right, that's up next here on dw, on 6. if you want more news, there's always the website around the clock, www dot com. and i'm a little outside for me in the team here in building. thanks very much for watching . take care of the at could be green, green. all this blue p s. twine twine. lovely. red. definitely. pull just the yellow. if that's what you present on purple titles, very special. that's in the georgia. choose your favorite color. what secrets

live? you discover new adventures in 360 degrees and explore fascinating boats. heritage selling dw world heritage. 360 now i'm just gonna throw this on. what do you think? fraud can you? that's quite us, david. he's talking to an a i chip bought the space for me. it's more it more like talking to an actual human being. but what does this really mean for us as use us? how well do they really understand us? and what are the risks that we face? checkbox becoming increasingly lifelike. take the trip to peachy for oh, by open a i. it can visually identify surroundings and recognize our emotions. the book even has some understanding of comedies. what do you call. busy a giant pile of kitchen means i don't know what am yeah,

i'm saying that's perfectly the layer. it is just how useful is the newest version of chips repeat to you. really cheap. t for o is more human than ever. it can communicate emotionally and see its surroundings . the old. that's what i'm the as an on the put sense. the model can work with text, audio and video making, beaten to action, more natural. take a look. and i wrote one last thing, i love it. you can take a look at, of course i'd love to see what you wrote, show it to me whenever you're ready. okay, so this is what i wrote down. what do you see? oh, i see i love chat to p t. that's so sweet of you. according to it's make of g p t for all cannot only express emotions. it's and recognize your emotions to it looks like you're feeling pretty happy and tearful. hey, i, systems learn to recognize emotions by being trained on corresponding data. so for

example, by analyzing text, that links the smile to happiness, that leaves tiers to establish and a i is train to up to like in an optimization process to recreate this data. but what happens when a i adults more human like traits cut the model become indistinguishable from extra human. we tend to be very trusting of computers decision, even if they contradict own believes that's according to research from as early as 1999 and more recent studies suggest the same checkbooks already convincing. but if they speak to us like a human being what they might be even more so, and that could be a problem because a, i can like it can invent effects if at least to be true. this is called to see nation test. so that a i playing poco a video games have learned to bluff,

to produce of favorable outcomes. these systems are known to be lying or to making up things. and if something is very convincing septic and making things up, just by, by natural behavior, i'm way more inclined to believe it best if it was dry objective text or, or reasoning that can be much easier to assess slow, such as nonsense or not. i don't know about you, but in a i chip or deliberately lying seems pretty scary. it could trick us into giving away personal data like credit card details. and even if that doesn't happen, a chip both could be the future of online misinformation. that's why it's important that a i is trained with error free data and is able to understand the nuances of language correct. in india, there are 22 official languages and over 19000 dialects. plus some topics like sexual health aren't open, lead, talked about. and that's with checkbooks could come to the rescue. the may not full

of books can also tricky questions about sex and health in india. but it has to understand cultural nuances and different dialects. not what he's in the one of the more thing with degree diverse areas with um you know, hundreds of languages. but within the languages there are so many sub languages, some die leg line normally sign within those languages, what's 9, but in defense within both families. so how do we then reduce on says an attack board, that is mandukes dog equipment that are maybe not listed to a standardized version of our language in advance loft language models and able to check both to comprehends and engage in multiple languages. the l. m. used by the may not be a little f r trained to connect with use us and then native language is like manase and even the sing list can be mixed with english. what, why people don't think that they would say something like maybe a mom or a dog those days. oh really, you know, so there's different ways of theme to themes and so pleasing bad ones using slang.

so things that does a paste look as well as that could be to be for to, for that this f c use instead to having goals metaphors from 5, lots of the back end states of 8th is needed. so the both can understand a problem correctly. and doesn't get lost in translation or even worth create confusion with incorrect information or to move the nation. this requires intensive ground with the checks balances that the guy bought provide. and if it does it all, then we have a good i portion that we create. and then we take all the questions and the correct answers, and bc to in fact into the market. and that is actually the 20 to 25000 questions as we are fact checking. and i think not only do you with our staff, but also with medical professionals and doctors and has professionals have been in the field for 10 years. the nations is an issue that almost across the industry. we're trying to figure out how to do that along with other things like it will be

by is gender by is kind of good bias that comes up in the data because it isn't the was the reason why this gender bias would be does. because in 10 devices, back to the basics, how doesn't they? i chip bought, learn to talk to its environment and know what to say. the most important components of chip thoughts is a so called l l m a lots language model. that's an a i program that can recognize and interprets human language. lots language model is an ai program that recognizes and interprets human language. l. m, 's use machine learning to understand how characters words and sentences worked together. creating neural networks with many layers. processing large amounts of data based networks learned by recognizing pets, tons, for example, in language, then waiting and scoring them. this network is learning to recognize the number 9, the more frequently of pets. i'm. of course,

the more weight is the network places on it's in this way important peasants are reinforced for feedback and repeat the training l. m. sloan to understand and generate and natural language. the name of the foreman depends on the quality or training data. and the names of trained with data from the internet, a pre select the dates of fed as most suitable here. because arrows can be filtered out beforehand in the middle. the $25000000000.00 us dollars. we're invested in the development of genuine to a models such as chess, u, p t, in 2023 alone, a times more than the year before. and the market for a i software that can read your emotions is also growing. it doesn't even have to look you in the face. natural language processing software is being used by companies worldwide to analyze their customers and employees emotional states. the

study conducted by jonathan chrysler during the cold with ben demick shows just how well the systems work. the co founder and president of the motion a. i start the rest of tippett t analyze the red posts of 6500 nurses and doctors with this proprietary software. looking at close to full 1000000 comments over the course of full years, cried luck. you rated the data base for research as worldwide the goal advance the detection of psychological distress. and trauma is actually an incredibly wide variety of insights that are available from our assessments. we can understand, you know, simpler concepts like personality. but we can also understand psychological traits and understand how people change in the moment, depending on the context. what we found was that even early in the pandemic, even before the world health organization declared depend on that, we could see some of the early indicators of what would ultimately lead to burnett things like elevated cognitive load. all that is needed is everyday communication,

such as emails. it's like messages or social media posts. but why existing through communications, the software doesn't just look for obvious descriptions of feelings, but also for the so called function words such as pronouns, how we use these can review a lot about all mental states. people with depression, for example, use significantly more 1st person singular pronouns, like i and myself. according to a psychologist, people who are depressed, often highly self aware, also focusing this applies across languages and cultures and a i could help support the so called language of depression, water. the patterns that exist that we're analyzing are the same across cultures in context. it's less of what people are talking about and how language can provide insight into how they're thinking about themselves and the context their end and how they relate to the context or in into the other people around them. it's very much a social psychological phenomena that we're focused on,

and that exists regardless of the cultures or language is you're communicating. pay, i can not only analyze text, but also voice messages. the use cases, plentiful, analyzing how people interact with checkbooks, splunk companies insights into the use this mindset. the really interesting recent use of our for technology was to help with the vaccine uptake and under privilege communities. so what one of our customers was finding is that if they could understand how some of these populations think and how they consume information, they can do a better job calendar and the messaging around the benefits. and really the reasons why this action should be used to this community and improve vaccine uptake. after conducting interviews with community members, the language model was used to create a detailed community profile. this allow to custom tailor the message to a specific part of the population. addressing benefits and comes sons. a message

that comes from advertising and marketing, and it's becoming more and more common. so called triple bots. now also using a motion, a i to also use us customized travel tips. just one example is the voice a i m a p d f is currently being tested and based on your wishes and personalities, it can put together a trip from new york city. for example. first, you'll have to write a few pictures and answer some simple questions. the up then creates a kind of psychological profile of you using that as a foundation who then receive personalized travel recommendations. but what does that mean for our privacy when a companies use our data? metal for example, officially says that it uses our publicly accessible facebook and instagram photos and takes to train. it's a i what happens then? one of our programs that can analyze or facial expressions who uses this data and what for regulation can hardly keep up with innovation,

leaving us exposed to potential dangerous. for a i just find acceptance and responsible use that people need to know what they are doing. and how they are working work they are trained on understanding leads to better assessment a i will continue to understand off and all emotions. betsel that's affect it will make every day tasks easier. but we will have to stay vigilant with how we developed and protect our data in the wrong hands. our data could easily be used against us. i want to know what you think about all this. leave a comment and let us know. that's it from us. this week, so next time the conflict zone as special edition on the european election. we have

a lot of new challenges on any of these outside of the role which has influence inside. how can people trust the elections? they don't know whether that voting for a russian pop it, or somebody who's just t s. t is the most popular party among young voters of john tim sebastian investigates. the question is democracy under threats? conflicts the next on d w. mars. around. and then just to see a quote, you can make sure i'm not sure. go ahead. so here's your sisters, have been in a coma that was caused by persecution and flight for con dreams of another world wake up on the 1st few minutes on dw conflicts, crises,

every single connection mapped out shows the geophysical reality. the on the board is what makes things the way they are mapped out, navigating a changing world. now on youtube, the economy is in january this year, mocking the mother of 3 fascist activists in rome in 1978, almost 18 years after the defeated the right. that can be few single gestures accent about designed to send such a shilling warning. and you see pictures like that you what goes to your mind. we are reminded on the doctor's

dollars stuff this continents when fascists.