is an old who stopped the police have enlisted the help of international authorities to find the thieves and their precious loot and you're up to date, but to stick around we'll have more international news for you. the top of the next hour. i'm here until through. thanks for joining us. the questions got any issues with a lot say who the is the most important stuff can be used across different geographies. the real challenge, it softens, needs to be incredibly scarce, waste and transforming business syllabus onto leo media

and lots, just green washing. what's now on? the artificial intelligence is part of our lives being on social media platforms or in the am a checkbox we use among other things a i make sure we're protected from harmful violence content. but there are real people behind this technology. we do not too far from the world litzy check homes, these data workers trend a systems by sifting through massive amounts of harmful data. it's tells you explain to me how does the flesh of a human face like while they carry out the awesome ruling task of keeping a systems clean, international corporations on the other side of the globe rate,

the profits are we'd be used to move before the unit is from kenya and germany. we set out to investigate this back story of a i on covering a web of global connections. that power a huge industry the without us there will be no end for that because we have the most important part of the, of this ai. what's life like for those training e i systems that the entire world now relies on our quest to find out begins in kenya's bustling capital of nairobi. here we meet junction. you are a single mother of one in her twenty's. she left university to support her family, bouncing from one job to another until 8 years ago when she began taking on data

annotation of work that she found on platforms online. for years, this work mostly meant labeling elements and images. you know, take maybe vehicles, cheese people like everything that is on level death is visible, including road signs, but not in other words, in the woods of san francisco, a jill in town to all sorts of content, these odd jobs came in around the clock. she was paid by the task awesome, just sense. ready over time the work became more and more sinister. ready as time goes by, the content that you've exposed yourself to, now it's fetched to happen. what does she mean by that? we'll get to that. but 1st, let's house for a 2nd to understand the role of data, entertainers like her in training, cutting edge a systems. most of today's may i analyze this huge amounts of data.

the goal is to identify patterns. for example, how to identify a cat or a dog. this is how the system learns that cats tend to have pointed ears and of triangular faces. while dogs tend to have brought our faces and fluffy years, but what is the data isn't straightforward. this is where human data imitators come in. their job is to make sure data makes sense to computers by teaching them. this is a cat. this has a dog. and these are neither a man, however, this often means dealing with questions. finally, we're done here. then cats or dogs is loud.

well, pregnant, she sometimes worked 18 hours straight until 3 in the morning, pushing herself to the brink. that's the bad thing. after i'm going to the fed up used, and that is when need be a be a line to that the content that have been exposed to has affected me to a point that i do not even know the world's bank estimates that hundreds of millions of people around the world are now engaged in online gig work often with uncertain income streams and lack of contracts. in the west, many check giants have started outsourcing this work to countries in the global self, such as kenya, data workers, they are reported to sometimes earned less than $2.00 an hour, compared to more than 20 dollars in the us. we meet another data worker somewhere called face. she asked us not to reveal her face to face trend large

language models without being told which specific a chat bought her efforts were supporting with time. also, her work became more troublesome to the team. give you a question. describe coming boys. if you have to put yourself in the shoes on the chat board and chat. and you said in that topic that you have been given faith chose as to jot allowed her to help her elderly parents pay their medical bills. but there was only so much she could take. it tells you explain to me how does the flesh of a human face like, can you boy, can you fry a human meet? i don't know. face realize this work was more than she could handle when she concluded her job was to train the chat. bob, to provide answers to user questions about these topics. i was like, i'm out if they are know more about projects that i can do, that not least i'm being then the amount,

but was she really training the track thought to instruct people on cannibalism? what's behind all of this? we ask an expert to explain large language models. the thing is they don't understand commands in a way we know this is all you all ship or chart a leading expert in a i and large language models. we show him our interviews and asked him what he makes of faith account for us humans. it's absolutely clear. i have an 11 year old daughter. she knows nothing about cannibalism. i assume to her i couldn't explain, i guess in 5 to 10 minutes. what it is about and from there on see what know. yeah . when it is okay to talk about it and what is out and what is what is not ok to talk about, but for the machine. so you have to give examples. unlike his daughter an a i system has no moral compass. that's why human data annotate her is like phase are hired to add their tax. they help to a,

i understand when it's okay to talk about topics like cannibalism, that it's okay to provide historical information for example, but not instructions on how to do it. it's all about the context. there are contexts in which you want to talk about categories. when you talk about war situations, but you don't want to talk about the county, but it isn't in the topic cluster of cooking recipes. and now this is the, the, the critic a point with. but these people we're talking about today come into play. so you would feed these examples to the tech but as a negative example. and that's 18 chad, but we used gave us no instructions and referred us to a with a p, a article when we asked it to describe how to boil human flesh. our research has shown that data workers like face, awesome, have little to no knowledge of the i systems they're helping to train. training them involve a complex network of companies face since she found her freelance work through

a silicon valley base platform called remo tasks, a subsidiary of scale, a i it's a company so successful is that it's c e o. alexander wang briefly became the world's youngest self made 1000000000 there, and whose customers include us, tech giants and companies from around the world scale a i declined to be interviewed in an email. a spokes person wrote that training gen a i models to prevent harmful and abuse of content is critical to the safe development of a i. adding that we have numerous safeguards in place, including the ability to opt out at any time. the company we are working for is making millions or millions that mix these by phone feelings each we've been used to make before being done. as like face gentlemen, spend years doing freelance work one day in march of 2024 out of the blue. her

employer remo tasks. ready operations in kenya. ready new skills, you do not have any security, especially in what you're working on like it's not even safe because how does a company where you're working, you've been working for for more than 5 years, just exiting. so in the low diesel, anything in an email to d, w. a remo tasks spokesperson explained that people were not notified due to operational errors. so when i say that platforms and components are predatory, that's exactly what i mean. this brings us to says ologist and computer scientist, milan pulse meat shelley in berlin. they go to a place as long as it is profitable for them as long as it is convenient for them, and then they leave whenever they feel. this is not great anymore. so what happens with rema task is just one case among many, many others that we have of serve. in the summer of 2023 here at the vitamin bomb

research institute, researchers launched an initiative called data workers inquiry we work with data workers, things so far and for continents. the situation on the ground depends on where these data workers live. but researchers say there does seem to be a common denominator. workers were often marginalized. in my experience, so i was doing this research in kenya. one of the work also told me that they were told when they, you, in the organization that we completed and playable, and this was something that was the cost of need repeat it so that you, you know, in your place, you come here. you do this within the full range, you know, one of the one you, i incredible and about to give it to you to grow. it is great. that's people who otherwise don't have a lot of chances get this type of, of, of jobs. the problem is not without the problem is with the quality of their jobs, their perspectives to actually grow. all the remo tasks has left kenya. several

other platforms continue to do business there. it's one junction new and has decided to join the countries growing data workers movement. on the one hand, it's about better pay the union. we advocate for sale working conditions, say a bit and push for legislation and policies for people watching in this field. on the other, it's about better psychological support. you work on very difficult projects, things like dead bodies and such but is not to me. and i see it affects the mental states within that thing is provided for taking care of that. like being for psychologist as an example of similar efforts by data workers to unionize suffering a setback last year when a local contractor working for us tech trained meta fired, 185 facebook content moderators. those dismissed are now assuming the company.

meanwhile, jonah and her comrades in arms are continuing their fight for union and data workers and other countries do the same thing. it's not that dumping place the yes we might have economical issues and crisis, but are you know, stupid? why is it so hard for us to be recognized? why is it so hard to be given for performing the proper medical coverage? why is it for hard for us to be, may be properly compensated and given with what conditions if it is being done in other offensive and it should also be done as the a, our revolution continues to ramp up. it's clear their flight is just beginning the, the force of the car in the you go to the

stop compact from the former yugoslavia conquered hollywood and many a hard why did it become an icon of failure? and why does this lemon remain? so here is this simple red on d, w, the purple and pushed in on the 1st we are far in cost of the options we didn't noise cost of it was ready to accept something like this. we took the step friends and let's try that, steve, and open the box and keep it open. i'm number mod upfold steps and is making a successful not only among the clear community your own mac, 30 minutes dw, no. understand. can have

a think like the right to present dw used on instagram and follow up or the machine, the way you go, a car from the communists world that even one over hollywood it made its mark far and wide. when i see it in a over how i feel like the whole the.