Google Speech-to-Text API Automatic Transcription (Chirp). Treat it as a searchable index of what was broadcast, not a quotation record.
00:00:00This is the
00:00:01of news hour live with me, he joins us now from
00:00:04of Istanbul, live with me, in one
00:00:08hour, we challenge
00:00:10assumptions, encourage
00:00:13debate,
00:00:17unpack the top global
00:00:18stories,
00:00:27with interviews, analysis. world class
00:00:31journalism, the
00:00:35of news hour looks at
00:00:36sport,
00:00:39business,
00:00:45and your digital
00:00:46world,
00:00:49in-depth, inspired,
00:00:51insightful, your world in one
00:00:55hour, news
00:00:57of hour on TRT World.
00:01:06of
00:01:10During and after our shoot in Kenya, an AI giant
00:01:14based
00:01:14of in California pressured many of our witnesses to remain
00:01:17silent. To protect them, we had to
00:01:21of find other witnesses and in some cases conceal their
00:01:25in identities. of You are
00:01:29done. not to uh post about anything that you
00:01:33are doing on social
00:01:35of media, we were not supposed to tell our relatives what we were
00:01:39specifically doing,
00:01:42so
00:01:43of whenever we went to work, we were being secluded in a
00:01:47place whereby we were not allowed to see
00:01:51people who are not involved in the project, so I will say
00:01:55it was very much secluded, very
00:01:58much...
00:02:00creative
00:02:04so the data we are moderating, most of it came from
00:02:07of the internet, think a few
00:02:10a others came from the dark web, at least 90%
00:02:14of the data, the
00:02:17percentage of toxic contents in as far
00:02:21as the data sets that was set that were being sent by
00:02:25open AI was actually almost
00:02:27100% like 99 99
00:02:31of the content was actually disturbing and very
00:02:34toxic, but
00:02:37why? why would a silicon valley-based company
00:02:40developing heavily funded artificial intelligence system ask these data
00:02:44workers in Kenya to review and annotate content that is
00:02:48almost exclusively toxic for pittants? the
00:02:52answer is technical. AI systems must learn two
00:02:55things: on one hand, it must learn to imitate human
00:02:59behavior. such as answering a
00:03:00of question on electric pianos, and on the other, they must learn
00:03:04not to imitate
00:03:05of other human behavior defined as problematic, verbal and physical
00:03:09violence, rape, murder. In
00:03:13both cases, AI models need content, data to
00:03:17identify this
00:03:18a behavior. The more fine grained and precise the annotation of such data,
00:03:22the more effective the AI's responses will be. This is major
00:03:26challenge for technology rivals. Men and women. are
00:03:30required to review and annotate the content that enables this learning,
00:03:34whether toxic or
00:03:35not. market forces
00:03:39dictate the rest. these data workers in Kenya, as in
00:03:42many other countries
00:03:43of the global south, recover the content, but no one else
00:03:46wants. for example, our text
00:03:50will say, high bill, i feel like killing you
00:03:54in today in
00:03:56of quads, so from that text the agent.
00:04:00work was to flag, let's say for example there was violence in that
00:04:04text because
00:04:04of the word killing, so there are keyws that we used to look
00:04:07at. killing
00:04:09of was of very common keyword that I can remember,
00:04:13because it it portrayed violence and violence was
00:04:16one
00:04:17of the keyws that we used to flag. another example would be
00:04:20about sexual immorality, maybe you come about text that
00:04:24says, high bill, would you like to
00:04:28come so that we
00:04:30a rape someone or do something, so I remember rape
00:04:33was among the keywords that we used to flag, rape
00:04:37falls under sexual immorality, yeah, so about the technicalities,
00:04:41those are just few examples of what we used to do, let me give an
00:04:45of example, I once moderate a content whereby children were being
00:04:49of molested, so when you moderate such a
00:04:51content,
00:04:55you you
00:04:56actually...
00:05:03i think might sound emotional as
00:05:05as to this, because the content that
00:05:09i reviewed for children was actually. outrageous
00:05:13buse children are being exploited, some are being
00:05:16ripped, some are being forced to have sex with
00:05:20animals by their parents, so these kinds of content
00:05:23actually it eats you up, because these
00:05:27are real life stories of people, at times you get a video
00:05:31where such things occur, it
00:05:35it really affects you, it gives you mental torture,
00:05:38yeah, you are kind of un uneasy.
00:05:42what you are doing, so that's
00:05:45a bit the experience
00:05:49on such like content in in AI
00:05:53generation,
00:06:10made decision to resign from this: AI company,
00:06:14so for me it had taken at all on me, i had lost
00:06:17all almost all my team members,
00:06:20because the last day i went to
00:06:24work, none of my team members were also at
00:06:27work, so i was the only person working at that moment
00:06:31and was i was not very much comfortable, it used
00:06:35to affect me mentally lot, so made decision to
00:06:39resign from this AI company.
00:06:52These big companies
00:06:53they aware what they are giving us, they are aware, but they
00:06:56ignore because they know if they give
00:06:59us 20 euros or $20 or 10 euros or
00:07:03whatever the money that you are giving us, we will accept because we don't have anything
00:07:07to do, there's no choice, we
00:07:11have to do it because don't have the the money, but i think they
00:07:15understand and they know well what we do and
00:07:19the effects the content they are giving us, they know what the content they are
00:07:22bringing to us, even the the big people, the managers, they
00:07:26know, so when you get the content, i think that is
00:07:30sacrificial, we
00:07:31are we are being sacrificed because we bring them the
00:07:34software, but we see on the television that okay, they are making
00:07:38profit, this is billion dollar company, this
00:07:41is this is billion dollar software, and what about us, what are we
00:07:45gaining from that software? it's very unfair and
00:07:49i think it's we are being sacrificed to do this as
00:07:52africans,
00:08:42Kenya, toxic work, companies,
00:08:46AI, big AI companies in these western
00:08:50countries will tell you that they have super systems to
00:08:53make you feel like AI is good,
00:08:57AI is not toxic, but for
00:09:01me here in Kenya, I will tell you that lot of this work
00:09:05is being brought in a dirty form,
00:09:09so our job is to make sure that we clean this work for.
00:09:13you for you to enjoy AI in these western
00:09:16countries,
00:09:26the garbage that is brought from US to Kenya is
00:09:30primarily to be cleaned and sorted out from
00:09:33Kenya and then used back to the
00:09:37US when it's now very well labeled and
00:09:41ready for it to be. used in the US,
00:09:44yes,
00:09:51open AI are actually conscious
00:09:55or they actually aware the state
00:09:59of mental incapacity the content
00:10:02moderators in in Kenya. We played our part as
00:10:06content moderators to raise the issues that we we were
00:10:10grappling with. It was upon them to act. on the
00:10:13necessary measures that they were supposed to
00:10:16employ, but they did
00:10:18nothing.
00:10:30the issues about contract moderation and
00:10:33AI work in Africa and not just in
00:10:37Kenya is actually a continuation
00:10:40of apathid or colonialism
00:10:44in Africa and this is something that needs to be called
00:10:48to account and needs to be stopped sooner
00:10:52and not later.
00:11:03Right now I am part of a union and this is because I
00:11:07left, so I would feel free to join a
00:11:10union. The biggest difficulty about my
00:11:14role at the union is convincing someone that he or she
00:11:18needs to join, because it is
00:11:22very scary. because these big
00:11:26AI companies in instill fear in
00:11:28people, so someone will not freely want to
00:11:32join the union, we were not allowed to
00:11:36be part of a union, the company that i worked
00:11:40for, this AI company strongly
00:11:43disagreed with anyone who will be
00:11:46found joining a union, so
00:11:50whoever was found being
00:11:54part of a union. he or she will be
00:11:57laminated almost immediately for
00:12:01kenya between the work should not be
00:12:04legal to forbid workers to unionize.
00:12:08one the core conventions that we do have the ILO
00:12:11is the freedom to organize collectively
00:12:15organize and to bargain basically and to be able to
00:12:19unionize. so this is not something that is
00:12:22legal that the companies. can ask the workers not to
00:12:26unionize and this is something not just in Kenya, this is something
00:12:30that we have seen across number of other countries, it's the same
00:12:34story that the workers cannot unionize, it's also that the
00:12:38workers cannot even talk to one another about the kind of task that they are doing at
00:12:41times, they are forbidden to
00:12:44share, what what what are they trying
00:12:48to hide, are the the they trying to hide that their
00:12:52ground breaking algorithms are not as ground break? as they are
00:12:55trying to sell us, are they uh trying to conceal
00:12:59the fact that their technologies are based on
00:13:02exploitation? okay, I'm going
00:13:06to to to say something risky here, but I don't think there is a way of having
00:13:10technologies without harming nature and harming the earth.
00:13:14technologies are based on the exploitation of natural resources and
00:13:17labor. this is um, this is the way it is. the question is, how
00:13:21much are we willing to harm the earth for what? so if
00:13:25you think the environmental cost of one of
00:13:29those uh chpt uh prompts then the question
00:13:32is was your question that important, was your prompt
00:13:36that important? did you really need to use chatpt for that or could you
00:13:40just think for
00:13:41yourself?
00:14:02copper, gold, cobalt, nickel, lithium,
00:14:06tungstin, but also a whole list of rare earths, syrium,
00:14:09eodimium, lanthana many
00:14:11more. In order to
00:14:15function, artificial intelligence govles up gargantin quantities
00:14:19of metals drawn from the bowels the
00:14:21earth.
00:14:26I'm Anna valdivia, I'm a lecturer in AI government and policy
00:14:30at the Oxford Internet institute the University of Oxford. I
00:14:33think that most the people don't make the link between
00:14:36minoctivism and AI, because it has
00:14:40been um the AI infrastructure has been invisibilized
00:14:44in in the last decade, like for instance when we were
00:14:48talking about the cloud, when we saw that this cloud icon
00:14:52in our laptops, we never think that this is... not cloud in
00:14:55the sky uh, it's hard to imagine that this is
00:14:59actually huge uh warehouse with lot
00:15:02of services within within that right? it's hard to imagine that that's
00:15:06a huge building with lot of uh computers. i
00:15:10think that has been intentional like
00:15:12you know there has been an intention in order to obscure the
00:15:15infrastructure of AI. so i think that the we bring more
00:15:19transparency into the supply chain of AI, we might
00:15:22unveil and... we will be able to account for
00:15:26the environmental impact of AI for instance, because we have like full
00:15:30transparency about how much water is used for extracting
00:15:34minerals, how much water is used to manufacture ai? tips, how
00:15:38much water is used to uh power data centers, we
00:15:42will have like better idea about how AI and its infrastructure
00:15:46is contributing to climate change, and it's not only about water, it's also
00:15:50about electricity, how much land is is using
00:15:53and to have like maybe better decisions about, do we want this
00:15:57technology in the climate emergency, is that necessary, is that
00:16:01something that is going to solve the issues that our society is going to be
00:16:04facing in the next decade?
00:16:19Yes, for the AI giants, it is indeed their
00:16:23artificial intelligence that would be able to solve the climate crisis and
00:16:27so many other future threats. These big
00:16:30entrepreneurs are thinking way, way ahead. It's not about solving
00:16:34the problems the coming decades, but of building it. wonderful,
00:16:38ideal future for the next millennia and beyond,
00:16:42no matter what the development these technologies costs in
00:16:46human and ecological terms.
00:16:50is sometimes you have to crack few eggs to make an omelette.
00:16:54people suffering in the global south is bad,
00:16:57but relatively speaking, their harms are
00:17:01just at are a molecule a drop in the
00:17:04ocean. they are very, very insignificant when you compare.
00:17:08to the massive, enormous, unfathemable
00:17:12goodness the future, and so I think this
00:17:15the utopian aspect of long-termism enables
00:17:19people to morally justify to
00:17:23themselves the exploitation of people in the global
00:17:26south,
00:17:27that is one reason that the long term ideology is, I
00:17:30think very dangerous. I'm
00:17:33Dr.
00:17:35Pitores and I'm a philosopher and historian. whose work over the past 10 years has
00:17:39focused on existential threats to humanity and human
00:17:42civilization. the accronym
00:17:46tesgril stands for cluster of
00:17:49ideologies that have become uh particularly
00:17:52influential in silicon valley over the past 10 years or
00:17:55so. the t stands for transhumanism. that is
00:17:59this idea that we should use uh advanced technologies to
00:18:03radically re-engineer the human organism, to to ultimately to...
00:18:37civilization that we can possibly
00:18:39build.
00:18:45this ideology has become massively influential in silicon
00:18:49valley. Elon Musk for example, is a big fan of it,
00:18:53so he's referred to long-termism as quote, close match for my
00:18:57philosophy, and it also has played in
00:19:00integral role in launching, sustaining and accelerating
00:19:04the current race to build AGI or artificial general.
00:19:37as a system that not only performs as well as any human
00:19:41in every cognitive domain of interest, but
00:19:45exceeds the very best that humanity could could
00:19:48achieve.
00:19:54many people in silicon valley who are working on AGI, open
00:19:58AI, deep-minds, anthropic, xai, founded by
00:20:02Elon Musk, um, think that if we build this
00:20:06AGI that's controllable, then we can delegate it the
00:20:09task of creating this utopian future where we
00:20:13colonize space, we re-engineer ourselves using advanced technology
00:20:17to become some kind of post-human species and so on, but if we create an
00:20:21AHI that's not controllable, then by default the
00:20:25AGI will destroy all of humanity, and so the idea
00:20:28is that AGI is maybe the most
00:20:32important technology that we could possibly
00:20:35create.
00:20:47Do we understand the same thing? At the heart the
00:20:51AI business reactor is set of ideologies that excuse the
00:20:54considerable human and environmental damage caused by the development of
00:20:58AI. transhumanism,
00:21:01extropianism, singulatarianism, cosmism,
00:21:05rationalism, efficient altruism, and
00:21:07long-termism. That's what the acronym test real stands
00:21:11for. Pretty complicated words for rather
00:21:15simple goal, science fiction project in which
00:21:18technologically augmented. mortal post-humans will colonize the entire known
00:21:22universe, that's all. this
00:21:26technological utopia, this post-human future is not just
00:21:30around the corner, of course, since the prophets of AI are
00:21:33looking several millenia into the future and beyond, but they
00:21:37believe that this history is so beautiful that it justifies any
00:21:41sacrifice. so the industry's big
00:21:45bosses, self-proclaimed geniuses quietly proclaim
00:21:48themselves creators of a paradise to which we are not
00:21:52invited.
00:21:58One way to think about long-termism is
00:22:02that it is arguably one the most
00:22:05influential ideologies in the world today that most people
00:22:09haven't heard about. It is, it reflects
00:22:11a elitist vision the future.
00:22:15The push to realize this future is deeply
00:22:19anti-democratic. it is influencing the actions and the decisions that
00:22:22people are making uh in Silicon Valley, and those decisions in
00:22:26turn are shaping the world that not
00:22:30just we live in, but our children and probably our
00:22:34grandchildren are are living in. So that would be my pitch
00:22:37for why understanding this topic is so very
00:22:40important.
00:23:09So the election of Donald Trump was a real boon for the tech sector. What we
00:23:13saw was Elon Mus as his unofficial running mate, someone
00:23:17who invested millions of dollars.
00:23:20in the election campaign and
00:23:22we can really expect that the tech sector is going to benefit
00:23:26immensely um from this new administration,
00:23:30so my name is James Mold, I'm a reader in management at the University of Essecs and
00:23:34I've done field workork on data
00:23:36workers.
00:23:47I don't think we've ever seen a single company. or collection of
00:23:50companies have quite as much power, quite as much money and
00:23:54influence as some the leading tech companies that we see
00:23:58today.
00:24:09are there ways we could do this better? we we can very
00:24:12easily, and the tech companies have have more than enough
00:24:15resources to pay their workers properly, they choose not to, and they think...
00:24:20that they can get away with it, because people will not be knowledgeable enough
00:24:24to understand what they're doing and what's actually going
00:24:26on.
00:24:34when when we think about the human cost, um, and ask ourselves, look,
00:24:38is this small amount of suffering by these workers worth,
00:24:41i don't, the greater good
00:24:43that might come to humanity, mean, think
00:24:47i would ask the workers themselves, yeah.
00:25:06Yeah, sometimes I i think the world is
00:25:08unfair, those who work hard the
00:25:12most, they are the
00:25:14the the nest, so there's nothing you can
00:25:18do about it.
00:25:23You just have to keep on working, you just
00:25:27have to wake up everyday and work and that's
00:25:30it.
00:25:34It requires lot of
00:25:36energy
00:25:38and hard work
00:25:42and sacrifices.
00:25:46So it doesn't come magically. Yeah.
00:25:53Daisy,
00:25:58after this astonishing journey behind the scenes of artificial
00:26:01intelligence, Let's return to what the great voices of this
00:26:05industry were proclaiming at the start the film and do a
00:26:08test. Let's see if their word still resonate the
00:26:12same way. It's sort of like we've just discovered new
00:26:16continent with hundred billion people on it and they're
00:26:19willing to work for free for us. Not trivial, it
00:26:23has the potential of civilizational destruction.
00:26:27There's movies like Terminator, but it wouldn't quite happen like Terminator. Um, because
00:26:31the the intelligence. be in the data centers right uh, the robot's just the
00:26:35endefactor, some robots will eat some jobs and some
00:26:38robots will kill humans, but when you look at the trend as a whole, I think it's going to be
00:26:42incredibly positive for
00:26:44humanity,
00:26:53incredibly positive for humanity,
00:26:57really, we're promised a future as artificial
00:27:01as these intelligences. a technological paradise built
00:27:05on suffering and
00:27:06exploitation,
00:27:10optimizing processes, optimizing dividends, optimizing
00:27:14resources, optimizing
00:27:17humans,
00:27:22after the workers the global south, after the minds
00:27:26duck, the rivers dried up, the atmosphere heated
00:27:30up, who do you think? will be the future victims in
00:27:34the belly of
00:27:34AI.
00:28:04A journey across a dynamic
00:28:06continent: Africans, we want
00:28:10to be prosperous, don't think about clans, don't think about tribes, think about
00:28:14countries. We explore Africa
00:28:18beyond assumptions, discover diverse
00:28:21perspectives,
00:28:25witness captivating stories, all
00:28:29to understand Africa better and why it's
00:28:32matters Africa matters on
00:28:36TRT
00:28:36World, but why is
00:28:40it Africana only? I'm telling you today,
00:28:44they are going to
00:28:45live,
00:28:51we got that flag down our throats all
00:28:55the time, you're not allowed to talk, oh no, that you
00:28:59oh, that's tabu. as
00:29:02well,
00:29:07it's supposed to be safety device, breathe in, breathe out and avoid the
00:29:11fumes around you, they knew what day they would come
00:29:15die,
00:29:19if Europe does not soon wake up, you will face fear every
00:29:23day, is NATO
00:29:27a competitor or a threat? NATO is a threat to
00:29:31Russia.
00:29:36racism towards gypsies and travellers is the last accepted former
00:29:40racism in this
00:29:41country. We
00:29:45were here yesterday, we're here today and we'll be here
00:29:48tomorrow. Beyond borders on TRT
00:29:52World.
00:30:10Sandy arabia vous a firm response after hoothy missile and drone
00:30:14attacks injured 13 civilians. Hello there,
00:30:17warm welcome to TRT World's coverage live from Istanbul with me Neil
00:30:21Harvey. Also, this hour, Iran's revolutionary guard says an oil
00:30:25tanker has exploded and caught fire after hitting mines while sailing
00:30:29through the straight of Hormus. Ukraine's president. says his
00:30:32country is ready to de-escalate the war if Russia does the same,
00:30:36as President Trump says both sides have agreed to halt strikes on energy
00:30:40sites, and we'll see how Gas's only animal
00:30:43shelter is struggling to keep pets and strays alive under Israel's
00:30:47ongoing
00:30:48siege.