Skip to main content

tv   The World According To AI P2  Al Jazeera  June 18, 2019 3:00pm-4:01pm +03

3:00 pm
for the. but i'm a clock and to hell with the headlines here on and egypt's only democratically elected president mohamed morsi has been buried at a summit dream car the day after he died in court security forces kept journalists away from the burial service egyptian state t.v. says it was he had halted sarg it being imprisoned since 2013 when he was toppled by the military well most surround turkey holding a funeral pros for mohammed morsi seen in kosovo this morning from istanbul. where if i think most of them istanbul and very symbolic place for gatherings and process especially with regard to the incidents and i'm just events in the middle east in muslim communities there is an absentee funeral prayer for the ousted president
3:01 pm
mohamed morsi and a call was made by turkey some religious by the religion affairs director great to have seen 2 funerals that's doing the mid say across 13 almost the smart people have gathered here. and there are dissidents and journalists so even more have been residing in turkey since their offspring began in early 2010 and people are here to give us support for mohammed morsi is because their prisons are gone is expected to attend the 2nd absent to feel that it's going to be held in the same place here in the later hours around 14 she today along with the n.g.o.s and other political dissidents of the egyptian egypt present time has been very vocal about mohamed morsi and the process that he has been through being in jail the executions ongoing in egypt he also last night criticize the west committees for turning
3:02 pm
a blind eye to the executions in egypt and he said about mohamed morsi who has been elected by the will of egyptian people is a march for democracy and what has been ongoing in egypt is answer to my question. yemen to the rebels say they've launched another drone attack on a port in the south of saudi arabia it is the 6th time the airport has been targeted since wednesday who things have warned they will continue attacking saudi airports in retaliation for the kingdom's actions inside yemen. russia's foreign ministry has warned the united states against deploying more troops in the middle east calling the move provocative washington is planning to send around a 1000 additional troops to the region amid rising tension with iran before making that announcement the pentagon released pictures that it says shows a leads iranian troops removing an unexploded mine from one of its 2 oil tankers hit by explosions last week. to carry has offered what she calls her most sincere
3:03 pm
apology for the way tradition bill has been handled she's under pressure from protesters who've organized large demonstrations opposed to allow suspects be extradited to mainland china says she heard the protesters demands loud and clear it is refusing to resign. pro-democracy members of hong kong's legislative council say that car alarms apology came to tittle too late they reiterated calls for her to step down saying that hong kong people no longer trust her every lamb is actually presenting to the rest of the world that the administration it's going to be a lame duck and ministration because she would have a very hot tiny governing hong-kong from now on she acknowledged that it's a war question of trust and. confidence in her administration which has become completely non existent a court in kenya is set to deliver
3:04 pm
a verdict for suspects charged in the 2015 attack can go to university for men face life in prison for the killing of 148 people most of them students some of the survivors say the attackers initially killed indiscriminately but later released muslims jews. gunman in democratic republic of congo have killed a chinese national in the colbys soldier in an ambush on monday increased violence has led to the displacement of hundreds of thousands of people from the north east and it's very province fighting between tribal groups has increased over the past week killing nearly 200. or outfit with headlines here on a desert we've got more news coming up right off the big picture i feel.
3:05 pm
the world is slowly waking up to the growing impact of ai. this exhibition in london is all about our relationship to computer technology but ai remains in its infancy still struggling through glitches in the system yet we're handing more and more decisionmaking over to the algorithm. i'm corey kreiter i'm a human rights lawyer i investigated a drone strike in yemen that killed my client faisal banally jabbers family what i found was a semi-automated targeting system used to describe who was and wasn't a threat but as the big picture showed previously on the world according to ai artificial intelligence isn't just used in military targeting but in everyday policing as well and this episode will delve deeper into the power potential and prejudice of ai as people seek to use it to shape our world.
3:06 pm
eat. we're. one of the best established blind spots where the police have been seeking to use algorithmic material to accelerate and inform their practice is predictive policing so the way that it works is this let's say that you've got a historically over policed community a community who are poor a community of color algorithm takes that data and basically turns around and says you went there before you probably want to go there again and so what everybody found and study after study has shown this is that it amplifies the next celebrate the process of over policing communities who are already over policed.
3:07 pm
in the united states the history of law enforcement has an unfortunate link to racism going right back to the era of slavery. in the 18th century groups of white men on horseback paddy rollers as they were known would go out on patrol looking for in slave people who'd fled to freedom. if caught those who'd run away would be horrifically beaten mutilated or killed. the paddy rollers would be assigned an area to patrol. it was known as their beat. there is no question that policing in the us contacts certainly has grown out of slave catching so let's go back to their original origins people who tried to escape and being enslaved were often hunted down and brought back into some type of system of control. when slavery was abolished
3:08 pm
the patty rollers would evolve into what became formalized police units and they didn't just abandon the role beats. just as before african-americans continue to bear the brunt of repressive police attention. with every crime recorded every stop and search every arrest every sentence a criminal history has been recorded data has been generated data that in the 21st century has been used to make predictions about the future of crime. we're now in the era of predictive policing. the got its start in los angeles with the l.a.p.d. around 201-2011 was actually the brainchild of an anthropologist and a mathematician brantingham and george miller were 2. of social scientists who are
3:09 pm
studying crime patterns and they realize that certain kinds of crimes are sort of viral are most contagious in their way so if there's one burglary in an area it's actually statistically more likely that they'll be other burglaries in that area and so they came up with an algorithm and they pitched this the l.a.p.d. and the l.a.p.d. would went with their 1st experiments to see if it worked and for them it worked enough that it began and it suddenly took off. the research that went into geoffrey browning and algorithm was for the most part funded by the pentagon. his initial work used military data to forecast insurgent attacks in places the u.s. had invaded such as iraq and afghanistan. but his focus shifted from battlefields abroad to supposedly enemies at home. he created a company called credible that uses data from us police departments to tell us police departments where to look for criminality. what happens is that when.
3:10 pm
vendors show these tools to the police the police say that's exactly what we were going to do. and sometimes we describe this as selection bias meets confirmation bias and they live happily ever after the idea that we use historical data to train machines and algorithm to predict into the future is very important so for example if you live in a community that has historically been over policed you are more likely to have your algorithms pointed right back to those communities that have been over policed if you have people who have been over arrested like african-americans and latinos and degenerates people then your ai is going to point to those people once you feed that data and say ok i see there's more crime here let me go find
3:11 pm
some more. well guess what you're going to find if you go look for more crime you are going to find more crime where you look for it and not where you don't look for it. but then they added in this extra element that every time there is a police contact you get an extra point. so in some ways they've created an almost self-fulfilling prophecy for the people because police are directed to go find the people with the most points then they go find the person who has most points and then the person gets extra points so tomorrow where are they going to go this guy's extra point the consequence of that is that the database starts accumulating a lot higher fraction of all the crimes committed by black people than committed by white people so you turn algorithm loose on that it'll say wow black people are really dangerous but we can ignore white people. so what happens is that. the algorithm. calcifies or embodies the bias in the police.
3:12 pm
by there for a really good summation and really going to be there literally. yeah let's go. on is one of the lead organizers of the stop l.a.p.d. spying coalition a collective that campaigns against what it believes to be a growing police surveillance and criminalization of the local community in 2018 the coalition took the los angeles police department to court forcing it to release the details of its predictive policing program there's 2 layers to predictive policing one is a community and a location based where algorithms are used and the company principal has developed that algorithm which was owned by jeffrey brown think it was a professor of anthropology and has a long history himself how this thing was created on the on the feeds of
3:13 pm
afghanistan and iraq directly coming from the border from the war zones and the other piece is operational laser which is a person in a location base for the police in program lou. stands for los angeles strategic extraction and restoration program and the reason why it's called laser is that the creators of lasers said that we wanted to go into the community with a medical type precision and extract tumors out of the community like lead from a decision that's where they came up with the act and that's how they came of the acronym as people are tumors the exact fact i think is not actually. what credible and laser claim to offer is a one stop crime prediction shop the pitch is to tell police not just where crime will occur but also who might commit crimes in the future. the l.a.p.d. was using these technologies to decide where to deploy their police patrols.
3:14 pm
focusing resources on so-called crime hotspots like by these. so this is all the hot spots for a particular time period hot spots are created by the algorithm the prep for longer where they use the information long term crime history or short term crime history and then they create these 505500 square foot hotspots on what basis how are they deciding this so will the put it very bluntly there's a lot of pseudo science and now it's being presented as these computers are really snootful and they would predict when crime may happen. but predictive policing doesn't just flag up a place with laser it also sticks to a person the l.a.p.d. maintain something called a chronic offenders bulletin these bulletins are undisclosed reports on so-called
3:15 pm
persons of interest people the police believe to be likely to break the law. this risk is calculated using a points based formula based on data from police records field interviews and arrest reports this is pulled together. and scored by algorithmic software created by the defense contractor pal and here a company with close ties to the u.s. military. so how do you get yourself on to the laser systems are these other things that identified these risks so if you're stopped a field interview cards filled one point so if you know there is it or if the police stop stop you've got a point one point immediately you get a point and this individual was stopped the same day 3 times so they treat him for 3 points right there if there had been a previous arrest with a gun 54 inch if you had any violent crime 500 parole and probation is 54 and and identified as again as gang affiliated 5 point. when it comes to the chronic offenders bulletin points can mean prison but it's not just about locking people up
3:16 pm
for hemant con the data suggests increased police attention at the borders of a historically deprived area called skid row which helps keep the poor contained from the more affluent neighborhoods nearby. so this is like a beachhead so think of the defense of financial district yet from poor people you know when we talk about hot spots you know you will see the dirty divide how the proximity of extreme wealth and extreme poverty co-exist right here about 2 blocks from. absolutely. he was gone oh man doing good to me and cory general. secretary of. the building right there. i meet steve richardson who goes by his street name general doe gun he's a former prisoner and skid row resident who now works with the coalition
3:17 pm
campaigning for greater protection for the local community. so you guys have been doing work on this predictive policing stuff right is what does that look like out here on the street to people who live here so predictive policing roles of a lot of weight because i mean skid row is ground 00 all experiments that have been you know so these poor folks of course so all a little programs l.a.p.d. spy programs everything that they come out with is 1st tested right here they say 1st last to say for seriousness right here was grow 110 x. disk a rope maker the most told this community not only in america but 2nd in the world to baghdad got all kinds of patrols on skid row so we got the cops on motorcycles we got regular cars we got. we got detail cops there's all polies in like a 15 block area what are cops on horses right smack in the middle of the house you know they have no things that continue to come out 5 here from
3:18 pm
a lot. about 80 percent of people here are black right about any sins people suffer from something. a disability maybe physical you know several minutes old and all of us is full of. the most the rest of parson on skid row was a woman a and moody sousa rested under 808084 violating 4118 right 4180 days is up when this bill called to say you can sit sleep a lie on public sidewalk so our only crime was she was homeless and have anywhere to go and was forced to sleep in public space she got arrested $118.00 times for being 108. for just for just being in public space and it was all over you know based on a lot of predictive stuff like that and the point you you joe gone and you haven't
3:19 pm
are making is that this is a practice that goes way back right over policing in this community goes back decades and then that information from that then gets fed into the computer and the computer turns around and says we'll go back and do some more of the same thing right now and the computer before the information gets in the algorithm is designed for policing so the algorithm would create outcomes that an agency wants to achieve and this is really the key point and the outcome that the agency wants to achieve in this community is cleansing and damage when. we walk further along the tense begin to thin out as to the local residents gathered on the sidewalk we're it's obvious we're approaching the outer limits of skid row the hotspot boundary hammond had pointed out earlier.
3:20 pm
this is like if on the storm a host of hot spots that a person from skid row would be walking into and this is where you will have more policing waving for people who are for him than waiting for people to give them tickets waiting for people to throw them against the wall right. for people to intimidate and harass and demand to believe the neighborhood. but few weeks after we left skid row the l.a.p.d. announced that it was canceling the laser program the pushback worked police admitted the data was inconsistent but the l.a.p.d. says the predictive policing tool pred pull is still in operation. so let's think about the incentive structures with some of the predictive policing tools that we've been talking about what does it say about the incentives and the problems we're going to have with these tools that you've got counterinsurgency software then essentially used for law enforcement purposes i hate to have such a sinister. interpretation but i think it's about opening up new markets to sell
3:21 pm
this software to 0 and law enforcement in the last is you know been a great market for lots of military technology is quite frankly i think there's actually the opposite incentive to get it they have the incentive to get it all predictive policing software has an incentive to make the sale with police so their incentive is to is to make predictions that are as close as possible to what the police already believe is correct so given that it's really hard to know. if ai has been tried on representative data or or not if we have real reason to suspect for example that there might be bias than isn't there a question about whether the system should be used at all well i think that's the fundamental issue is that we're seeing the deployment of all kinds of automated decisionmaking systems or ai over we want to kind of characterize it and we don't know the effects until after the fact after the damage has been done is primarily
3:22 pm
how we're learning quite frankly about what doesn't work and i think it goes far beyond bias i mean we're talking about aggregating data about us building data profiles that for close certain types of opportunities to us and what's more dangerous i think in the digital age about this is that you know in the 1950 s. if you tried to get a mortgage you were you were black and try to get a mortgage at a bank and you were discriminated against you were very clear about what was happening that discrimination was not opaque and when it moves into a software modeling system what instead you have is a banker who's like you know i'm sorry dr noble you just can't have it and i don't really know why and so that lack of transparency is one of the things that i think we're kind of trying to contend with here and this just becomes a wholly normalized process we don't understand or with you know the the models for
3:23 pm
actuarial science for determining whether you're going to pay more insurance for example because you live in a particular zip code doesn't even account for these histories of racial segregation housing covenants real estate covenants so just because we look at the zip code that doesn't tell us about this long history of discrimination that has sequestered people into particulars of codes those are the kinds of things that i feel like over time become harder and harder to see i think one of the things that i find. worrisome is that we talk about data being collected for these kinds of systems and for the most part they just collect of some completely different purpose it just happens to be there in policing data is created by the police doing what they do they're driving around they're stopping people there occasionally arresting people and so forth that data gets produced and then is used in a predictive policing model it's not collected for the print predictive policing model that's a 2nd order effect that's used because the data is already there and it turns out
3:24 pm
that it is a terrible way to predict where future crime will be because what police do is not collect a random sample of all crime they collect the data they can see this is true in most of the places where people are applying i think it is useful to detect where bias is happening and simulation can be important that i think that's true however it doesn't necessarily allow people to have again this conversation that i have been discriminated against it's just sort of leaving the expert analysis to make that discovery when in fact there are a whole bunch of people that wanted to be homeowners or you know wanted to move house and they don't really understand why these decisions are happening so as a data scientist what's your take on this how do we build a kind of test for when it's appropriate at all to use machine learning and when it's not the question should be who bears the cost when
3:25 pm
a system is wrong so if we unpack a particular system and we say ok we're building a machine learning system to serve ads and the ad that we're serving oh this customer's searching for sneakers but we served or boots at. oh dear we were wrong there no one cares that's a meaningless meaningless problem the consumer could care less we get along ads all the time we're trained to ignore them let's compare that to a system which makes prediction about whether or not someone should get credit. in a credit based system if we're wrong the consumer who should have gotten credit doesn't get it or the consumer who should not have gotten credit does get it in both cases and in particular in the case where someone who should have gotten credit does not get it that consumer bears the cost of the air she doesn't get whatever it was that she needed the credit for to buy a house or car or something else the company that failed to offer the loan may bear a small cost but there are a lot of customers so they don't really bear much of
3:26 pm
a cost and so when the customer bears the harm. we can predict that the harms will be greater because the people deploying a systems a little incentive to get it right. we know that if people of color are over police report people are over policed and over arrested they are also likely to be over sentenced. machine learning isn't just used to predict crime it's also used to decide whether a person should be given bail or how long a sentence a prisoner serves. criminal courts in the state of florida and use a predictive sentencing program called the correctional offender management profiling for alternative sanctions compass. in 2016 journalists at the us news outlet pro publica investigated compass and discovered an apparent racial
3:27 pm
bias at the heart of its algorithm. investigative report and one of the things that they found in their hand a review of all of their records was that african-americans were twice as likely to be predicted to commit future crime i found it incredibly interesting for example the story of died one of the reporters told that there was a black woman a young black woman who had taken a bike in one of her neighbors front yards and kind of ridden it around and the person here on the bike said bring that bike back and so she did it but a neighbor called the police on her and she spent 10 days in show and the compass software gave her a score of 8 out of 10 that she was likely to commit a crime again and that and they looked. white man who had
3:28 pm
a history of violent crime of history of being in a. now out of jail and the software gave him 3 years so he was more likely to be replaced but. once again the bias in society was revealing itself in the machine. let me take it. like no place on. wild west the rain. and the strongest false steps down most skilled. nation the concert that in stadiums that are robbed in quest. discover our newest destination defeat the women's world cup france 29 t.
3:29 pm
. in a war torn city in iraq a magic documents the stories of the survivors recording bare hopes and dreams for a peaceful future after american troops withdrawal. but the conflict is far from over. he turns the camera on himself when i saw take control and his family off forced to flee no where to hide a witness documentary on al-jazeera. i've always been fascinated by space but the story of the space race isn't just about the men who risked their lives to travel into the unknown but the ones who held those lives in their. grandfather and his colleagues worked on the space search they designed the spaceships apollo 11 was his triumph. and the perfectly designed space suits for his legacy putting man on the moon on al-jazeera.
3:30 pm
i have a clock in doha with headlines here in al-jazeera and egypt's only democratically elected president mohammed morsi has been buried at the cemetery and car day after he died in court security forces kept journalists away from the burial service and state t.v. says that morsi had a heart attack he had been in prison since 2013 when he was toppled by the military . the rebels say they've launched another drone attack on ports in the south of saudi arabia it is the 6th time the effort has been targeted since wednesday who thinks of war and they will continue attacking saudi airports in retaliation for the kingdom's actions inside yemen russia's foreign ministry has warned the us
3:31 pm
against deploy more troops to the middle east calling the move provocative washington is planning to send around a 1000 additional troops to the region amid rising tensions with iran before making that announcement the pentagon released pictures that it says show elite iranian troops removing an unexploded mine from one of the 2 oil tankers hit by explosions last week hong kong's me to kerry has offered what she calls her most sincere apology for the way an extradition bill has been handled she's under pressure from protesters who've overnights large demonstrations they oppose moves to allow suspects to be extradited to mainland china says she heard the protesters demands loud and clear but is refusing to resign pro-democracy members of hong kong's legislative council so that caroline's apology was too little too late gary lamb is actually presenting to the rest of the world that the administration it's going to be a lame duck and ministration because she would have
3:32 pm
a very hot tiny governing hong-kong from now on she acknowledged that it's the one question of trust and. confidence in her administration which has become completely non existant. gunman in the democratic republic of congo have killed a chinese national and the congolese soldier bush on monday increased violence has led to the displacement of hundreds of thousands of people from the north east and it turi province fighting between tribal groups has increased over the past week killing nearly 200 people in your state the headlines no news hit right off to the big picture.
3:33 pm
the. you know. the risks of bias baked into machine
3:34 pm
learning aren't just confined to law and order upon release prisoners must reintegrate into a world that is increasingly automated. today for them as for you and me hope a computerized systems will help decide their access to state welfare to private finance and to housing to credit scores these are shorthand for a person's financial trustworthiness in many ways credit scores are the gatekeepers to opportunity and increasingly they're produced by algorithms fed on data blind to context and history. if that credit report comes back with a low score that means this individual is supposedly a high risk so you begin to sort of just go around in a circle. low credit score criminal background can't get housing because you don't have housing you can't get a job because the job that you're applying for requires
3:35 pm
a permanent residence and therefore again are stuck in this cycle of an opportunity you're at the whim of a machine driven system. decides on the basis of different criteria that are on the nonce to you. this is one of the darkest topics of. their human biases in targeting on the on the battlefield their human biases in who gets loans their human biases in who is subject to arrest and human biases are horrible couldn't we fix it with algorithms that wouldn't be biased but then it turns out the algorithms are perhaps were. worse the algorithms have refined the worst of human cognition rather than the best because we don't know how to characterise the best.
3:36 pm
i went to the work rebooted conference in the heart of the tech industry san francisco california to see if a i could be used to bring out the best in human endeavor some people are going to do well some people can do less well i met ben prng who heads the center for the future of work at cognizant a multinational corporation specializing in i.t. services. i know a lot of people anxious about the whole notion of bias within the algorithm and so one of the jobs we've speculated on the to be creative is what we call an algorithm bias order to which could be a sort of morphing off of the traditional kind of cool order row to make sure all of that the reason unconscious bias within. algorithms are going to production environments within big business is so that people can reverse engineer the
3:37 pm
citizens made by software you do look at job opportunities opening up you know have said that you do anticipate some job losses in certain areas yeah occluding some that actually people you think haven't seen there is a class of new software there's a motion the last couple of years in the industry it's called republic process automation. and you can get a team of 500 people down to 50 people that's the reality of what's going to happen in big business is that a lot of that kind of white cold you know skilled semi skilled work mid-level mid skill level what is going to be so you know is replaced by this kind of software in the snake denying that some people will be kind of left behind in that in that transition so what other jobs do you think that ai might. open up in 5 or 10 years time so we came up with this job we call a walk or talk which is this idea that you know in a lot of. towns around the world certainly where i live in
3:38 pm
a lot of seniors they're very isolated so what if there was an imbecile platform where people in the neighborhood could log on to the platform i've got a spare hour on a tuesday afternoon or saturday morning i could go and walk and talk with a senior in my neighborhood so people living in the kind of gave the economy a living a kind of portfolio style set of jobs they maybe drive. they maybe drive a lift they may be due to their house through air b.n. b. they may do things through task rabbit what if they could literally monetize that spare time they have to go and walk and talk with a senior that doesn't sound like a technology based job but that would all reside on a ai infused platform in the same way that being the. most of the people who do care work are women and women of color guess what guess who has been taking care of other people's kids since they were in slaves and brought to north america black
3:39 pm
women this idea that somehow these historically oppressed suppressed communities are now in some type a better situation because there's an app interface between them and then the new people who want that work done and then call it a fascinating new gig ng opportunity i think is just completely nonsense the experience of marginalized people basically foretells what's to come for the entire population degrees of control or lessening of autonomy. a real difficulty in confronting. and sometimes resisting these systems. some say if you want to know what's to come with ai you need to look to china.
3:40 pm
the chinese want to be the primary innovations center for. both a potential driver of more social instability but at the same time the chinese state thinks that i can use this tool to call social address. china is home to 1400000000 people its capital beijing has more surveillance cameras than any other city in the world facial recognition technology is woven into everyday life getting you into a bank your residence checking you out at a shopping till 800000000 internet users and weak data protection laws the chinese state has access to colossal amounts of data and china's credit scoring system aims to go far beyond finance. there is this ambitious goal to have a national unified social credit system that would assign
3:41 pm
a score to citizens to judge whether they were their behavior was politically acceptable or was socially desirable. the plan is for all chinese citizens to be brought into the social credit scoring system in 2020. and uses data everything from financial records and traffic violations to use of birth control and processes that data through algorithmic software to give people a score for their overall trustworthiness. a high social credit score could mean better access to jobs loans travel and even online dating opportunities a low score can mean being denied some of the modern benefits of said. probably the most troubling aspect of social criticism is not necessarily the social credit system itself but actually the application of some of these facial recognition technology is to expand the surveillance state and to check behavior of citizens in
3:42 pm
the western region of job where ethnic minorities waders have been disproportionately targeted in terms of their location being tracked 247 whether they're going to mosques which areas of their travelling to and that has been empowered or is in the process of being empowered by facial recognition algorithms being connected through security integrators. autonomous region is home to china's weaker population. and ethnic muslim minority that has faced systemic forced dissimilation. a small fraction of the weaker resistance to this oppression have turned to violence . including attacks on civilians. leading president g
3:43 pm
jumping to embark on a so-called people's war on terror. aimed at stamping out weaker separatism and imposing a secular ideology. new ai led technologies particularly facial recognition are the latest weapon in xi jinping crackdown. some reports have indicated that it was a database that tracked 2600000 residents of. tracked where they were going and that database had labels of sensitive locations like whether they're going to a mosque or whether they were going to this particular region job so that was updated. on a 24 hour basis and that database had i believe more than 6000000 records so it showed it was tracking these people real time. are now in reeducation camps.
3:44 pm
so that's a pretty significant departure from normal life where you're forced to study in a camp and repeat party monstrous. it's a stark picture of how artificial intelligence can go wrong the chinese government deploying ai to track and suppress its own minority populations. recognition checkpoints engine junk use deep learning technology to identify individual leaders cross checking them with data collected from smartphones to flag anyone not conforming to communist party as unsafe a threat to state security. has become a test bed for authoritarian ai. this harsh system of control may seem a world apart from the west but systems like social credit actually have some parallels. in some ways if you think about the origin of some of the the social credit coming from some of the major private businesses in china how
3:45 pm
different is it really from a kind of experience or an equifax or one of these set of credit rating agencies that actually do collect also very granular date on westerners and that data is then shared with all kinds of other entities and used to make consequential decisions in current operation i would say that they are different i think the difference will be when it's not just your financial behavior one it's also your social and your political behavior that gets observed and oftentimes the social credit system becomes a projection of our own fears about what is happening in our societies in the west where it's not necessarily what is object really happening in china that is important but it's about using what's happening in china as a way to project. what we're fraid of.
3:46 pm
so when i think about china's millions of wiggers being tracked 24 seventh's by i am potentially put into reeducation camps i think about the black community in the united states i think about predictive policing and i think kind of east and west one of the problems and worries i've got with ai is the way that it gets road tested on communities of color in the for. the ways these technologies are being developed is not empowering people it's empowering corporations they are in the hands of the people who hold the data and that data is being fed into algorithms that we don't really get to see or understand that are opaque even to the people who wrote the program and they're being used against us rather than for us. there's this incredible
3:47 pm
informational imbalance isn't there that even as a handful of companies are acquiring more and more and more detailed information about each of our intimate lives we've got in some ways less and less information about them in the way that they operate it's nonsense when you think about the way in which fraud and corruption are words that get pointed poor people who get tracked into these high highly surveilled systems. if they don't participate they actually have no other option you don't get food if you don't participate you don't get to go to school if you're not in the system properly tracked and so i think these kinds of things are. are the questions again and that also might need to be regulated you know beyond kind of the technical regulations one of the limits there is that so much of the pressure or focus in those movements is about perfecting the technology instead of thinking of more broadly about like what are
3:48 pm
the values that we're trying to implement and. who are they in service of now see to you worked a little bit with the obama administration didn't you on trying to determine how we make some of these automated systems more accountable to us did you find that that was a useful exercise how did that go i think there was a genuine interest in thinking about what might be harmful what might be helpful what should we think about it now in order to forestall or prevent particular outcomes that we can't undo that's right further down the line and so there was a lot of interest i think what's happened since then is there has been this increasing crescendo from industry saying these technologies are inevitable whether or not you like it they're coming. and what that creates for
3:49 pm
members of the community for citizens consumers is. increasingly a sense of despair or resignation we might not be able to do anything about it and given that increasingly it looks like governments are actually punting to corporate governance structures or cut corporate governance bodies it can create a sense of despondence i'd like to pick up a slightly different but i think related part of that to do with a the kind of supply chain and actually the labor that is involved with some of this artificial intelligence because i feel like the phrase ai sometimes kind of hides like a lot of human labor that's used to make a given system work right so you've got people in kenya who've got a label images to train software for self driving cars or people in phoenix right on very little wages looking at videos that would come up on you tube looking basically all day every day at a stabbing or a beheading so that that stuff can be taken off and you and i don't see it on our
3:50 pm
social media feeds is that one of the kind of problems we don't see a hidden problem of some of the artificial intelligence economy that there's a lot of human labor that is required to prop it up surely you don't see the data janitors who pay the day they're not the ones that are. you know in our line of sight as things like silicon beach expands we don't see the sort of this aggregated geographically dispersed nature of these ai companies and who all is involved in making and cleaning data right and i think that's that's highly problematic because it is contributing to this sort of magical or that surrounds a i can do all of these things efficiently and instantly and yet there's this whole kind of body of people that contribute to that and the fact that in many cases their labor rights are being disrespected i think is also a cause for concern. yeah i mean i think we know now for example from
3:51 pm
researchers i think of my colleague at u.c.l.a. sarah roberts has done all this work around commercial content moderators bringing them out of the shadows so that we actually understand that there are huge. dispersed global networks call center like environments where people are doing this kind of moderation that you talk about. you know one of the reasons why i think we previously didn't know about them is because there's such a deep investment by you know the sector and thinking at least in the us contacts of the internet as a free speech zone for example and that anything goes but of course we know that anything doesn't go i find it always interesting when i hear the machine learning experts talk about how how crude in many ways things like kind of visual mapping is like you know is a table a table is the cat a cat right still trying to figure out these really rudimentary kinds of questions
3:52 pm
and yet when we see tech leaders in front of congress they say things like we're going to take down you know damaging content of violent content content you know live murders live suicides with i sed to protect workers and i think you know that's really interesting because ai is not there. corey right to me. the role of ai in medicine is to. make better predictions but to make the doctors lives better if you just look at the camera and smile. but there are some fields where ai is already there and has the potential to do great good. it could well transform the way we practice medicine a lot of what we're trying to do machine learning or big data in health care is to predict these healthy to disease transitions so really tracking your trajectory over time right and use those to guide treat. so let's have you think about bill's
3:53 pm
on here at the lab 100 clinic at mount sinai hospital in new york city and relax i go through and i've driven health check that generates a heap of data so now i'm going to hold on to your results right providing a more complete understanding of my physical well being. with access to this kind of information doctors could save lives and potentially millions of dollars along the way one of the most mature areas in medicine is the application of ai to imaging data and the actually deep learning came from image analysis and video analysis it was really well tuned for that type of thing so an example looking at radiology images and diagnosing a tumor or you know finding a hip fracture those tools were already well tuned for that task i mean finding cats and videos but i think what's clear is is that the ai is at least as good and men insert many cases in reality as a human is equivalent to a human radiologist might be more like airline pilots in
3:54 pm
a way so airline pilots are kind of there for you know takeoff and landing and then the plane flies itself for the most part but i think what radiologists are going to basically be doing is looking at the radiology image and basically rubber stamping it for. legal purposes really until we solve that problem with ai. ai that prevents disease what could be better. but in a world where people have to pay for health care what would be so great is that private companies use your ai health profile to charge you more with. the future of ai and health doesn't just depend on the tech it depends on our values. is health care a human right should a person predisposed to heart disease or cancer because of their low income or ethnic background have worse care than those better off. the aim should be decent standards for all not a 2 tiered system. there
3:55 pm
are so many positive potential applications of artificial intelligence that would change the world for the better one is a very obviously the pattern recognition that ai is good at has proven incredibly good at spotting malignant tumors an incredibly powerful and inspiring medical advance that i've seen some papers on just in the past year but the technology is going to shortly underpin all aspects of our daily lives very shortly some form of machine learning artificial intelligence will determine whether somebody can loan whether somebody gets a mortgage whether somebody gets bail whether somebody gets paroled and as we've seen it may well determine matters of life or death in a military context so the stakes could not be higher the quality of your decision making absolutely depends on the quality of the material that is coming into it and we have seen and other uncertain human contacts such as policing that math data
3:56 pm
runs through machine learning algorithms has a distressing tendency to replicate and accelerate all of our preexisting human biases. if you have a whole. technological culture infrastructure. language that emphasizes the lack of human responsibility and instead emphasize a system where there are these artificial agents we pretend they have agency but what's really going on is received a trick to manipulate each other then they'll be more and more trickery and manipulation and if we want to reduce things like turn attacks we have to empty. since human responsibility. the drone attacks that killed my client's family showed just how much responsibility we're handing over to technology. but we can take that
3:57 pm
responsibility back our curiosity and drive to innovate has been pushing the bounds of what we can do with artificial intelligence for decades. as ai is used to make more and more decisions about us from targeting to policing to social welfare it raises huge questions while i would be used to target minorities or clean up our air will it destroy our privacy or treat disease will it make us more unequal or fight climate change these are questions that should be decided in the boardroom of a software company what happens with ai is everyone's business the world according to ai is our will and it's up to all of us to be sure it's a just one. we're
3:58 pm
. going to weather the storm surge by the time. hello there we've had some very very heavy rain over parts of south america recently most of it has been over parts of argentina and 3 year ago i as well here's the cap responsible the main focus of energy from the system there is now pulling away towards the east so it's taking the worst of the rain with it there will still be somewhat weather around though as we head through the day today but as we head through wednesday most of that clears away and as it does so it will allow the temperatures to rise a little bit as well as one as always should get up to around 15 degrees as we head through wednesday elsewhere and there's a good deal of bright weather around and santiago will see some cloud at times that 17 should be our maximum further north some showers here and they really get going as you head up through parts of venezuela and columbia very heavy showers here
3:59 pm
russo seeing plenty of them further north as well as some of us in mexico we're seeing some heavy downpours as we go through a parts of cuba to those over mexico will still be coming and going during the day on tuesday and wednesday and then further east this area of rain over cuba it won't try and pull away eastwards as we head through the day for the north we've had some very very heavy downpours over posts and north america in fact some of us have been under flash flood warnings thanks to this system which is falling on already saturated ground that is going to try and pull away but still bring us rain for tuesday. the weather sponsored by cattle and ways. they wanted 43000000000 pounds with a weapon that was 6000000000 pounds in commission. there's no hope of any more because there's always a small cobbles people for really really good business. in assonance we
4:00 pm
in the united states have privatized the old public function more shadow on al jazeera. this is al-jazeera. that i welcome to news. and come up the next 60 minutes the family of mohamed morsi say egypt's government denied their request to bury him in his hometown all 3 died and called. president morsi isolation and el treatment might actually amount to torture 0 rights groups say morsi was killed slowly by poor prison conditions but government particularly in the west a largely silent also ahead. my most see.

20 Views

info Stream Only

Uploaded by TV Archive on