Skip to main content

tv   Click  BBC News  May 6, 2018 4:30am-4:59am BST

4:30 am
the american nra gun lobby that the 2015 attacks in paris could've been stopped if civilians had been armed. former president hollande called it ‘shameful‘. mr trump also used the level of knife crime in london to defend us gun laws. the most successful manager in british football, sir alex ferguson, has undergone emergency surgery after suffering a brain haemorrhage. his former club manchester united says the surgery went very well, but that the 76—year—old now needs a period of intensive ca re to recover. almost 1,600 people have been arrested across russia, including opposition leader alexei navalny, during protests against president putin. police used teargas to disperse protesters, some chanting "down with the tsar." the rallies come just two days before vladimir putin is inaugurated for his fourth term in office. now on bbc news, we look at the digital world in click. this week, the data that might keep the streets safe and keep
4:31 am
the high street in business. plus, some beautiful music that won't scare the sheep. we're becoming aware of how much data we give away without even realising it. our habits online can say a lot about our personalities. but when we're out and about, what does our behaviour in the real world say about us? well, in the uk, we're all getting used to the fact that we're being filmed by cctv a lot the time. but although a human can tell a lot about a person just by looking at video footage, that is a really hard job for a computer to do. that said, this system is having a pretty good guess at who it's looking at right now. it's a very flattering guess, actually, but probably no worse than a human would guess given the same footage. this system is an artificial intelligence that researchers at southampton university have trained to estimate the gender, age and a description of everyone that falls under the gaze of their test cameras.
4:32 am
we have millions of images, different types of people captured from different environments, from all around the world, and we hand—label these. they're labelled by human beings as male and female, or 16—30, and the magic part is that we can feed these into a machine, into a computer, to learn what it means to look male, what it means to look female, just from the visual cues alone. the team are pitching this at retailers, as an improved way of measuring footfall — instead ofjust counting the number of customers who come into a store, this can tell shopowners whether the right kind of people — the shop's target audience — are being drawn in. rather than know that 100 people came in and they sold 50 things, we can tell them that 50 people of their target demographic came in and they sold 50 things.
4:33 am
so now they know how well they are converting on certain types of people, who they may be targeting to sell products to. or maybe tell them that their sales are down even though people came into the shop, they're not selling as well because the right types of people aren't coming into the shop. retailers would get a breakdown through each day of the most popular spots in store and the most popular routes taken through the store by each type of customer. and with this high street view here, you can see which shop windows are more grabby and how long people dwell in different areas. in this example we can see carphone warehouse actually has a higher proportion of females walking past it than gap, on the other side of the road, which might indicate that maybe gap should move their placement along the high street as this side of the street sees more of its target demographic. this kind of profiling of humans by computer systems has many uses, from this kind of anonymised retail analysis, to other areas that might say more about us as people. one of the more controversial uses for al is in policing. marc cieslak travelled to durham to find out more. peterlee police station in county durham.
4:34 am
the early hours of the morning. the man pictured here in the station's cctv — let's call him steve — has been arrested for possession of heroin. i'm arresting you on suspicion of possession of a controlled substance. here in durham the police are trialing ai software which could help make decisions about suspects being held in custody. leave your shoes there, and we'lljust do a quick search. after steve's been processed, the custody sergeant will enter his details into a system called the harm assessment risk tool, or hart for short. it is an artificial intelligence tool designed to help custody sergeants make decisions about what to do with a suspect. what it does, algorithmically, it uses all of our data to tell us who's high risk of reoffending, medium risk of reoffending, and low risk of reoffending. it's not absolutely perfect, but it gives us a really clear indication of who might commit crime in the future. the ai is trying to identify repeat offenders who, rather than being sent to court, will be entered into
4:35 am
a rehabilitation scheme called checkpoint. we all recognise that when we place offenders into the criminaljustice system there is a revolving door of criminaljustice, so we wanted to look at, is there a better way, is there a different way, that can potentially produce more positive outcomes? this is the first time an ai system like this has been used in the uk. so what's the process then? i put my name in here... the system is now searching for the prisoner called meecham. the sergeant will click on the list. the system has decided i am low risk, as a detainee. that is the screen sergeant be presented with. it says, "low—risk — the subject is not likely to commit an offence in the next month." what happens next? it can go either way, really. so if the sergeant has decided this person will be remanded in custody, they will be charged, refused bail and go to next available court, generally the next morning.
4:36 am
it's a different story for steve. he has a long list of previous convictions — including drug possession, assault and driving offences. do you want a solicitor? nah. the ai assesses the data that durham constabulary holds on him and makes its recommendation. he can only enter checkpoint if it rates him as a moderate risk. a final decision around steve being sent to court or entered into checkpoint rests with the custody sergeant. we will continue to use our nous, our nose and our gut like we always have done, but what the hart does on top of that is, we have looked at ten years of data and that allows us to target the people, in terms of our interventions we have developed, to make sure they don't commit crime in the future. everyone i've spoken to who works with the hart project is very keen to stress that the information it provides is advisory only and that a human custody sergeant makes all final decisions. but the use of this technology in general is, for some people, a cause for concern.
4:37 am
this kind of artificial intelligence system in policing relies on big data. and that means people's privacy is at risk, and it risks us moving more towards a surveillance state. but it also risks discrimination, because patterns that exist in data already risk being perpetuated and repeated. and there's very little accountability over it. cambridge university's criminology lab is the birthplace of the technology being used by the police in durham. it's been in development for five years. the decisions that hart makes are based on historical data. it uses this information to predict future outcomes. our artificial intelligence, regardless of whether it is the criminaljustice system or somewhere else, is looking to the past and the patterns we've observed, to predict the future. we know that the past was not perfect — human beings
4:38 am
were making decisions. by the way, humans were making predictions at every stage of the criminaljustice system, from the first time we had thief—ta kers and the bow street runners. the system makes use of 3a different predictors, which include things like age and gender. but it also uses data from the credit reference agency experian, and it's this data that privacy campaigners find troubling. experian‘s data profiles individuals based on their postcode their household and even down to an individual level. and some of these profile stereotypes include terms like "disconnected youth", "asian heritage", "dependent greys". and people are sorted into these categories. that's bad enough, but then that kind of profiling is being used to make decisions about freedom and justice in the uk. i think we really need to put the brakes on and say, why are we doing this? it turned out that experian variable was highly important, if we removed that variable from the model we would have lost a lot of accuracy.
4:39 am
i will say that we are now about to install the next generation model which will not use this code. it's just a contractual reason. the durham constabulary's contract with experian has come to an end. back in peterlee‘s police station, steve has been assessed and accepted into the checkpoint programme. he ended up spending around eight hours in police custody. i will grab my stuff, come round the front and we'll give you a lift. yeah, we'll go out the front. what is the future of this kind of technology? in terms of how i see it developing, i think over the next five years there will be a proliferation of these kind of tools, i think, going forward, society will come to accept them more, but human beings, government, society, needs to stamp its foot a little bit about it. and the reason i say
4:40 am
that is because we need to have more regulation. it's taken me about a decade working in this field to realise that the real question is, what i always called the "now what" question. i can build you a model that will predict whatever outcome you're interested in, criminal justice—wise, but what areyou going to do with it? once that box comes back and you have a red box on your computer screen that says "high risk" — now what? hello and welcome to the week in tech. it was the week in which the world's tallest tower in dubai created a virtualjump off the top — should you choose to accept it. iranjoined russia and blocked telegram and, with the harvest over, cambridge analytica shut down. ford revealed it's planning to let the blind see through car windows. this gadget converts the window to greyscale, the window then vibrates differently depending on how light or dark the scene is where you touch it.
4:41 am
rather usefully, an ai assistant also tells the passenger what they're looking at. the boss of whatsapp is off — jan koum's not happy that owners facebook are trying to commercialise users‘ data. lg had its biggest phone launch of the year, the g7 thinq sports an ai camera with 19 shooting modes. don't worry, the ai picks the one you're most likely to need. and finally, take a look at what could be the most beautiful epic fail ever. this is the stunning display created by chinese firm ehang, who broke the guinness world record for the most drones flown simultaneously. the problem? well, the aircraft failed to spell out the date and the record—setting number of drones perfectly. an "epic fail" according to the country's own south china morning post.
4:42 am
tough crowd! so, we've been talking about the power of data and probably no—one has more personal data on you and me than you—know—who. yes, at facebook‘s f8 conference this week, they were telling us their plans for the platform for the year ahead. there will be new face filters — hmm, i wonder where they got that idea from(!) a new vr headset, and a dog on a laptop. thanks, jiffpom! nice work! but there wasn't just a dog in the conference centre. there was one very big elephant in the room as well... welcome to f8! this has been an intense year! privacy. it has been centre—stage for facebook so far this year
4:43 am
with zuck appearing in person before congress to defend facebook‘s privacy practices. and this week, the company sharing that scandal spotlight with facebook, cambridge analyticia, announced that it was shutting down, saying it had been vilified by unfounded accusations. nevertheless, facebook‘s feet are still very much to the fire as questions heat up around how our data is used. and what i've learned this year is that we need to take a broader view of our responsibility. facebook‘s opening salvo, a new feature called clear history. we're starting with something that a lot of people have asked about recently, and that's the information that we get from websites and apps who are using facebook‘s advertising and analytics tools. the new tool, which will not be available for at least a few months, is an attempt to answer some of the harder questions put to him by congress. facebook already offers this tool which lets you download the data that they hold on you.
4:44 am
the problem is, it's not all the data that they hold on you. via facebook like buttons and also using hidden facebook trackers installed by the owners of many websites, facebook can follow you as you browse much of the web, and it's accessing this data which is uniting privacy advocates and congress alike. "clear your history" is bit misleading. all they are doing is actually disconnecting the history, the browsing history, from the profile, from the account itself. the data will still exist. the information that you went to website one, then website two, then website three will still exist and will still be stored in facebook‘s data bases. the only thing is that it will not be linked to you directly but it will be linked to some other identifier and possibly alongside that other identifier will be kept other data about you. and while facebook might not be so lucky in privacy, maybe they could be luckier in love?
4:45 am
we are announcing a new set of features coming soon around dating! cheering and applause. now, this is gonna be — this is gonna be for building real long—term relationships, all right? notjust hook—ups. zuckerberg promised to bear these privacy issues in mind with the new service but would you like the idea of facebook getting even more intimate with you? a few years back, wearables became the buzzword. but this year, there's been a lot of talk about hearables. now, that could mean a personal training session or real—time translation directly into a pair of earbuds. but this week, i'm putting to the test a few devices that aim to give you a hearing test and, as a result, optimise what you're listening to.
4:46 am
even for those of us who are unaware of having any defect, the chances are our lifestyles will have had some impact on our hearing and due to this, we don't all hear things in exactly the same way. we have a huge range of hearing, from 20 hertz to 20,000 hertz, and there are various causes of different type of hearing impairment, so we could have a hearing loss through ageing, it could be through noise—induced hearing element as well. but can some headphones balance this out, and do we actually need them to? i popped into one of our peaceful radio studios to find out. so first of all, i have these audera headphones. now, i need to do a hearing test to create my profile. what we're doing here is finding the quietest part of my audible range throughout all the different frequencies. so, let's get started. that's it. it takes about ten minutes to go through this whole process for a full range of sound on both ears.
4:47 am
ok, so the results of my test, i can tell that each ear seems to hear differently from each other but beyond that, obviously, it doesn't mean a huge amount. but what i can do now is i can put the headphones to the test. ok, so it says here the first time that you experience this personalised sound, it may be very different from anything you have ever heard before. so it's — oh, it also says that it may be "overwhelming" for some people. shouts: ok, sorry if i'm shouting! this is at 0% now! let's apply 50%! wow, it sounds a lot cleaner! the sounds a lot sharper! you can hear each element of the sound really clearly! the clarity at 100% is pretty good.
4:48 am
ok, so in optimising the sound for me, it seemed clearer but obviously, my test is pretty subjective. so what do the experts make of the concept? with the auderas, it uses a basic form of pure tone audiometry. this is something that we perform every day in the clinic and it involves presenting different pitches and different tones and then measuring how loud we have to make it in order for someone to be able to hear it. auderas are stating that they take that into account with the settings on the headphones, and then incorporate that, giving extra sound in the areas to compensate for any hearing loss. the mimi music app works in a similar way and has been around a few years. after creating your earprint, it aims to optimise the music you listen to through any wired headphones. just inputting your age does the job to some extent, although i ended up more impressed by the way you could tweak the sound around you. here i have the nurophone headphones which look pretty interesting for a start as they're sort of in—ear and over—ear. but they work differently.
4:49 am
instead of asking you whether you can hear something, they do the test for you, and they do it in just 60 seconds. they claim to do so by sending sound waves into your ear which will trigger a reaction, sending electrical impulses to your brain and sound waves back out your ear. it's through measuring them that the app will create what it calls your hearing signature. in my not—very—scientific experiment here, i am noticing a much bigger difference between the generic and the personalised in these headphones. the music sounds totally different. it's different parts of the music that sound louder. and that's mixed in with just a real clear, crisp sound. but our expert had some questions over this upgrade of sound quality. they do give you the opportunity with the headphones to be able
4:50 am
to compare the generic settings with those taking into account your personal profile. there was a hugely marked difference between those two things when i actually tested them out on my hearing, which actually
4:51 am
possibilities for our citizens.
4:52 am
now when we have more and more tourists coming, they are also quite interested in our language somehow, but we were not present on google translate this time. so we thought that we — well, we should give google another go and ask them if they could put the fa roese language on google translate. they established faroe islands translate, an online service in which requests were sent to faroese volunteers, who recorded the translations on video and posted them on a dedicated website. as well as a useful tool to promote the islands, they also hoped it would again add pressure to google. try it, and you will learn that we have a0 words for fog. "toka". "skaltova". as yet, google hasn't bowed
4:53 am
to pressure a second time. we don't have the budget the uk have, for example, so we need to find something that works for us and this works very well. it gives us the attention we wanted but most of all, it shows our culture and our people. but why is such a remote part of the world so keen to stay connected? a part of it is to tell all the world that we're here and we have a language and we are a country and you should come visiting us, or you should go and google us on the computer because we are also here. hello! chuckles. that was paul in the faroe islands, and that's it for this week. don't forget, we live on facebook and on twitter, @bbcclick. you can find loads more from us there every single day. thanks for watching and we'll see you soon. hello there.
4:54 am
saturday was a glorious day for much of the country. we saw some fine sunsets up and down the uk, like this one here in wakefield in west yorkshire. not all areas, though, were sunny and warm. coastal parts along the irish sea were disappointing, around cornwall and devon and into western wales, south—western parts of scotland as well. this weather front has been plaguing the north—west corner of scotland, bringing strong winds and outbreaks of rain. as we start sunday morning, there should rarely be any rain here, just a bit of cloud. elsewhere it is a clear start, quite chilly across eastern parts of england. sunday promises to be another glorious day for much of the country once again. sunshine from the word go. a bit of cloud across scotland. i think we are looking at a better day through the central belt on sunday afternoon. top temperatures reaching 2a or 25 celsius, so another warm day on the cards.
4:55 am
even warmer air is imported off the continent on bank holiday monday. that will nudge temperatures into the upper 20s celsius, potentially being a record—breaking may bank holiday. we could still see a little bit of low cloud and some murk in the irish sea. maybe the odd shower across the northern half of scotland. that said, most places will be dry and sunny. a bit cool across the north—east of scotland. that cool air moving up towards aberdeen. this is likely to be the high temperature on bank holiday monday. it may very well break the last bank holiday record, achieved in 1999, that was 2a degrees in hampshire. tuesday, we start to see some subtle changes. this area of low pressure will bring cooler air into the western side of the uk. a weather front which will be very weak, bringing outbreaks of rain to northern ireland and western scotland. the eastern side of england, another warm day on the cards. we could be looking at 27, maybe one or two places seeing 28. a cooler and fresher feeling further north and west. next area of low pressure moves in for wednesday. this one will have more to it,
4:56 am
stronger winds, that will make things very wet in western ireland and northern scotland. still some sunshine across southern and eastern parts of england. a relatively fine day here. temperatures typically 20 degrees. but a much cooler and fresher feel out west. hello and welcome to bbc news. i'm nkem ifejika. us president donald trump has
4:57 am
caused outrage in france by suggesting the 2015 attacks on paris could have been stopped by giving people guns. he mimicked gunmen summoning and shooting victims one by one, using his hand to imitate a gun being fired. the former french prime minister manuel valls called the comments ‘indecent.’ andrew plant reports. usa! it has become an annual event, president trump addressing america's national rifle association.
4:58 am
4:59 am
5:00 am
5:01 am
5:02 am
5:03 am
5:04 am
5:05 am

32 Views

info Stream Only

Uploaded by TV Archive on