tv Click BBC News September 29, 2020 1:30am-2:01am BST
1:30 am
for viewers in the uk and around the world. the world health organization and the us have announced plans to rollout a rapid and new covid test. it could transform the global response to the pandemic. it should cost as little as $5 and show results within 15 minutes. a second wave of coronavirus cases hits europe and thousands across the continent are rejecting new lockdown measures. a day ahead of the first presidential debate, donald trump has dismissed claims he paid just $750 in income tax the day he won the presidency does make the year he won the presidency and is substantially in debt. the new york times say he avoided paying his fair share of federal income tax for several years.
1:31 am
it is just after 1:30am. it is time now for click. this week: youtube cracks down on misinformation. nick sneaks up on a robot tractor. and omar's mask—wearing masterclass has been postponed due to unforeseen circumstances. theme music hey, welcome to click — i don't know about you, but the weather has turned very autumnal around these parts, and as expected the restrictions are being tightened up as the numbers here rise. so here we go then, on the way into winter. lara, how are you doing? well i've been a little bit under the weather this week — don't worry, not coronavirus —
1:32 am
but it did get me thinking that if i'm ever unwell it would be quite handy to just have a spare version. so i have been working on a virtual version of me. i mean, we are a technology programme. yeah, we are, but we have always said this is the danger, right — if ever they create a convincing version of us then we're out of a job, right? hi spencer, i'm virtual lara. ok, you may not be fooled but do you think anyone else will notice? what was that? where did that come from? i was created by ai video generation platform synthesia — it seems anyone can have a virtual them. what do you think? i think that is absolutely incredible. i think the only thing that didn't fool me was the lip sync, your mouth was doing something weird, like you had just come out of the dentist. but that is brilliant. can we bring out the real lara? what have they done here, have they basically just reanimated your mouth? yes, but it was quite frightening to see how wonky my mouth actually is, i have never
1:33 am
noticed that before. but the purpose of this is it could be used for something like corporate training videos, where a familiar face or even calling people by their names could be helpful. 0k, how did you make it? i had to make a video of me presenting the exact same thing five times. now to do this i needed to follow their guidelines, which included sitting rather still, not moving my arms and not getting too animated. which for me felt rather unnatural — so i was expecting the end result to be a little bit strange as a result of that. but once you have made that video, they are able to use it with all your different mouth movements to add whatever sound you have given them, making it look as though you are actually saying it. now they can do that using one of their voices, or your own. and it also means that there is the opportunity to be able
1:34 am
to give you different accents or make you speak in different languages. american accent: hi spencer, i'm the new virtual lara. speaks spanish si, si, excellent spanish! oh, thank you! someone who can tell us a little bit more about the purpose of this is synthesia's ceo victor riparbelli. hello, victor. now, we have looked at your technology before on the programme, at the moment there is an option of voices on the platform or you can pre—record your own. but how about an artificial version of our own voices, is that something that's likely to be possible soon? yeah, absolutely, that is the natural next progression of these types of technologies. we can now sort of replicate your image in a very believable way, and soon it will be your voice as well. the voice technology is kind of there today, but it requires a lot of audio to really work, usually the kind of, the rule is something like 15—20 hours in a professional recording studio of you reading aloud a text. why is it so much harder to do voice than video? there are several reasons, but one of the reasons is, if you think about what a computer sees, right, then if you are replicating a video of you in the kind of style that you have seen with lara, there is, the kind of space of possibilities is quite constrained, right?
1:35 am
you're replicating a video of someone sitting here and they are speaking and you sync their lips and head movements and things like that. if you try to do something like this with a voice, then the training data will be a lot more limited, so let's say even if you have two hours of you talking, that is still a very, very little part of all the things you could possibly say. so could i use your platform to upload a video of anyone, not me, and create a virtual version of them? no. why not? we need written consent from whoever is being uploaded. so everything goes through manual review and we only do custom avatars for corporate clients. so clearly you have protocols in place, but is the very existence of this technology, the fact that all of this is possible, not a pretty big risk in itself, that the wrong people will use it? we have seen these kind of things happen before. definitely, synthetic media as a whole definitely comes
1:36 am
with potential for misuse, that is something that is really important that we address. both from a technical perspective but also from an educational perspective. i think any technology can be used for good and bad, i am not i think that deepfakes or video is uniquely different from being able to forge text or images, or tweets, which has been possible for the last 20 years, but it is certainly something we need to be really aware of. thanks, victor. so the reality is that making fake videos has become a lot more sophisticated. and we are already all—too—familiar with how false information can cause real harm, and has damaged people's trust in news. and even though most of the conversations have been centred around facebook, youtube has also been a big contributor to the spread of misinformation.
1:37 am
now the upcoming us election is going to place even more emphasis on this issue. and so the bbc‘s specialist disinformation reporter marianna spring has been finding out more. youtube has an incredibly smart algorithm. at least smart enough to make the company a lot of money. more eyeballs on a video and more engagement equals more cash. it's simple. but what is not simple is how potent this combination has become. videos from fringe groups with extreme ideas surfacing to the top and influencing users' views. the pandemic has pushed the issue to new levels, with false information about the coronavirus often finding a much larger audience than trusted sources. as the months have gone on, youtube has become inundated with documentary—style films from pseudoscientists promoting conspiracy theories. this plandemic documentary received millions of views. despite efforts to remove it from youtube, facebook and twitter, users were constantly re—uploading the clip.
1:38 am
and recently demonstraters have taken to the streets in london to promote some of the most popular coronavirus conspiracy theories mainly seen online. a man who would like to remain anonymous got in touch with us after his mum decided to go along. mum had two posters, one read "arrest bill gates for crimes against humanity", the other had a qanon hashtag, "save the children". she was first taken in by coronavirus conspiracy theories on youtube and now she has been radicalised by qanon on there too. it is so hard to have a normal conversation. qanon is a conspiracy theory that suggests president trump is waging a secret war against satanic paedophiles in government, media and business. since the last us election, the social media giants have had a bit of a reckoning. they are all trying to clean up their act to stave off heavy—handed regulation. facebook with its oversight board, twitter with more robust warning labels, and now youtube says it has added a new tool
1:39 am
to its arsenal to combat misinformation. it has partnered up with fact checking sites to warn users if the phrase they are searching has been refuted, and pointing them to information by a trusted source. put simply, it is a tool where when users go to search for a particular topic on youtube, if there happens to be a fact check, these fact checks are generated by third—party publishers, that will trigger in the results of that particular query, right at the top of the query, and it will link out to that fact check, saying whether that particular claim is false, oftentimes fact checking publishers have a rating of that type of claim, that will be displayed as well. this is one of the many tools youtube says it uses to reduce misinformation on the site. but none of these measures seems to be proactive and doing nearly enough to reduce exposure to misleading information. the challenge about misinformation, as you know, on a platform like ours, which is an open platform, we really do value the fact that it's an open platform
1:40 am
where there's a diversity of voices, is that with the changing nature of information around this pandemic, what is the new type of misinformation that might pop up? before there was the conspiracy around 56 and covid, who would have guessed that those two kind of randomly, separate pieces of technology and science would be linked together? well that happened in the course of the pandemic. so we had to adjust very quickly to change our enforcement guidelines. you say that these conspiracy theories are new and evolving, but a number of the ones that are still being promoted on your platform have been around for months. especially in foreign languages, including in russian and hindi, of false conspiracy theories about the origin of the coronavirus or even doubting its existence. for content that might not clearly cross the lines of our policies, we reduce that content in our recommendations. so if it is borderline content in nature, because as i said misinformation can be murky but we still want to reduce the exposure to it.
1:41 am
youtube says it has removed millions of videos containing misinformation from the platform, in many instances before anyone has viewed them at all. but there have been numerous examples of videos promoting false claims and conspiracy theories about coronavirus, especially in foreign languages. and those have stayed online for months, accumulating hundreds of thousands of views. the mozilla foundation behind the firefox browser has decided to act, this week releasing a new extension for both chrome and firefox called regretsreporter. the idea is a user can report any recommended content they find unsavoury. mozilla says this helps crowdsource research into youtube's recommendation problem, and it hopes it can force youtube to become more transparent about how its algorithm works. i think that youtube's algorithm is really one of the most opaque ones out of all the platforms out there, which is why they have a lot more work to do.
1:42 am
so youtube has come out with a lot of statistics like saying they have increased, introduced policy changes that work to decrease recommendations of borderline content by upwards of 70%. but the problem is that there is really no way to verify whether or not this is the case without involving the public, without involving researchers, sociologists, people who have knowledge that goes beyond just the boardrooms of silicon valley, that together we're better able to solve this problem of misinformation surfacing in recommendations. hello, and welcome to the week in tech. this week: us companies oracle and walmart are to take a 20% stake in tiktok. donald trump cast doubt on the deal unless they could have total control.
1:43 am
microsoft announced the $7.5 billion acquisition of zenimax, behind fallout and the elder scrolls. xbox said the publishers would be added to gamepass. an elon musk reviewed a new battery designed for tesla, which has claimed it will give the compa ny‘s vehicles a 16% greater range. police and the us and europe joined forces in a series of co—ordinated raids against illegal dark web marketplaces. the operation known as ‘disruptor‘ saw 179 people arrested and the seizure of over 500 kilograms of drugs, as well as 64 guns. could this be the future of flying? aerospace giant airbus has unveiled plans for it hailed as the first commercial zero—emission aircraft. the company said its hydrogen—fueled passenger planes could be in service by 2035. and finally, in ‘if movies have taught me anything, this will end terribly‘ news... a 18—metre tall 25—tonne robot has shown that it can move.
1:44 am
the gundam robot, based in yokohama,japan, was initially due to be the centrepiece of a new factory. but the coronavirus outbreak has delayed the opening. around the world, each country has their own social distancing rules. here in the uk, they're constantly updated. but there are a couple which seem set in stone. and that is: to maintain a two—metre distance from others where possible, and to wear a face mask in most indoor spaces. but, let's be honest, not everyone sticks to the rules. so, a number of tech solutions have been brought to market to try and help people stick to the guidelines. let's take a look at a few. ocucon have come up with an artificial intelligence solution that identifies when someone is walking towards the store and counts the number of people entering and exiting. once the maximum occupancy limit is reached, the doors lock and only open up again once there is more room inside. though it is probably one
1:45 am
for the bigger chains, as this solution will cost around £6,000. so, in a bid to make people feel safe enough to go back into stores, even smaller businesses are adopting high—tech solutions. videcon have created a system that can detect whether someone is wearing a mask or not when entering a premise. if they are, great! you can walk on through. and, if not, a screen displays a warning to remind them to put one on before coming in. the a! has trained using thousands of images of people wearing masks, but is still fairly new. so, just how accurate is this? so, as long as that stays green, it's working. and this works again. what if i cover my eyes? hold on. right. let's give it a go. green? green. oh, it didn't do it that time. i am gonna see if i can trick the ai wearing this.
1:46 am
man! i thought i was being smart. i thought by printing this out i could trick it into thinking i was going in without a mask. nah... that's...quite annoying. laughter. nose, mouth, nose, mouth. like that. my days, my days! so, i was able to fool that a couple of times, but the company does admit it has some work to do. we've found the system to be incredibly accurate using a whole host of masks. the mask you're showing me there has a picture of a nose and face on it. if we find there's problem masks, we can very quickly teach the system what that mask is and it will therefore be able to identify it. but this system is still quite expensive, costing about £3,000 to install. but with both solutions we've looked at so far, they can be valuable in another way, too. when members of the public are approached and requested or asked politely to put a mask
1:47 am
on, i think that can be quite a personal thing. they can start acting a little bit hostile in the worst—case scenario. since we've had the system installed, it's taken the confrontation out. so i definitely would say mental health both with the team and with the customers has been improved. now, what about the outdoors? this temperature—sensing helmet is something we featured briefly on the show a few months ago. and we've finally been able to get our hands on it. just behind the visor is an augmented reality eyepiece that displays an overlay on what is in front of you, telling you the temperature of a person. and if they're over a certain limit, it could be an indicator that they may have a symptom of covid—i9. developed by kc wearable, this helment is primarily used by police forces in china, but now also being used internationally to survey areas with a large volume of people. now, there's something in common with all of these products. maybe you've noticed it. all of them were designed before the pandemic,
1:48 am
with other users in mind, and are now trying to quickly adjust their devices to market them as something to help to fight against the spread of covid—i9. so, there needs to be a way to test these products, to see if they've been repurposed effectively. and that's why we're here today, at a research event where they be looking at these distance—sensing devices, to see how well they alert you when you're within someone's two—metre proximity. and they're doing that by running through three different scenarios. the first is ‘the chat in the kitchen‘, where two people will move in closer to someone standing by the sink, to see how quickly the devices go off, once they move within two metres of each them. beeping. second is ‘the walk down the corridor‘, where two people will walk past each other. this is done at different speeds, to see if the devices can pick up such a quick passing by. and the third is ‘the close but not close enough‘, to see if the devices can tell that a wall separates the two people in close range
1:49 am
and instead won‘t go off. now, these accessories use one of three different kinds of radio waves to detect another. the well—known bluetooth, the lesser—known ultrawide band, which is used by many tracking devices and now also appearing in smartphones, or a combination of both to maximise range. bluetooth devices didn‘t farewell and failed all the tests. it isn‘t very accurate if the devices are obscured. ultrawide band ones performed very well, as its accuracy is roughly 5—10 centimetres, and the combo products did well, too, but all of them failed the wall scenario. and that‘s not good enough. there is a lot of hype around iot, and the fear barrier is really low. so you are going to find a lot of devices pretending to do some things, that actually, they don‘t deliver. proper evaluation of devices and solutions is important. many companies are bringing
1:50 am
in their solutions to the market, and as keen as they are, whether their main intent is to sell or to help, the most important thing is that it works. that was omar. how interesting was that? and now to something else that i‘ve really been looking forward to. it‘s only nick kwek‘s continuing journey across the australian outback in search of tech. yes, he‘s back in a field. this time in queensland‘s central highlands, with some robots for company. plus, a spot of cutting—edge al to look at. we're a start—up technology company building autonomous robots. and we're also a fully—fledged farming operation here as well. so we grow wheat, barely, chickpeas, mung beans, sorghum, and we run cattle as well. we're farmers and we actually wanted to change the way we farm. we got bigger, we got heavier, and i think we were compromising the way
1:51 am
we were doing things. so, yeah, we started swarmfarm to make things — or make our machines a lot lighter. our spray rigs — they started off being about seven tonne, i think we were up to 27 tonne, the erosion that those wheel tracks were making was huge. jochie and andrew bate realised their traditional tractors were compacting their soil too much, and in the face of increasing herbicide resistance, decided to make their own. and they really went for it. weighing in at a slender 2,000kgs with a 75—horsepower diesel engine, fifth—generation bot november autonomously prowls paddocks on the hunt for invasive weeds. so, november‘s got 13 different infra—red cameras dotted along its back boom. and when it dries over weeds, it sees them and sprays some herbicide out one of its individual 65 nozzles. with such precision it can target those ha rd—to—kill weeds and i‘m told it can cut harmful herbicide usage by up to 98%, drastically reducing run—off.
1:52 am
it can work day and night, it doesn‘t need a lunch break, and unlike me, it‘s not heading to the pub at the end of its shift. and what are these here? so, at the front of the machine we‘ve got these cameras, so we‘ve got three different lots of cameras, and they‘re actually our obstacle detection. so, what they do is build a ground plane out in front of the machine, and then any obstacles are built off that ground plane. so, if there‘s something that comes up that‘s unexpected, it will see it and stop. we try and feed the robot as much information as we can before it gets out there, but this is just one of those safety measures we‘ve got in place. so i don‘t get run over? yeah, so you don‘t get run over. developing the software on the farm is — puts us leagues ahead, i suppose, in terms of the practical applications of the technology. if we were in inner—city or somewhere where we could only go testing every three orfour weeks, there‘s already a delay in that feedback of that software. with the training ground on their doorstep, it means
1:53 am
developers in there can rapidly prototype and see their code in a real—life scenario right out here in the field. for us, we can write something up and have it out on the robot within an hour. out in the real world, within a couple of work hours after that. working here at gindie, it makes the process a lot faster. we actually want to get them out of air—conditioned comfort and into the paddock. they can see how it's actually gonna respond to the environment or whatever is in front of it. you know, wash—outs or — i think every paddock and every industry and every situation is so different. and the bots are customised for different industries. they‘ve got them on macadamia nut farms, planting cotton, mowing grass on turf farms, and they‘re now turning there laser focus to orchards — literally. the technology we‘re working on using uses computer vision and looks at each tree and says this tree needs thinning today and this tree doesn‘t need thinning. the end result of that will be better—quality fruit, better yields and potentially less chemicals used in the system.
1:54 am
and eventually, we want to get rid of using chemicals altogether. if we‘re able to go through it and you just pluck a weed out or chip a weed out, or microwave it, for example. there‘s, yeah, there‘s lots of opportunities in that area. sarah wants us to lock in a time... a husband and wife team? that's really normal in agriculture. a lot of people say, "0h, goodness, how is it to work together?" it's what we do. agriculture, you see it all the time. so, it's no — nothing different for us, yeah. i was going to say, is it a pain? laughter. 0h, sometimes. that was nick, loving life on the farm. and that is it from us for this week. as ever, you can keep up with the team throughout the week on social media, find us on: youtube, instagram, facebook and twitter at bbc click. seriously, she‘s already gone and left her virtual self to do the work? honestly.
1:55 am
well, from the real me and the not—so—real her, thanks for watching, and we‘ll see you soon. bye— bye. theme music. hello there. a window of fine weather awaits many today — not all, but for most, there will be some good lengthy spells of sunshine and light winds, and it will feel pleasa ntly warm for late september. the reason, well, we‘ve got a ridge of high pressure moving in behind the cold weather front, which hasn‘t brought much rain but it certainly brought cloudier skies southwards through the night and some drizzly rain, which will stick across eastern england for a time before finally clearing. the ridge of high pressure towards the north ensuring a pretty cold start to the day again, with temperatures close enough to freezing to give us some grass frost. the other thing we need to be
1:56 am
aware of in northern and western areas is some patchy fog — of course, it lingers through the rush hour this time of year. now, we have got the cloud, the low cloud, the drizzly rain with our weather front as well in the south, really rather great start for parts of eastern england. as you can see, with time through the day, it‘s starting to pull away, the skies will brighten. also, just the outside chance of a few showers in north—west scotland, possibly northern ireland. it‘s more likely the cloud will thicken towards tea—time here. with light winds, despite temperatures a little lower than they were on monday, the north will still feel very pleasant, and approaching 18—19 in southern areas, rather where they should be for the time of year. but that ridge of high pressure is a brief ridge of high pressure, because by the night, as you can see, and tomorrow, we have got more rain rushing in. and it‘s a slow moving band of heavy rain. some parts of southern and western scotland could see 40—60 mm before this rain finally clears away. you can see, it continues to fall throughout tuesday night and wednesday. strong, potentially gale force winds with it. ahead of it, some drier weather, and this mild air
1:57 am
at least, but it will be a fairly wet affair wednesday for most parts of the uk. then a brief respite again, potentially on thursday, little ridge of high pressure, still some showers around. now comes the uncertainty. where will this deepening area of low pressure come to lie? potentially southern areas, bringing us increasingly windy and wet weather and lasting into the weekend as well. it certainly looks a lot more unsettled by that stage. so, yes, definitely feeling autumnal, and feeling quite chilly as well after the fact that we‘ve had some sunshine and it‘s felt quite warm. look at the temperatures — really dipping down by the end of the week. as i say, perhaps a little respite on thursday, but some very unsettled weather as you can see in the outlook. as ever, there‘s more on our website.
2:00 am
a very warm welcome to bbc news, with the latest headlines for viewers in the uk and around the world. my name is mike embley. an unwelcome milestone. the number of verbal coronavirus deaths officially past i million. -- official coronavirus deaths. the world health organization and the us announce plans to rollout a rapid and new covid test. it could transform the global response to the pandemic. one, two, three, four, five. into the test. twist it three times. the adhesive is pulled off. and you wait 15 minutes. and that is the test. less tha n less than 24—hour was before the first presidential debate mcdonald from dismisses accusations he has hundreds of millions of dollars in
31 Views
IN COLLECTIONS
BBC News Television Archive Television Archive News Search ServiceUploaded by TV Archive on