you bought this moses, you opened it, of course this change can only be achieved with small steps. when you might have thought, well, we shouldn't just stay silent again. a karate me nor money mulch it. although the media regulator blocks tiger dot info a year ago. the website reaches its audience through vpn encryption and they want to continue reporting on russia in russia. and set you up today. i'll have all well news at the top of the our british, prodigy has news asia. next. i'm good with species populate our planet. most of them are here long before us. it was on when humans came along that the drinking extinction began. an amusing

diversity of species still exist on the with the british isles. and they are calling to preserve this habitat story of hope. while dials starts on d, w, i sent it up there news, a shock coming up today, artificial intelligence and a journalism, a match made in heaven auto cautionary tale. i news anchors of the latest example of artificial intelligence being applied in the news industry. but how did that really work, and how credible is the journalism that comes from these systems?

ah, i'm british manager, welcome to the dublin news asia. glad you could join us. today. marks world press freedom day and share on d. w. news asia. we thought it an apt opportunity to discuss the future of journalism in a time of artificial intelligence. why? because it's already happening. take the example of india's 1st, e. i powered news anchor gold, a santa she was launched last month by a major media group. she's multilingual and promises to begin working immediately. and this is what she sounds like. joining us. how much been going on the edge after freshman linda stone building during b. g. b yaga in hopefully security widget roundup internet suspended vandalism. austin and stone building on gun budget session resumes so not phase of not

just conroy is defeated, sunrises. handlebar by 72 runs into fullest match. my roy challengers, bang lot trash will be indians might eat wickets at home. like many countries in india to television news anchors have a 2nd connect with audiences. so a bureau, delhi asked people what they felt about a powered summer. i watch news because i wanted to feel excited about something towards the end of the day that i had an i thing that a i just bought a i journalist cannot bring that out. i anchors don't feel, i'm not able to relate to them as much. they see more monotonous than you know, a human and good and i would much prefer watching human and good then and i am good at even if he is making the news then sure i have, i have a lot of concerns about this, but he is delivering the news that a journalist has curated afforded, i'm okay with that news deliverance in the sense of writing an article or delivering an article like this 8, whatever model lady was doing for me. i'm ok with that. but if you're telling me

that this news has been sourced and model object fact checked, aud even summarized by an artificial intelligence. this, i have a problem. but if the preference is given that i will obviously want a human anchor because it was healing dylan's can also have its own biases. who see the when advice has come coming to factor, then the reporting can also be very difficult to which ones pick one for one sector, which artificial intelligence white cell cannot improve. the joining mean office of context is felix simon, communications researcher and doctoral student at oxford university. felix is currently researching the implications of in journalism and the news industry. phoenix, welcome. are news anchors like me on borrowed time. i mean, are a news anchors the next big thing in journalism. i'm personally skeptical that he will be out of work anytime soon. in many ways, the use of a i can use i can, is useful. it can be time saving,

it can make things more efficient. you can, that's in theory contents and audiences, especially on versus mobile apps. but i think what people really value about needs, and because ultimately is the human touch, it's the human element to see which is the sponsor, nancy, and that's something you can really replicate with a i. so i think for now you're fairly safe and your job. how do i generate a news anchors work? i mean, where do they get that information from? how do they know what to say when? so this is a complicated process, which in many ways means you have various ai technologies and all these intersecting side for the m as part of the image generating a model for the language part those or you're generating a models and they all come together in something more or less looks like a human depending on how the system works. in terms of information. i think it really depends on how it has been built, missing some sort of loud lodge language model, underlying it,

which produces the content in addition to the voice on the visuals. that of course, the information, everything that has been gone up by the system at some point, if the anchor is basically just a representation visually in order to visually with the text coming from human done, it's still a human journalists doing the work of reporting and researching and then just feeding a text through that system and then the system will produce something up more or less, looks like a human and sort of read the news and reports. and so what you're saying is you're going to have an option in which this system automatically generates its own text and a video, or you could have a system which is fed this information. and this is just a non human entity, digital entity that, that regurgitates that and presents that to the public is, is that right? yes, i mean, i'm stuck in series of both ways. you could use it and especially with the

advancement of things like chassis, p, t o goose boss, and recent months. i mean, in theory, you can have a news anchor that's digital and which is not really a digital entity in of itself, but also sort of produces contents through an automated prices of some solar, although you will still in the need. some humans alert and to the problem, said she is what you're supposed to talk about. that's one way. the other way is of course you just have a digital representation journalist for feedback. and it really depends on the use case been dose as possible. all right, at this point or if it makes sense to stay with us because i'd like to bring up what the man, commonly called, of the godfather of a. i has to say about artificial intelligence. we're talking of course of dignity, psychologist and computer scientists. jeffrey hinton, who recently told the b b, c and i code right now they are not more intelligent than us as far as

i can tell. but i think they soon may be and we need to worry about that. felix, should we be worried when it comes to a i and a journalism, especially if it's a self learning system as to really hard question. lessing as lot of elements and there which we have to untangle of some sort. i'm personally, i'm not a computer scientists on the other. people are much smarter than me have from the interviews on if we will have artificial general intelligence, which is really what jeffrey hinton is alluding to. i'm personally on everything i know about this, much more skeptical. i don't think we will see conscious machines any time soon. i think we are to your question about the dangers and again, it really depends on, on where you sort of look at and in some context. yes, we clearly have to be careful how the systems are being used regardless of if they are learning or not. so i'm take error is being produced by something like chassis bt. if you sort of lab loose on producing gymnast

a copy of some sorts on the errors of creep into the file outfits and no one checks that, that is the big problem that has nothing to do with the system or the intelligence of the system itself. it's more the way it's being used and if there some sort of oversight, ah, there's other elements where this also applies, but generally is of the system itself. it is relevant, but it's much more important to le kurtz. and the way it's being employed in jealousy context and, and he's on vacation and the choices us humans often ultimately make about the way it's been used. and that is producing our problems down the line. not so much that he knows enough itself at 1st for in simple terms, the information potentially that goes into these he, a systems could be manipulated, for instance, so that he, i news anchors in this example, say what you want them to say that could happen correct. i mean, yes, that's the technical possibility, something and so it's prompting called prompt injection,

where basically this error is actually do that in a context, as we've talked about before. well, you have an a news anchor, which also will also magically produces what they say through some sort of loss language model. and it's not information within the model which they get fed into. it is 40 that of course you will have era in the else. but then again, it's still human decision and the 1st place to have such as this, which is completely automated from the beginning to end. and as long as you have human side, as long as human sorts produce a content, while a can still be errors, but it's human error, it's not some sort of automatic thing which comes from the data. and the way the model is trained physically seen so far, i anchor from india acquainted news outlet has also debuted and a anchor named fair, the trying to stay things agency. she was launched on a i and go back in 2018. do you sometimes get the impression that this is a trend that organizations are following without actually thinking through what

this means for journalism about issues such as oversight, for instance, or issues such as controlled? yeah, i mean in many ways there's lots of news organizations who actually make the effort to think about these things. i haven't told teams, i have people devoted to thinking about this and that has been happening for quite some time. so many of them are so decently prepared and of course is always a question if the vailable resources and not every organization can afford this. and there's also a huge differences between organizations over north of resources. plenty, well, you probably have the time money to tell a team or one person to think about this for a while. and so i think 3 problems and issues that micro about versus organizations in the south or even local news organizations where you don't have the same have abilities in the same resources. so i think that's, that's definitely one thing people do. think about it, of course, at the same time you have lots of extra mutations and you also,

and that's always been the case. you have to care much less about the problem suit and rethink through the implications base for the in jail isn't. but also for the news of lodge and stuff, always been a problem with new technology sort of arriving and using that has been the case of social media. and in many ways, we see that with a i as well as the best we can do as a community of journalist and research sort of our civil society access is to really demand thoughts. these things are being thought through not just mindlessly applies. and while those demands are being made, what should the, the, the, the, the common public, the audience is be aware of. i mean, how credible is this journalism that is emanating from these systems? what should people keep in mind? i think the most important thing for audiences, and this is where news organizations have an important role as to help them better understand what this technology is, what it is and where they can come to and also how it's being used in news process . i personally, i would not be in favor for instance,

and i news anchor not being labeled as such because there could be misleading to audiences with downstream consequences. and luckily many organizations take that's a place where they say, if we use i, i in some ways to shape our content to produce content, to the services we will let people know you will inform them about it through labels 3 by lines. and all these things, nothing of the right step to do, but in many ways they also have a duty. i think 2 is not just about how they are using it, but those might be in the future. and what signals is, because a lot of misunderstanding, a lot of lack of knowledge, which is no one's fault, really represents the sort of clear gap, not just news organizations, but researchers as well can mason sort of interjection, and hopefully help people to better understand what is happening and of what will be happening in a couple of years time. we leave it there,

but thank you so much for that context. felix simon, that communication researcher and doctoral student at oxford university. thanks so much, felix. and that does it for today because of cost more on our website and as ever you can follow us on facebook and twitter. i'm british manager and i'm not. i don't see you soon. by the gigantic waynes that swallow ships were long thought to be tall tales told by sailors and it's magic. it's the kind of magic. now the mystery has been solved. what turns waves into monsters?

40 to the answer to almost everything. mm. mm. what people have to say matters to us or me. that's why we listen to their stories reporter every weekend on d. w. ah, ah, unpredictable uncontrollable but so beautiful. waves of always fascinated us. but for me, what special about ways that they're so.