google, how's the traffic to my home?

- On your way home, traffic is light as usual. It is 28 minutes by car.

OK, Google. Remind me to call mom when I get home.

- Sure, I'll remind you on your phone when you get home.

When that works, are they really at the human level, the same way that humans interact between themselves, then these types of technologies are available to everybody.

00:10:04:19 Anna Ukhanova – Head of Technical Programmes, Google A.I.

I have an optimistic view of the future, and I believe that there are tremendous opportunities in artificial intelligence. We are still in the early days of the field but we have seen that it can tackle some of the largest problems in the world. The technology has an ability to find patterns in complex data sets that humans could never do alone. So, this has huge implications and the technology has an opportunity to help researchers in different fields to tackle some of the most difficult problems in the fields of medicine, transportation, astronomy and elsewhere.

 

00:10:45:16 Voiceover

 From healthcare, to transport, to the stars, the applications for this technology seem limitless. To find more concrete applications, we head to a so-called 'Smart Town.' On the way, we noticed airports' increasing reliance on AI systems to manage the flow of passengers. Planes themselves will soon integrate AI systems to monitor weather patterns, altitude, and fuel levels in real time to adjust the flight path. Managing the daily rhythm life of life is a growing problem for urban centers. Dublin, Ireland, is proud to declare itself a Smart city, having bet heavily on an investment in AI systems for logistics. The old docklands are now the Smart docklands, a model for future urban development.

00:11:46:17 Jamie Cudden – Head of Smart City, Dublin City Council

We are the Silicon Valley of Europe, in Dublin. You just look at the Docklands here, some people call it the Silicon Docks. We've got world leading tech companies. Within 15-minute walk you've got Google with 6,000 employees, We've got Facebook, we've got Airbnb, all creating excellent value for the city and building great partnerships and relationships on our Smart city program as well.

00:12:13:02 Jamie Cudden – Head of Smart City, Dublin City Council (Cont.)

There's lot's of big trends, like big data, Internet of things, machine learning, artificial intelligence, but what you really need to do is take that data back, create insights from that data, and in almost real-time, make decisions based on that data. We're seeing it across mobility and transport, across waste, across emergency responses, and the true Smart city is the city that actually can act on the data, make better decisions, and create better outcomes, at the right time, for its citizens.

00:12:44:21 Voiceover

Some applications are quaint, like this smart bin that notifies the local services that it needs to be emptied. Other uses are aimed at more complex problems, like managing the flow of traffic in a town experiencing an economic surge. The council has installed cameras and sensors across a number of key locations. They record the flow of pedestrians, bikes, busses and cars. But how do they manage this huge flow of data? Brendan O’Brien runs the team responsible for traffic light operation. Many towns use a similar system, but here, adjustments are made in real time, based on the flow of traffic. Every bus is equipped with sensors.

00:13:38:11 Brendan O’Brien

Every 20 seconds, every single bus reports to us where it is, what it's doing, what route is it on, is it in congestion, is it stopped at a bus stop. We'll see the slow down and the bus movements and that will actually alert our operators here to the fact that those problems which the buses are experiencing, but that in turn means that all their traffic is experiencing it as well.

00:14:02:01 Brendan O’Brien (cont.) and employee

Hi. So, what's happening this morning?  It looks a bit a congested there in a few locations.

-          The data shown was actually congestion here this morning due to delays on this route here, so they should be coming into view as we come down through here now.

That's a fairly longish queue at the moment, isn't it?

-          It's actually building up quite severely, yes.

Our system is a self-learning self-calibrating system. Most of the time, the system works away in the background by itself and in an automated way.

00:14:27:21 Voiceover

The system still requires some human input to manage effectively, but there is another great technological leap forward on the horizon, a piece of software that will predict problems before they arise.

00:14:38:19 Brendan O’Brien

Can we predict what's going to happen next? We see all these vehicles coming in, you know, can we go forward the next half an hour and say, "Oh, I think here will be congested, here will be congestion unless we take some action.” That's kind of using all that information and using multiple data sources to really find a pattern. And once they find a pattern, they can then see what to do next.

00:15:04:05 Voiceover

The city authorities are sitting on an enormous pile of data, but they don't have the means to fully utilize it. They have therefore joined in partnership with IBM, which opened a research station just outside the city in 2011.

00:15:20:07 Wendy Belluomini, Director IBM Research Ireland

Once that investment is made, then you can do things like sensorizing entire road networks and actually collecting that data. And then once you have it, then you can actually start applying artificial intelligence to that data to make predictions, to make suggestions as well.

00:15:39:18 Voiceover

Data scientists are the backbone of AI development. They're constantly refining and analysing a huge range of inputs to find hidden correlations amongst data sets.

00:15:54:09  Jakub Marecek IBM Research Ireland

We ingest a lot of data from various sources. The main challenge is that the data are very heterogeneous. Each tells you a small part of the complete picture. Each tells you a small part of the complete picture. It's very hard for us to actually know what has happened. So we may be able to detect that this road is blocked. But the precise reason why the road is blocked is often obscure.

00:16:25:00 Voiceover

Often the solution lies in adding a new source of data. IBM has begun integrating posts made on Twitter to its algorithm's data feed.

00:16:36:14  Jakub Marecek IBM Research Ireland

The Twitter often allows you to associate Geo tech tweets about an event such as, there was a concert and 70,000 people are leaving this arena, with the information that this whole area is seeing much slower travel times than usual. So sort of in order to provide the explanations, the Twitter is very useful.

00:17:03:15 Wendy Belluomini, Director IBM Research Ireland

So a lot of artificial intelligence now, I think, comes down to making predictions, saying this is going to happen. And then leaving it to the person to decide what to do about that. As we go forward, we'll see a lot more solutions, as opposed to just predictions.

00:17:26:14 Voiceover

Many developers are hard at work creating programmes that leverage predictive technology. This team is working on an AI assistant for drivers.

00:17:38:14 Martin Mevissen IBM Research Ireland

You have now the event detection in the video stream, right?

-Yes.

-So that's something we could use for the companion, right.

00:17:48:22 Martin Mevissen IBM Research Ireland

What we want to provide here is an assistant that gives information that's relevant to you in the context of this strip. So the idea is, essentially, the companion that watches over you, like a guardian angel that sees all your behaviours because there's only some advice that's relevant for you.

00:18:16:10 AI in car

Oh, it just detected something.

It seems you're going to use UCD Bellfield, is that correct?

That is correct, yes.

You should enter UCD through the North East gate. It's still Oregon road. Is that okay?

Can you explain to me why?

Because the UCD internal Gates are closed now. Can I suggest the best parking spot to you?

That will be great, yes.

00:18:44:22

The assistant can intuit what the user needs, without having to ask a direct question.

00:18:51:05 Martin Mevissen IBM Research Ireland

This data coming from the car and from the driver, like where the car has been, what trips I've taken in the past, where am I headed now, but also a lot of data from the environment, right? Like traffic situation, weather situation, maybe even data from social media or the publicly available data.

 00:19:10:12 Cameraman, driver, Martin Mevissen

What's the camera in front of you?

So with this camera will be able to detect certain conditions of the driver and included these conditions in our AI model and expand our risk mitigation strategies so as to really include the cognitive state of the driver.

So if you can detect that somebody has a lot on his or her mind, then a companion may use that information to be more proactive than a situation where somebody is really focused and not distracted.

00:19:52:04 Voiceover

AI wants to be your best friend, a friend that reads your emotions, counsels you, guides you and follows you. It seeks omnipotence, but whether this will be for better or for worse remains to be seen.

00:20:30:10 V.O. Cont.

We head North in search of the cutting edge of digitally enhanced medicine. Finland was an early adopter of data driven healthcare. AI is a key asset in providing treatment. Helsinki's general hospital, the largest in Finland, is developing an algorithm to efficiently process patients. Markus Leskinen was an early advocate of such research. His unit deals with premature births that weigh less than one and a half kilos, whose lives lie in the balance.

00:21:31:22 Markus Leskinen – Research Director, Helsinki University Hospital

We have collected data for several years, actually over the decade. We monitor data, heart rate, breathing rate, oxygen saturation and blood pressure, laboratory values. We are trying to develop algorithms that can detect the different diseases and as a model we have so far used sepsis.

00:21:55:11 V.O.

The immune system of premature babies is very weak. They are particularly at risk of contracting septicemia. This infection can have serious repercussions on mental development and can lead to cognitive impairment. In 10% of cases, it can even prove fatal.

00:22:13:19 Markus Leskinen, Helsinki University Hospital

In the beginning, the computer needs to know whether this patient had sepsis or not. So basically we are telling them that, okay, these patients had sepsis, these patients didn't have sepsis, and asking computer to find what differentiates these two groups. And then the computer is trying to find the rules. So the machine is in a way learning by itself to detect sepsis. What we are hoping to do with machine learning is to get the tool that would give the same kind of eyes that an experienced doctor has, that tell also less experienced doctors that okay, now you should perhaps be worried about this baby. And actually in this project we were able to detect sepsis 24 hours before, doctor had such a high suspicion that we took blood complex samples that confirm the diagnosis.

00:23:06:23 V.O.

This advanced detection system allows treatment to begin 24 hours earlier in a situation where time is of the essence. Data accumulated throughout a patient's lifetime is what forms big data. In Finland, for the last 20 years, this information has been stored in a centralised server. Helsinki's general hospital wants to deploy AI to get the most out of this data, and has launched some 30 pilot schemes.

00:23:41:09 Prof. Anne Pitkaranta – Research Director Helsinki University Hospital

We collect all the information from the patients from the electronical health records, from their imaging pictures, from their laboratory tests. The amount of data nowadays is so huge that there no man can handle that. So we need some help for that. Artificial intelligence would be the normal life, not only for researchers  but also for the doctors and it would be a tool for a hospital directors and nursing staff. So now it is time to use it efficiently.

00:24:24:14 V.O.

Another strand of research is focused on brain trauma patients. Often in a deep coma, it is difficult to assess their medical state, which can change drastically in a short space of time. Doctors are therefore reliant on their instincts. So here, an algorithm is being tested to see if it can produce more objective responses.

00:24:55:07 Doctor Rahul Raj – Department of Neurosurgery, Helsinki University Hospital

So here we have that 60 year old severe head trauma patient non-operated, external ventricular drain to control high intercranial pressure. IVD is threshold to 15 water centimetres and we will continue like this one more day.

00:25:13:19 Interview Doctor Department of Neurosurgery, Helsinki University Hospital

We are collecting all data that's being monitored from patients treated for severe head trauma in the intensive care unit. It includes everything from all laboratory results, respirator, invasive blood pressure to intracranial pressure, cerebral perfusion pressure and other so brain oxygen levels. So that gives us hundreds of thousands of data points from one patient.

So you can see that he's obeying commands, the motor responses are right. It's all good.

It's impossible for the individual physician to take into account everything and make an objective reasonable estimation of the patient's prognosis. So that's where the artificial intelligence model comes in. It's capable of taking all the data, putting it together, and give us one number, a percentage basically that every doctor understands.

00:26:05:06 V.O.

To make sense of this enormous amount of data, doctors turn to data scientists. Together, they construct a predictive algorithm, which evaluates a patient's probability of survival in real time.

00:26:21:09 Doctor Department of Neurosurgery, Helsinki University Hospital

These are all the parameters and variables that the algorithm includes. And then it summarises everything, so we can see real time predictions of the patient and how he's doing. At the moment, the model predicts probability of 30 day survival. It's the initial step. And if we see that the probability of survival is going down, patient is doing worse, then we know that something is now happening to a patient. For example, has the brain bleeding expanded? Is it bigger now? Should we operate it? It's functioning like an alarm bell, telling us that now something is wrong. We have to look at the patient more carefully and do something about it. The algorithm is never alone taking the decision. It's not the purpose of that algorithm. The purpose of the algorithm is to give us an objective measure of how the patient is doing so that the doctors in charge can make the real treatment decisions. At the moment you can never blindly trust the AI, but you have to know the information behind it.

00:27:34:23 V.O.

Another source of big datasets is our own genome. Six billion different data points can potentially predict a person's predisposition to different diseases. Finland has just launched a programme of unparalleled ambition: The FinnGen project. The goal is to analyse the genome of 10% of the population, that is 500,000 Finns. Within these storage units filled with liquid nitrogen, we can perfectly preserve blood and tissue samples for decades. The DNA extracted from the samples will allow us to enter a new age of personalised, preventative medical treatment.

00:28:19:10 Prof. Aarno Palotie, Research Director, Finngen Project           

This is among some of the largest projects in the world. I think what's key in genomic medicine is that it teaches us about the disease mechanism. It provides a stepping stone for, for instance, biomarkers, better diagnosis, potentially new treatments. Without better understanding, there is no new treatment, there is no better health. So we aim in the FinnGen project to collect 500,000 Finnish people, which is 10% of the Finnish population.

00:28:58:06 V.O.

Just under a dozen biobanks across the entire country work tirelessly to achieve this monumental goal. In Helsinki, a specialist nurse approaches each patient to ask them to donate a DNA sample.

00:29:20:14 Kimmo Pitkanen – Helsinki Biobank Director

The hospital has set the target that - a slogan - to collect every patient entering the hospital into the biobank. We are not quite there yet, but we are aiming at hundreds of thousands of samples and patients in the biobank.

00:29:40:06 V.O.

The pitch is made. Some apprehension is expected. The patient is required to know who will have access to their genome and if their privacy will be protected. Once convinced they sign a consent form approved by an ethics panel.

00:29:53:06 Patient Masi Tammela

From my father's side we have quite a lot of heart problems and heart diseases, so it could be maybe helpful to see like what's the real reason behind there and if I could already know for myself there is an increased risk, maybe I would take better care of my body with sports and with food. But on the other hand, I think the bigger benefit might be from my my data, hopefully for others and maybe for my family or my, my children in the future or something like that.

00:30:32:15 V.O.

Genetic data is ultra sensitive. It contains insights into a person's medical profile, information relevant throughout an entire lifetime, not only for the donor but for their immediate family, also. Here the samples are archived. Only the bio bank can identify the donor. Once the genome has been sequenced, it will be made anonymously available to researchers. 95% of donors agree to share their information.

00:31:03:14 Kimmo Pitkanen – Helsinki Biobank Director

The high level of content comes from the high level of trust. I mean we pay high taxes, yes, ut we get quite a lot from the society. Healthcare is models free or test the minimal cost are covered from the patients. There's a long tradition of trust in Finnish and probably, I don't know the societies towards government, universities, researchers.

00:31:32:17 V.O.

Genetic information falls the same rules as big data. The more you have, the more successful the application. Each genome is unique with its own independent variables. That is what will make healthcare of the future infinitely more precise diagnostics as well as prescriptions will be tailored to the individual making standardised treatment obsolete.

00:32:01:09 Prof. Aarno Palotie, Research Director, Finngen Project         

Thinking of personalised medicine with the idea that we wouldn't treat mean values anymore. Very few of us is a real mean value. We are persons and we have our personalities and each of us likes to be treated as individuals. To be able to analyse an individual in a clinical setting, as an individual, we need to compare it to something, and we need large numbers to compare. It's not enough that we have 10 people as your reference. We need hundreds of thousands. To find new things, we need large sample numbers.

00:32:44:19

By analysing billions of data points, AI gets to the heart of our biological processes. It can observe how our bodies react to pathogens or a new treatment. It's the great hope in the face of the deadliest diseases still afflicting mankind in this century. Cancer will kill almost 10 million people across the world this year. The challenge ahead lies in understanding how to improve treatment whilst mitigating side effects. Olli Carpen is a doctor and researcher at the university of Helsinki. He too is betting on AI technology.

00:33:33:08 Olli Carpen, Department of Pathology, Helsinki University Hospital

I'm a pathologist and what people think about pathologists, they can do everything and they know everything, but they are always too late.  My interest is in ovarian cancer and what we want to understand is how cancer affects the metabolome. What are the outcomes? How do patients respond to treatment? What is very important is that we understand the disease as a whole, not just what the cancer cells do, but how they affect the entire body.

00:34:15:19 V.O.

In order to understand the reactions of the body, one must study the metabolism. This is the entirety of the chemical reactions that take place inside us every second. This process releases the molecules that give precise information on our biological processes. We call these biomarkers.

00:34:42:06 Olli Carpen, Department of Pathology, Helsinki University Hospital

We want to learn how chemotherapy, the treatments that the patients receive, affect liver functions and other functions, and thereby improve the probability of successful patient treatment.

00:35:03:07 V.O.

The key to further progress lies in developing algorithms that can bind genetic and metabolic data.

00:35:11:23 Olli Carpen, Department of Pathology, Helsinki University Hospital

Genomics is something that is stable but metabolomics changes all the time. That's why it's very important to combine the different information and then the health data.

00:35:29:02 V.O.

In order to obtain this precious metabolic information, Ollie Carpen calls in a startup founded just four years ago. A blood sample allows over 200 biomarkers to be recorded in a matter of minutes. AI then allows us to determine the effects of a given cancer treatment. Cardiovascular disease and diabetes can also be detected much earlier.

00:36:05:05 Teemu Suna, General Director, Nightingale

If a doctor prescribes to you a lifestyle intervention, it can be exercise, it can be diet, it can be anything. We just kind of follow up on a biomarker level what is happening inside your body. We can actually show how your risk of getting a chronic disease, like diabetes or cardiovascular disease, how the risk is changing. This is entering the era of personalised medicine.

 

00:36:37:06 V.O.

 With AI, each patient will benefit from tailored treatment. Preventative advice can also be given to help patients make lifestyle adjustments. As healthcare becomes evermore technologically sophisticated, the ambitious goal of predictive treatment moves closer to becoming a reality.

00:37:11:21 V.O. (cont.)

Welcome to Shenzhen. 90% of consumer electronics sold worldwide are produced in this city. Smart phones, tablets, and video game consoles are all manufactured by the bucketload. All of the US tech giants produce their goods here in this colossal workshop. Boutique electronics are assembled and packaged alongside their bargain basement imitators. The city centre boasts a gigantic electronics bazaar, which offers such a huge selection of goods that this group of visiting Swiss students don't even know where to begin.

00:37:48:01 Yann Feo, Software Student

It's amazing what you can find here in just one day. You go one morning and you find everything you're looking for, It's like going shopping for groceries, but you're buying electronics. They have everything and anything here, you just need to know where to look.

00:38:06:21 V.O.

These students are studying electronics, design or project management. For them, Shenzhen is proving a bigger draw than Silicon Valley. Back in Switzerland, they worked on developing a number of projects. In Shenzhen, they work on adopting them to the realities of the Chinese market.

00:38:31:22 Chloe Dickson, Microtechnology Student EPFL

What we learn here is how to streamline the prototype, so it's cheaper and easy to produce, and also quicker. Things move a lot faster here than in Switzerland, it's quite different. So everyday we come back with a new design that's simpler, and easier to manufacture.

00:38:48:07 V.O.

here they work in a maker space, a startup style, digitally focused workspace. AI technology is their bread and butter. This engineer has come from Jordan to market his invention, a pocket spectrometer.

00:39:06:11 Inventor

We opened the application on mobile. We scan some food and you have here some tomato.

Is it real?

-Yeah, I can eat it, also.

Just like point it to the device like this, and here you get some...

-Nice!

like some information about tomato, so you can see how much carb, protein and fibre is inside this tomato, and the most important thing, what is the freshness for this. So that's why it's very useful for people. It's like the same algorithm used by China to identify faces by getting one face, two faces out of 1 million or 2 million. So that's how we use deep learning and AI, the spectrum analysis.

00:40:00:00 V.O.

In China, it's all about economies of scale. With 1.4 billion inhabitants, over half of which have smartphones, China is the home of big data. The whole country is investing in artificial intelligence. China is no longer satisfied with imitating its competitors. It is looking to take on the US for first place in the race for global technological supremacy. With a balance sheet worth $40 billion, the Alibaba group is an online retail giant, dubbed the Chinese Amazon. Alibaba is investing $15 billion in research facilities in China, but also in the US, Russia and Israel.

00:41:03:16 Xiansheng Hua – Head of A.I. and associated Technologies, Alibaba

It's very important for the company to invest and without technology, without artificial intelligence, it's going to be difficult for the company, not only for Alibaba, but for many companies. So this is why we invest a lot to help customers, to improve their experiences.

00:41:27:02 V.O.

Like the look of this dress? With one click, Alibaba offers a myriad of options. It takes mere seconds to review millions of items listed on its website.

00:41:42:01 Xiansheng Hua – Head of A.I. and associated Technologies, Alibaba

So basically the technology by hand to recognise what is in the picture and where the product is, we know where the product is, and then, uh, and then extract features. And then we compare with, or the product that we have done send back, uh, the same or similar products to our customers.

00:42:06:17 V.O.

The Chinese are now even more dependent on their smartphones than Western countries. A host of vital services are centralised within a few online platforms like the widely popular WeChat, which boasts 1 billion monthly users,

00:42:29:10 Jiang Fei Li

WeChat is first and foremost a social media platform, but a lot of people use it to communicate at work. There are so many apps for doing all sorts of things; there are so many service on offer. For the most part, no one carries cash. Everything is done through our mobile phones. When people do their routine grocery shopping, they pay with their phones. if we're going on a trip, we use it to reserve a hotel. For example, I have an app for booking accommodation online. Nowadays, no one can get by in daily life without a mobile phone. Obviously apps collect my data, if I rent a bike, the app can Geotag me. And since I make all of my payments with my phone, all of that information is collected. But I trust in the security behind it, so it doesn't bother me.

00:43:42:03 V.O.

Consumers can't get enough and show little concern for their own privacy. All the better for ambitious corporations who strive to understand and adapt to consumer behaviour. This firm is specialised in processing data from smartphones. The numbers are dizzying. 731 million users are scanned every month. Whole towns are put under the microscope. This is Harbin home to 10 million people. This graphic shows the flow of commuters block by block, but data collected from smartphones can tell us so much more about an individual's behavioural patterns.

 

00:44:34:07 Tony Yan, Vice President, TALKINGDATA

Because there are some sensor data on the mobile devices so we can focus on what they are doing now. Maybe they are running, maybe they are walking. For example, if their location is near a shopping centre, so we make inferences as they are shopping. And if they are in the park, so maybe they are relaxing. If they're often travelling, maybe it's a businessman. So we use data to make some inferences based on the data we collected.

00:45:04:20 V.O.

Profiling in this way greatly enhances the potential for targeted advertising but it doesn't stop there.

00:45:14:23 Tony Yan, Vice President, TALKINGDATA

For example, Yum, the biggest fast food company in China, if they want to open a new store, so they can circle an area, and based on what the data we have collected, we can see the users come to the area, can match their current. If the two groups of people are similar maybe it's a good place to open a new store.

00:45:46:00 V.O.

The amount of data transmitted by cell phones is constantly on the rise. Thanks to an ever increasing number of apps. Ant financial is a subsidiary of retail giant Alibaba. The company has developed software that assesses damages after a car accident. The app finds and assesses the damages in the blink of an eye. The new report is sent to your insurer with the click of a button. The app also calculates how the claim will affect your premiums. Then it selects local garages to carry out the repairs. It saves the customer time while saving the insurer money. Since the app can detect inconsistencies and flag potentially fraudulent claims.

00:46:39:05 Xin Guo Senior Engineer, ANT Financial

Normally there will be like 10 to 20% operational costs is caused by the fraud case. So once when we solve this fraud case problem, the insurance company can save tens of millions Chinese Yen each year.

00:46:59:09 V.O.

The algorithms being developed by ant financial can also offer fast and convenient credit services. This scooter vendor has just taken out a loan to renovate his storefront and purchase new stock.

00:47:14:21  Pingen Zeng, Electric Scooter Sale

At first I tried going to the bank, but they needed lots of documentation from me. There were so many procedures to go through, it took forever. It would've taken a week if not a month to get the loan.  On top of that, they require a guarantor. Whereas with this, I don't need one - I sort everything myself.

00:47:40:07 V.O.

In a few moments. The app evaluates his profile and accepts his application.

00:47:45:19  Pingen Zeng, Electric Scooter Sale

I click on the app, then I select 'online loan', I look through the available funds, and I hit 'borrow'. Then I choose the, amount I need, this shows the payment window, six or twelve months. I recieve my payment code, and it's done. The transfer will come through soon.

00:48:14:00 V.O.

This App is aimed at the millions of Chinese who live without a bank account. By offering them a chance to borrow on credit. Behind this express financial services system lies a complex mesh of algorithms. It scans all given data on a client to build a predictive model that evaluates the liquidity of any given startup. Entrepreneur.

00:48:45:22 Wei Chu, Senior Technology Expert, ANT Financial

We use AI to create what we call a 'credit access program'. We create a profile on our customers. We use a lot of data drawn from their online behaviour, and with their consent, we add more personal data to the model. This allows us to define predictive behaviours for our new customers.  Using AI to process this information, we can study these complex relations, and produce predictive behavioural categories. So when a new customer approaches us, we can assess their application and the risks involved by comparing their data against our own models.

00:49:32:14  V.O.

These kinds of systems exist also in everyday payment apps such as Alipay. These algorithms not only measure your ability to pay back money borrowed, but also integrate your consumer habits into their ratings.

00:49:50:14 Jiang Fei Li, Electronic Management Student

Alipay has a credits system called 'Sesame Credit'. If you use it every month, your score goes up. As you buy things, your score keeps rising. I think that this credit system is a good thing. It looks at my credit history, and calculates my score based on a number of criteria, like my solvency. It also lets me look at other people's score. This lets me know whether they are trustworthy or not. It's a great system.

00:50:23:02 V.O.

This app allows you to judge others based on their credit score. A once private piece of information turned public. A new social evaluation tool. Many companies are developing such algorithms which assessed citizens on opaque criteria, a development roundly denounced by this economics professor in Beijing.

00:50:55:04 Prof. Lian Cheng – Chinese academy of Social Sciences

If you're got a very low score that means you will be excluded form many services, many online businesses and even in real life you will be excluded from both business opportunities and also it’s even hard for you in daily life sometimes. Because, if some people think you don’t score high in the ranking system they think you are not a good friend. Also in some dating websites they also can show your citizen credit data score to people and if they saw a very low score they would think you are not reliable and they will refuse to meet with you.

00:51:42:08 V.O.

Finance apps aren't the only ones scanning personal data. Public life is under constant scrutiny by video surveillance. China is home to 200 million security cameras, often equipped with facial recognition technology. This intersection in Shenzen is one example of its applications.

00:52:16:09 Tourists

See, there’s a camera there. Look what happens when you cross on a red light. Look at the screen. It’s a total invasion of privacy.

It’s a bit off  - the government has a database of everyone’s face and can use it to identify them.

To be able to match people in a matter of seconds is impressive.

Especially when you consider all of the data it has to search through.

00:52:46:14  V.O.

This installation is a tool for social control. As soon as a pedestrian crosses on a red light, the camera picks up their image and searches the state's databases for an ID match in a matter of seconds. The name and photo of the offender is displayed for all to see on a giant screen. Facial recognition technology relies on a massive data collected by the authorities on each citizen. The transition to the digital age has been a gold mine for developing algorithms that work alongside smart cameras as the state partners with the private sector.

00:53:33:19 Allen Lu Fan – Cofounder of Senscap

The Chinese government owns a lot of a big data. Faces, for example, or the personal datas. So they would love to be involved into this big development of AI by offering those big data to work with start-ups like us. So this is a system that we developed for spots for the police department. We have a camera in the glass to find people in the crowd. So if you are in the database, for example, your fugitives or missing people like missing children, if you show up, that machine will shake. What remind you that you found the people in the database.

00:54:30:10 V.O.

Cameras can also be taught to analyse a person's movement patterns, allowing them to identify suspicious body language.

00:54:49:23 Allen Lu Fan – Cofounder of Senscap

All the public areas, The system will understand some certain patterns. For example, people start to run or people start to fight and uh, the camera and the sound can work together. The vision and the sound can work together. If the system figure out, detect any special sound, for example, a gunshot are people yelling for help, then the system will automatically turn the camera into the direction of the sound so that the system will know what is the accident or incident happening. The whole city or the system will become like a game. In the game, Whatever you do, you are recorded and the system understands what you're doing. So in the future of the city, we're understand the people's behaviour. If any crime happens, it will be stopped in the real time rather than afterwards.

00:56:07:05 V.O.

Total digital surveillance is no longer a hypothetical relegated to the pages of science fiction thrillers. The Chinese government has announced the implementation of a so called social credit for 2020 each and every citizen will receive a score based on data derived from their behaviour and habits. To find out more, We spoke to a rare dissenting voice. This renowned historian resigned 20 years ago from the Social Sciences Academy. Authorities are pressuring him not to speak with foreign media. He agreed to meet us in a safe place away from the all-seeing eyes of the Chinese authorities.

00:57:00:18 Zhang Lifan, Historian

When I was using my phone to do certain things, I noticed that my phone would start to heat up, and my GPS system would suddenly be switched on. That is surveillance at work. As things stand, I try and avoid meeting up with friends, so as not to draw attention to them. The authorities can see everything I do, every move I make, thanks to the cameras. I believe that we are living in a situation comparable to George Orwell's novel, '1984'. Except that this is like '1984 2.0', a more extreme version.

00:57:55:19 V.O.

Social credit will integrate your online behaviour and social media presence. If your friends, the sites you visit or the content you share are deemed non-conformist, your score will decrease. Allow your score to fall too low, and you can be denied rental accommodation, a bank loan or a job interview. Even long distance travel will be curtailed.

00:58:22:23 Zhang Lifan, Historian

In the past, the government monitored people who it believed were a threat. Now it has extended that level of surveillance to every single person. This will limit a lot of people's freedoms. If you are barred from catching a flight, or making online payments, your life will be entirely under the control of the state. I think it's much more serious than anything that came before. Before the advent of these new technologies, it was less all encompassing. Traditionally, the Chinese people think it natural that the state should have the power to manage them, and that they should obey. That is why, in the past, the Chinese have rarely struggled to secure rights for the individual. For the time being, the majority of the population remain indifferent.

00:59:32:22 V.O.

Is it ignorance or indifference that has kept many Chinese in the dark regarding this sinister new development in state surveillance? Either way, one can't help a sense of foreboding when confronted by the full spectrum of potential in mankind's ambitions for the future of big data.