this is bbc news, the headlines: prince andrew has told the bbc he categorically denies having sexual relations with virginia roberts, the woman who says she was forced to have sex with him when she was 17 years old. he's spoken in detail for the first time about his links with the convicted sex offenderjeffrey epstein. chinese communist party documents leaked to the new york times have revealed new details about the mass detention of uighur muslims in xinjiang. china officially describes detention

facilities where uighurs are being held asjob training centers. but the internal documents illustrate the calculated and coercive nature of the ongoing campaign targeting muslims in xinjiang. protests have erupted across iran after the government unexpectedly announced it was sharply increasing the price of petrol and rationing supplies. the country's interior minister says security forces will act to restore calm if protesters damage public property. now on bbc news, it's time for click. this week: the perils of the gig economy. checking for allergens. and...fake birds.

it's kind of hard to remember a time when we didn't have taxi—hailing apps. and when i say taxi—hailing apps, even though there are many more players out there, it is uber that comes to mind first. at its conception a decade ago, uber was really disruptive. i mean, the idea that you could have a taxi to you within minutes, it would know exactly where you were, you could see where it was and you didn't even have to have any money on you. i mean, it was revolutionary. the company grew at a rapid pace, becoming the highest valued start—up in the world. this without ever turning a profit. in fact, in the last three months alone, uber lost an eye—watering $5.2 billion. undeterred, uber continues to expand

and has its name stamped onto many apps that provide different types of services, all part of the so—called ‘gig economy'. now, it has faced a lot of backlash in many of the countries that it operates in from taxidrivers who have been losing out because of the platform's aggressive pricing strategies and from city authorities who've raised concerns over workers‘ rights and passenger safety. here in london, the transport authority says it too has concerns about passenger safety, and it will decide later this month whether to renew uber‘s licence. in the meantime, carl miller has met up with an uber driver who has concerns too, although this is about how uber controls his livelihood. it's monday morning and i'm catching a ride with hadi. another one. like so many parts of the digital

world, the gig economy was supposed to be a liberation. you 0k? apps like uber were supposed to transform how you worked, work when you want, where you want. but now, many fear that whether it was either in the platforms and how they work or the data and how it's collected, they don'tjust represent a liberation, but also something else — a potent new form of control as well. what i was told is that the closest driver gets thatjob, but i don't believe that to be right, ‘cause what happens is i've seen customers sitting in my car, trying to book a ride and it's not bouncing to me, it's actually going to drivers who are far away, five, ten minutes. that was something i really couldn't believe, so we gave it a go. but although i was physically sitting next to him, the job went to someone several minutes away. uber has now introduced a system that aims to reduce the waiting time

for everyone, not just a particular passenger. and this may lead to the counterintuitive situation where your driver can get to someone else quickly and another driver can pick you up soon too. confused? well, so is hadi. and although the driver app gives some information, he's struggling to understand what factors really determine how work is allocated. in his five years of driving, the work has become scarcer. it's becoming even more important to hadi to understand the algorithm that actually allocates the work that exists. important, but also unknown. you drive around all day, thinking maybe that's the best way to beat this algorithm or to meet up with the algorithm that has set. carl, i don't know what's going on. on an average, i used to work six to eight hours, five to six days a week. the number of days have not changed but the hours have increased. it will still be ten to 12 hours, five to six days a week. after costs are factored in, hadi

says he and many of his colleagues are often struggling to make even the minimum wage. i've seen drivers sleeping in theircar parking, actually, and you see them, basically, they have gone to sleep because they have overworked. not only for hadi, but plenty of other drivers as well, is actually, if you think about it, the algorithm that lets him feed and clothe his family. it's cold, hard maths, but with tremendously human consequences. unfortunately, we all depend on the algorithm. what we want from it — to be fair, to be transparent... that's the most important thing. there was only one way for hadi to actually figure out what's been going on — asking for his data. and when he got it back, it made things even more confusing. james farrar established the worker info exchange to help people across the gig economy to actually make sense of their data. he told us the information hadi

received refers from everything from speed to battery level, but, crucially, doesn't reveal the things he really wants to know such as rates of pay or the actual time spent on the platform and how to optimise his chances of earning more money. drivers always want to understand that they're getting a fair deal, that the value, the quality, the quantity of the work is fairly distributed. well, uber has always proposed to its workforce that the workforce, drivers are their own boss, they're free to make their own choices, they are effectively running their own business. but if that's true, then i must be able to access the endless amounts of data i'm creating for uber every day. but could it also be a matter of the complicated and hidden nature of the information algorithms involved? the way in which, for instance, drivers and passengers are matched via a platform is a complex calculation, and so the gig economy today is considerably more complex

and that means that the questions around how scores are calculated, how data are utilised and how that can be taken into account by service providers — that question is more important than it ever was. but a joint study between oxford researchers and uber itself found that on average, drivers earned above london living wage and reported they were happier than the average worker across the city. critics question, though, whether the full costs of being an uber driver have really been factored in when those figures were arrived at. the same arguments now playing out in the streets of london have happened in the city after city across the world. in what might have been a global first, the powerful taxi and limousine commission in new york didn't just ask uber for data, but demand it, and until uber handed it over, they were banned from operating. what we found out was that conditions were worse than what was being

described to us by drivers. 96% of drivers were making less than the city's minimum wage, but without that information, you only have anecdotes, you have stories from drivers about low wages, but you have no way to really quantify that, and without quantifying it, you can't create a policy to bring those levels of wages up. in response, uber said: another brilliant report from carl there. and, carl, if you could just sum up your thoughts about whether the gig economy really is eroding workers‘ rights, notjust pay, but workers‘ rights in general? i mean, in some sense, it really has to. i mean, these are kind of old, old issues that seem to be being now raised in different ways. you know, it used to be struggles of employees over rights. now, it‘s the struggle of workers to be recognised

as employees at all. it used to be struggle over control of the shop floor, now it‘s struggle over data. these are issues, really, that if you think about it, stretch back hundreds of years. but they‘re now being raised in all these strange new forms as the work increasingly transfers onto these platforms. do you think we‘ve been foolish to believe that the gig economy will be better than having traditionaljobs? well, it‘s that paradox, it‘s both a deliberation and a new form of control both in the same thing. so, we believed part of it and we were right to — it is bringing flexible new forms of work, but it‘s also bringing these new hidden forms of control, i think, which because they‘re kind of working through the tech, through the complex way it works and in data, it‘s actually quite hard for us to see or to kind of recognise is there in the first place. ok, if we can move on — in a few weeks‘ time, you‘re going to be doing what i think is quite a groundbreaking report for us here on click. so, in the week of the general election here in the uk, carl is going to be looking at the type of outrage and anger

that flows across social media to and from members of parliament and candidates — is that right? that‘s right. i mean, this is the free—for—all election. i mean, i don‘t think we‘re ever going to see a larger gap between the influence of the digital world on the outcome of the vote and the influence of laws on the digital world. it‘s a free—for—all. the digital world has become unbelievably important for campaigning, for the candidates and, of course, for everyone to learn more about politics. at the same time, the laws just simply haven‘t kept up, they haven‘t extended to the digital world. so, we‘re going to desperately be trying to track, i think, one of the major social outcomes of consequences of what‘s happening, which is simply that everyone is getting angrier with almost everyone else across society. it seems at the moment that we‘re massively polarised and we‘re also very angry. do you think the two go hand—in—hand, and maybe in a future where we‘re not so polarised, we won‘t be as angry, or do you think everything and everyone isjust getting more hyperbolic, angrier online? and that will lead to just more polarisation forever. i think it‘s this powerful yet intricate interplay between the digital world and everything else

happening in society. so, there‘s lots of things which have nothing to do with digital technology for why we‘re angry. of course there is. but i think it‘s the platforms we live on, the way in which information flows around is actually part of that story as well. we‘re only seeing certain kinds of information, we‘re only speaking to certain kinds of people, we‘re seeing cherry—picked facts from the other side being slotted into our own timelines and so on, and all that is part of the story. the platform engine has visible rules which determine what you see, what you experience online, how these platforms work — that is definitely, i think, leading us towards anger rather than the opposite. look at the glee in his face! we look forward to it, carl. good luck with your research. thanks. hello and welcome to the week in tech. it was the week disney officially entered the streaming market. well, it didn‘t quite go to plan. disney + finally went live in the us, canada and the netherlands, but customers reported technical issues with many unable to connect. disney said demand had exceeded its highest expectations. maybe ralph really did break

the internet after all? in the fastest backflip and u—turn since. . .well, sonic, the updated and redesigned hedgehog has been officially revealed in a new trailer for the upcoming live—action movie. the original trailer drew a deluge of complaints and mockery over the original cgi design of sonic himself, forcing animators — quite literally — back to the drawing board. from spinning hedgehogs to backflipping robots. these footballing flipping robots from mit are called the mini cheetah. its creators claim it is virtually indestructible and can right itself if it falls down. as well as some smooth soccer skills, it‘s also capable of working over uneven terrain twice as fast as a human. let‘s hope it can‘t climb trees. and finally, in other robot news, if you‘re one of those people that don‘t like speaking to shop assistants, maybe you‘d rather direct your questions to one of these welcoming faces instead. this humanoid shop assistant from russian company promobot can apparently show emotion and they claim they can make photorealistic clones like these

arnold schwarzenegger and albert einstein dolls. greeted by these in store, would it be hasta la vista? or will you be back? you decide. for those with serious food allergies, knowing exactly what you‘re eating can be a matter of life or death. when it comes to packaged food, the ingredients are normally clearly on the label, plus a warning if it may contain traces of nuts or any other allergens. but when it comes to eating in someone else‘s house or in a restaurant, things get a little bit more complicated. so, if you want to add an extra level of checking what those ingredients are, well, i‘ve been putting some technology to the test that might be able to help. this is nima. now, there‘s a version that tests for gluten and another that tests for peanuts. the idea is that you put in a small

sample of the food that you‘re eating, as small as a pea, into one of these capsules, that goes inside the device, which syncs up to your smart phone, and you can find out whether the ingredient you can‘t eat is in it or not. i‘m going to put both of them to the test with this cookie, which should contain gluten but shouldn‘t contain nuts. the device uses antibody—based chemistry born out of mit technology to detect proteins or allergen. the company‘s algorithms then translating complex science into a smiley go ahead and eat it face — or not. this is a pricey occupation, though. each one—time—use capsule currently setting you back five whole dollars. and the company does advise that this is an extra level of checking on top of your normal due diligence and, of course, carrying any medication. ok, well, i can confirm that the device definitely got this correct. it says that gluten has been found. it comes up here on the device

and you can see here on the phone, 12:30pm today, gluten has been found. if i tap on that, it gives me the option of notjust making a note for myself so i remember, but also, sharing the data to the nima database. and, of course, as more people use these devices, that database will start to become a lot more valuable. let‘s give the peanut tester a go. you can do this with liquids or solids. and we have a result in the form of a smiley face. during my limited experiments, the results were accurate. but i have of course only tested a small piece of each bit of food. so i‘m working on the basis that the ingredients are consistent throughout. whilst they are an entirely different entity, food intolerances can have a huge impact on people‘s lives, too.

i tested this at prototype stage but now, i have the finished version of foodmarble here. this is a digestion tracking device. now, the way it works as you breathe into this little hole here and it will track how much hydrogen there is in your breath. the idea behind this is that if you‘ve eaten something that you haven‘t managed to digest properly, then a small amount of hydrogen is released into the bloodstream. that makes its way into your lungs, and from your lungs, into your breath. so you can figure out which foods might be affecting you negatively. once you have a reading, you can sync that data up within the app with any sleep data or how stressed you‘re feeling, so that you can measure up the factors to see if there is any correlation. you can manually make a note of symptoms, too, and you can log all the ingredients that you‘ve eaten in a meal. so, how much use are these readings, taken at any given time? i think that the danger of that kind of result or variability would be that it might end up with you not knowing where you are and restricting your

diet in all sorts of ways. and you could compensate so that you didn‘t lose weight, for example, and become obviously malnourished. but subtle deficiencies of micronutrients that are important. but there is a secondary use the device hopes to fulfil and that should be able to really isolate the issue. the device can also be used to do food intolerance testing. now, for that, it would require some fasting and you use a different mode that‘s within the app but it comes with these samples of inulin, lactose, sorbitol and fructose to test the common intolerances. it does seem to be as accurate as the big machines we use to measure hydrogen in the clinic or in the laboratory in the hospital. my problem with it being practical is that there are many people who have primarily methane—producing bacterial populations and what if you‘ve actually got something really badly wrong with your intestine? and you‘re putting down all of these symptoms —

irritable gut, nerves and intolerance of foods. and yet, you‘ve actually got crohn‘s disease or coeliac disease. or what if you are a slightly older person and actually, you know, you think you‘re too busy to go to the doctor‘s with these symptoms that are abdominal and yet, you‘re sitting on a cancer somewhere? so, of course, these devices don‘t eradicate the need for a doctor‘s diagnosis or checking what‘s in your food. but for some, maybe they could provide an extra layer of reassurance. that was lara. now, i wonder whether you‘ve seen this sort of thing going on recently. this is my face on donald trump‘s body. it was made using deepfakes — an ai algorithm which is capable of taking a face from one video and animating it onto someone else‘s head in another video. it‘s very easy to get hold of the software, but it requires technical know—how.

but recently, researchers in israel have published a paper that says faceswapping can now be done without all of the complicated training and processing. we train the different components in our algorithm on a vast number of human faces, such that it is not adjusted to a particular person but actually, adjusted to any human face. one of the main dangers of this kind of technology is that if it progresses much, much more — and it will — and it will be utilised, but those who are interested in undermining the public‘s trust in news, then it will be able to undermine a new domain. as the tech behind faked videos gets more sophisticated, faked speech is also becoming scarily realistic. mouths can be animated to match words and the words can

be faked, too. now, most of the deepfake videos are done either by cutting together genuine audio recordings of the subject or by using an impressionist, as this amazing video illustrates. voice of arnold schwarzenegger: dogged by his own echolalia. voice of morgan freeman: better to crumble into dust than wind up an impressionist. voice of robert de niro: the definition of cheap thrill — like watching farm equipment rust is watching an impressionist. but we‘ve also seen software which can listen to your voice, analyst it, then make you say things you haven‘t. even birds aren‘t safe from this fakery. well, in one art installation, at least. lj rich has been to somerset house to meet an artist who is deepfaking birdsong in a darkened room. birds chirping. so in this room, we‘ve

real birds in one area and fake birds generated by an ai in this area. that‘s a fake bird. whispering: that's a real one. birds chirping. listen closely, and you can hear several species like great tits, robins and chiffchaffs. artist alexandra daisy ginsberg trained her ai birdsong using a generative adversarial network, or gam. this system is often used to make new video from existing video footage, commonly called deepfakes. it‘s not the first time she has used tech to make the impossible exist. at her studio, there‘s a digitally revived rhino, a computerised mars landscape — as you do — and even the scent

of a long—extinct flower. but creating deepfaked birdsong is a new challenge for the artist, both technically and ethically. so this is what we used to train the dawn chorus bits. mm—hm. deepfakes themselves are worrying in some applications, deeply worrying. so you may have watched the mark zuckerberg sort of false testimony videos. ..all their secrets, their lives... the words look like they‘re coming from his mouth, and they‘re not words that he said. the system has been trained to mimic him. and in the same way in this project, we‘re using the same technology, essentially, to mimic natural birds. it‘s like trying to drop a pencil on its tip and expecting it to stand up on the end of a pencil. and that‘s how hard it is to train one of these networks. so, how do we make a deepfake bird? start with thousands of real sound recordings — a job in itself. and that‘s day one. static.

not — at this point, that‘s not going to be a great sound installation. next, play all of these snippets to the gam programme, which then argues with itself about whether something it makes sounds enough like, say, a robin. by day seven, we maybe have something more like this... chirping with static. add in generous amounts of computing time and you‘ll get something that sounds a bit like a bird... ..eventually. and then by day ten, our chorus is much more sympathetic. various bird sounds. it‘s sort of an ethical issue with this project as well. it‘s that we‘ve advanced the technology of deepfake by working with them. you end up with something that is very lifelike but not real, and that‘s an uncomfortable thing to listen to. now, reasonably keen ears can spot the real songbirds from the imposters. and although the end result is a technicalfeat, it also sounds a little strange. and as deepfakes continue

to improve, our best defence is to educate ourselves, and each other, on how to discern what‘s fake and what‘s real — whether it‘s birds or people. that was lj, tweeting furiously from somerset house. and that‘s all we have time for this week, but if you do want to get hold of us any time during the week, don‘t forget we live on social media. on facebook, youtube, instagram and twitter, @bbcclick. thanks for watching and we will see you soon. in the news last week, you might remember that venice was hit with severe flooding,

and we have got similar weather conditions on the way for today. later on, around midday, we will see the high water coming in around 1.6 metres above normal, that makes it one of the highest of the last 90—odd years — perhaps not quite as high as it was on tuesday, but it will still be widespread, big flooding in the city. those flooding conditions are caused by an area of low pressure in the mediterranean, drawing up these strong winds, pushing a bulge of water from the adriatic. there will be some thunderstorms as well fairly widely in italy. that low pressure will get close to us as we will see injust a moment. for the time being over the next few hours, where we have clear spells, we are seeing things turn pretty frosty, particularly in scotland, the odd patch across parts of southern england as well. so for some, a chilly start of the day on sunday. for others, a cloudy start, with fairly persistent but mainly light outbreaks of rain affecting north east england, becoming quite patchy in nature as it works across northern england into the midlands and wales.

probably no huge rainfall totals building up, just some dull and damp weather. brighter on the south, we should see some sunny weather moving into scotland and northern ireland during the day, although there will be some showers across northern scotland driven in by a north—west wind. it will feel quite chilly here. through sunday night, the skies will clear further across the northern half of the uk really, so as that happens, we will see temperatures plummet. there will be a widespread frost. temperatures down to around “4 in edinburgh, colder in the countryside. some mist and fog patches to take us into monday. then that area of low pressure bringing flooding to venice later today, here is the low coming around europe and coming quite close to the uk on monday. this area of high pressure, though, it looks like it will stop its progress, but it is something we are monitoring carefully, given the saturated ground conditions and the flooding problems we have in places, any rain would not be helpful. at the moment it looks like it will be mostly kept at bay, so for monday, it will be

a cold start to the day, widespread frost and mist and fog patches, a lot of dry weather with sunshine, highs of around three degrees in edinburgh, you could just about see some of that rain getting into parts of eastern england. after that as we get into tuesday and wednesday, low pressures from the atlantic are going to take over. as that happens, the wind will back around more to a south—easterly direction, which means we will see rain spreading into eastern wales across the midlands, which of course is not great news. that‘s your weather. bye for now.

this is bbc world news. i‘m james reynolds. our top stories: prince andrew denies having sex with a 17—year—old girl. in an exclusive bbc interview, the duke of york speaks publicly for the first time about the allegations against him. i can absolutely, categorically tell you it never happened. do you recall any kind of sexual contact with virginia roberts, then or any other time? none whatsoever. angry crowds take to the streets in iran as petrol prices jump by 50% and fuel is rationed. leaked documents reveal new details about the mass detention