tv Click BBC News November 4, 2018 4:30am-5:00am GMT
4:30 am
this is bbc news. the headlines: in the last weekend of campaigning ahead of tuesday's mid—term elections, president trump focused on economic growth and his hard—line stance on immigration. democrat leaders have warned the character of america is at stake. polls suggest the democrats may win control of the house of representatives. iran's supreme leader ayatollah ali khamenei has attacked the united states after its decision to reimpose economic sanctions against iran. on monday the trump administration will reinstate all us sanctions which were removed under the 2015 nuclear deal. leicester city players and officials are flying to thailand to pay their respects at the funeral of the club's owner, who died in a helicopter crash last week. the players will attend the service after winning their first game since the crash. now on bbc news — click in kenya. this week: artificial
4:31 am
intelligence creating jobs. creating trends? and — scoring runs. ai. that's what the future is about, if you believe the hype. computer programmes that learn from past experience, that improve and that sometimes, learn to solve problems in ways that even we hadn't thought of. well, here at microsoft, future decoded events, ai is at the top of the agenda.
4:32 am
these days, there are very real examples that al are starting to be able to do things that were once only the reserve of humans. it is learning to drive, to play games. it has learned to paint. it has learned to understand what we say. each ten year or so we seem to have a breakthrough moment where we take a piece of human ability and defeat it with machine. ‘96 it was chess, go, last year — and we all worry. what that is demonstrating is that our al's are extraordinarily good and superhuman in tasks that we can specify and understand. they can improve and self improve. the challenge is this whole idea of general intelligence or transfer of cross tasks and that proves much more challenging, much more difficult. we think it will take many decades to unfathom that, and the old adage was, you cannot teach a machine to do sentiment programming, but if you have a learning capacity
4:33 am
in the system that allows it to go beyond the performance that was originally given to the system. and it is certainly true that al is already replacing us in particularjobs. we will talk more about that later. but we thought we would start with an interesting phenomenon which is happening in certain parts of the developing world, where ai is actually creating jobs. see, in orderfor artificial intelligence to learn, it needs to have access to loads and loads of data. for example, self driving cars need access to images where all the objects in them are correctly tagged. that work is being done by humans. david lee sent this work, not from california, but where the artificial intelligence journey really starts. this is the kibera slum in nairobi, kenya. more than 1 million people live here. i am 10,000 miles in what feels
4:34 am
like an entire universe away from the lush campuses of silicon valley. how are you? hello! the people i am here to meet are every bit as vital to the next wave of cutting edge tech as anyone you could meet in california. you have your brother living here? yes, my brother, my daughter and my mum. are they all supported by you? yes. they are supported by me. this is brenda, a 26 year old single mother, who has lived in kibera her entire life. how does it feel to be creating the technology that is going to change the future? it feels so good. at least you get to do something unique from others. at least with my work that i am doing, i believe i work for something that is go to help me. not even me in the future, but it will help someone in the future. every workday, brenda travels for around two hours to a building
4:35 am
on the other side of nairobi. she is among a team of around 2,000 people who work in this building for samasource, an organisation that recruits people from the very poorest parts of the world. in some cases, that means those who are earning less than $2 a day. here, they earn around $9 a day and there are importantjob is to give artificial intelligence its intelligence. when artificial intelligence works, it sometimes feels like magic. but really, what it is is data, lots and lots of data. if you want a self driving car to know what a person is, you have to feed it loads of pictures of people. if you want it to know what tree is, it takes millions and millions of pictures of trees. that is what is called "training data", and it is here where that data is created. so, depending on the instructions, we are going to basically tag, or annotate, items of interest. right.
4:36 am
from the street to the vehicles, the buildings, even to the sky. right. how is that? that's good. is that good? not quite right? not quite right. laughter. the item needs to be squarely inside that box. if we zoom in... turns out no pixel can be out of place, or unaccounted for. the sky and the street signs, the pedestrians and the lanes, everything needs tagging. once the work is done, a supervisor will check it is up to scratch. the quickest, sharpest annotators in the team will win prizes, such as shopping vouchers. samasource‘s clients include google, salesforce, ebay, yahoo and many others, working on everything from self driving cars to online shopping. one recent project from microsoft's bing search engine helped it become better at identifying certain types of clothing. the building in many respects feels
4:37 am
like a typical silicon valley campus, complete with subsidised food. 52% of employees here are women, in a country where having a child can typically rule you out from having a career. like a lot of people say, if you have a man in the workplace, he will support his family. if you have a woman in the workplace, she will support her family and the extended family. so you actually have a lot more impact if you can impact women and allow them to work as well. while most of their employees are of course in the developing world, the compa ny‘s headquarters can be found in san francisco's mission district. when i first started this business ten years ago, very smart people in the tech world and in the world of big philanthropy for it was a wonderful idea, but it would never work. lila touts her companies record on quality and security, reasons why tech firms come to them. but of course, there is a very obvious reason why these tasks are outsourced to places
4:38 am
where wages are rock bottom and people are desperate for work. some of your clients are the biggest, richest companies in the world. can they not afford to pay more than $9 a day for this work? we make a guarantee to every single worker at samasource that they are paid a living wage. if we were to pay people substantially more than that in some of the markets we are in, we would throw everything off and it would have a potentially negative impact on the cost of housing, the cost of food, et cetera and the communities in which our workers live and thrive. so, for us, we are on average, increasing our workers household income by over 500%. that impact can be seen right in the heart of kibera, where lessons in basic digital literacy are run by samasource at a location within the slum. the people who take these jobs often do it for only a couple of years, but it is a starting point for much better things.
4:39 am
it has changed my — everything. it has changed my perspective, it has exposed me to see there is hope beyond just living here. i was living in such houses, but right now i am living at a good place. so it has improved your quality of life? it has improved, sure. i think there is an obvious moral obligation that the companies that offer this work make sure the workers are valued and safe and protected. i think we have a moral obligation as well, as the people who enjoy the products they create. because ai isn't going to be seen in a place like this for quite some time, but at least what is happening, is education. from what i have seen, kenya isn'tjust getting ai trainers, they are getting ai experts. hello and welcome to the week in tech.
4:40 am
it was the week that apple nearly sent us to sleep when it announced new super—expensive, great big ipad pro, better spec'd macbook air and mac mini. oh, sorry! uk police in the west midlands grounded its own dji drones when a fault was identified that made them fall out of the sky. dji said it was thoroughly reviewing reports of power issues. and google staff across the world have staged a massive walk out over lenient treatment of sexual misconduct allegations in the company. it's chief executive, sundar pachai has said... norway's postman pat may soon be out of a job, as an autonomous mail delivery robot is tested on streets. the droid can travel around six
4:41 am
kilometres per hour drop off mail to 100 people per day. unfortunately it won't pop the post through your letterbox, instead an app tells the recipient to retrieve an item from its robotic guts. hmmm. and finally, tired of ordering delivery and your food arriving cold? the tundra pie pro, made by toyota and pizza hut, cooks pizza on the go injust seven minutes. so it will be fresher and hotter on arrival, but is an autonomous chef really crustworthy? the power of ai is increasingly being harnessed to create tools to aid accessibility. here is an app which helps users with visual impairment to work out what is around them, so it's basically object recognition. two faces, you can point it that way. there you go.
4:42 am
three faces. you can even programme it to recognise individual people like me. spencer, 2 feet away. two faces. it's also describing me when it says two faces. don't tell anyone. it is still under development but this app does do objects, handwriting and text as well. probably a close—up of a cell phone. ai — maybe it should stand for assistive intelligence. that's what this guy thinks. joseph surosh is the microsoft chief intelligence officer for al and i caught up with him for a refreshingly honest take on the technology. someone once said to me that intelligence is what we use when we don't know what to do and i think that is the best description. my favourite definition of artificial intelligence is it is the opposite of natural stupidity. what is artificial intelligence
4:43 am
and what is just data crunching, just looking at lots of historical data and processing it really quickly and looking smart? these days, when computer scientists adopted learning from data, they are crunching text and translating it, they are crunching all your input and doing speech recognition, they are crunching in majors and recognising things in it so all of that now gives a perception of humanlike capability. but it really still at the end of the day is data crunching, just different kinds of data and in the last 10, 20 years as mobile phones came and cameras became incredibly cheap and everything became digital, this was not the case is all of this became data. the world became data. so now, when you do crunching on that, well then, it's hard to look very smart and intelligence. you've got pretty high standards, i like that.
4:44 am
what would a system have to be able to do for you to say that is actually artificially intelligent? to be able to do really complex reasoning with it, and like humans do, and do so in the face of uncertainty. and do things like one—shot learning, instant learning. i have to see a thing only one time and i know what it is and i know how to react to it and even the smallest of living things seems to know how to react to things they encounter. that is the kind of thing that i want to see. do we have anything like that? no, and i don't think they have even achieved that. we have the computational power to simulate neural systems but we haven't figured out how an ant navigates in the world and really survives in an uncertain environment and reacts to all kinds of stimuli. creatures that live on the planet have evolved over billions of years and not through teaching,
4:45 am
not through ai, they have evolved in a real world. and that has led to intelligence. even the smallest creatures have intelligence. i think when i see something like that starting to emerge, maybe then i will start believing in the spectre of artificial intelligence but until then, it is software and artificial learning. what is yourjob? it is the cto of ai, i know. is one of the trends to recognise the limits of ai and say no, not yet. completely. it's called assistive intelligence. ai is most powerful when it is augmenting humans and what we are enabling, the power of all our devices, whether it be mobile phones, it, the cloud, algorithms, is enabling humans to be far more empowered in area action. and it's actually not that different
4:46 am
from any technology. the mobile phone empowers people. ai is technology which adds to that, layers on top of that in terms of extending human reach. what sort of things should today's children be studying in order to work in the world they will grow into? creative work. the sort of things that only humans can truly do. innovation. really, the history of humanity is about leveraging technology, standing on the shoulders of giants and reaching for something new. that was joseph surosh. and more ai now. as sports go, cricket has been an earlier doctoral technology with an armoury of hi—tech tools for umpiring but now a former indian captain has come up with a technology that could give
4:47 am
fans an insight into what is happening with the cricketer‘s favourite weapon. we sent cricket crazy david reid to mumbai. it is too small for my big head. you know the way you remember you are good at something and it turns out you weren't? this is me discovering that with batting. luckily indian cricket legend anil kumble was on hand to realise how it was done. the most important thing is that speed and how much twist at the time of impact and the quality of the shot itself, how close to this sweet spot here. that speed, is and how close you are to the sweet are now measurable thanks to new artificial intelligence technology, power bat. it is being developed by kumble‘s company, spektakom. it is hidden under the bat. this is a cluster of sensors, you have a bluetooth area,
4:48 am
but also gyro sensor which measures the twist of the bat, the velocity as well. there is also a sense of vibration so you can detect where the ball is hitting in approximation to that all—importa nt sweet spot. the sticker sends those measurements for the speed, the twist, the quality of the shot and they are being combined to calculate the power. it is essentially the energy you get into the shot. the system aims to take fans watching the game up close to what's happening on the pitch. the first use is to enhance fan engagement. everybody talks about timing of the shot. this was powerfully hit. this was sweetly banged. what does that all mean in real—time numbers, in real time data?
4:49 am
the amateur version of power bat communicates directly to a mobile phone via bluetooth but because you can't carry a mobile in professional games, with the pro set—up, all data runs to a device hidden behind the stumps. this is how the professional system works. the data comes from a tag on the back of a bat and then comes to stumpy, the stump bot buried underground. you can see the antenna, this bit will be poking out from the top of the pitch. the data is sent down the cable to the cloud where an algorithm does its work. the pro power bat might end up being used by umpires, especially for broadcasters now using it in the indian professional league, calling out bad on—field umpiring decisions. the mobile app is more about cricket fans and amateur players getting closer to emulating their heroes. for the fan out there who wants to look up to his hero,
4:50 am
what is the benchmark for him to achieve? can you emulate him? can you be as close to power that he can deliver, can you do that? that is something you can do. at the end of the day, in a polite arrangement, anil let me bowl him out. yes, it was fake by the got first ball. it felt great. watch this for a second. see, all you needed was just one ball. that was david reid in mumbai. well, we've all had our embarrassing fashion moments back in the day, apart from me, fortunately. emily bates has been looking how ai may be picking the fashion faux pas of the future. the fashion world moves fast. but some high street stores are changing their collections every
4:51 am
week to stay up—to—date with the latest trends. here in the uk, experts estimate we throw out over 300,000 tons of clothing every year. that's 1.5 billion t—shirts. but the clothing industry is now looking to ai to reduce fashion waste. many companies are now using artificial intelligence to predict what consumers will be buying in the future, the hope being this knowledge will cut down on garment production that won't sell, thereby reducing waste. and of course, increasing profits. but can technology really predict the intricate world of style? fashion pocket from japan certainly hope so. so this is the system where we are collecting all the clothing information from all of the world and by accessing different media, e—commerce sites, which shows quite a few different trends depending on the people, and we already collected more than 25 million from the pieces of pictures so that we can analyse the global trend of the clothing items.
4:52 am
by drawing all this data in, fashion pocket say they can predict fashion trends six months in advance. knowing the future of fashion can reduce waste by a hopefully making sure that the garments that are actually produced our garments that people want to buy, that they are in line the trends, with the feeds that people are looking at, that they are clicking on. clients are basically using this information like a process, based on those trend analyses, our customers, client companies, are coming up with what kind of clothing items are going to put in the next season. the ai system detects the colour,
4:53 am
shape, pattern and size of the clothes and the images and also the overall combination of the outfit. all this data is then analysed to show trends based on region. so this is going up. there wasn't much purple and now purple is on the rise. purple tops now. yes, exactly. especially in japan. people don't wear blue. but they used to. but they are over blue. totally. no more blue. so historically it would have been the brands dictating what the fashion trends up the next season so the designers would be doing all the research and be going on inspiration trips, they would be putting products onto the catwalk and that would be putting their stake in the ground and saying, this is what is going to be
4:54 am
on trend the next season. that's all changed. it's no longer brand—centric, its consumer—ce ntric. we as consumers are setting the trends. will it end up being a self—fulfilling prophecy, limiting the creativity of designers? it will certainly be a designer's time when you can look back to how much time they would have to spend research in a collection, as opposed to now how they can use their time or creatively to experiment with exciting new styles. so you can tell me what i should buy for six months from now in the uk and i will be completely on trend? i have all the information. looking sharp, emily. emily bates in tokyo. that is it this week. don't forget, we are all the social media. we live on facebook and twitter at bbc click and instagram at bbc click. and don't forget our youtube channel.
4:55 am
youtube.com/...guess what? bbc click. we will see you soon. hello there. the weekend so far has brought a real mix of weather. dry for some, wet for others. for a few, very wet indeed. this is the satellite picture. look at this beautiful swirl of cloud. we've got a deep area of low pressure just here, drifting to the north—west of the british isles. this band of cloud, a frontal system which has brought quite a lot of rain in some places during saturday, and is still sitting in place for the first part of sunday. the rain fizzling away, at least for a time. some cloud and patchy of rain across lincolnshire, the midlands, east wales as we start the day. sunny spells throughout the day, parts of the south—east getting away with a completely dry weekend.
4:56 am
and further north, sunshine and a much drier, brighter day across scotland on sunday. lots of sunshine to be had. the odd shower in the north—west. black wind arrows indicate wind gusts in excess of a0 miles an hour. the wind still quite gusty across the north—west. 13 in glasgow, 12 degrees in belfast, a nice looking afternoon for northern ireland. a bit more cloud across england and wales, with the odd spot of rain. east anglia and the south—east escaping dry and fine. this rain pushing into the far south—west of england, likely to turn heavier as we get into the first part of the evening, drifting into west wales. maybe getting a bit further east. i think for most places, if you're off to a fireworks display on sunday evening, it should be dry and relatively mild as well. that sort of theme continues for bonfire night itself. let's pick up the story weatherwise as we go through late sunday evening into the early hours of monday. rain fringing it's way across western parts of the uk. further east, largely dry.
4:57 am
the odd patch of mist around, and those temperatures not dropping fast, mild air continuing to come in from the south. 7—10 degrees. this is the pressure pattern on monday morning. another area of low pressure, a new one, developing to the south—west. this front just grazing its way across western parts of the british isles. parts of south—west england, west wales, particularly northern ireland and western scotland. we could see a bit of rain through the day, that rain coming and going, not raining all day long. further east it will be dry with spells of sunshine and very mild indeed, 15 or 17 degrees. really, it's more of the same on tuesday. bands of rain scraping their way into western parts of the uk. further east, largely dry with some spells of sunshine and, wherever you are, once again, it will be mild. mild is the big theme to take us through the week ahead. wind and rain at times, but not all the time. this is bbc world news i'm duncan golestani.
4:58 am
our top stories: the rhetoric heats up ahead of next week's midterm elections. president trump says the american economy is the best its ever been. a blistering attack on donald trump from iran's supreme leader, as the us prepares to reimpose sanctions on the country. eight children have been injured after an inflatable slide collapsed at a fairground in the uk. and a minute's silence as leicester city play their first game since the death of the club's owner last saturday.
67 Views
IN COLLECTIONS
BBC News Television Archive Television Archive News Search ServiceUploaded by TV Archive on