Years, African Musicians have a long history of being recognized at the Grammy Awards for many different styles of music. But this year, for the 1st time ever, they have their very own category. And its dominated by nigerian afro beats. The genre caught the attention of the Recording Academy after exceeding 15000000000 streams on the music platform. Spot a 5. 0, is there a level of quality that we look at, but also Critical Mass of the amount of recording is being created and consumed . We know just looking at data, looking at, uh, streaming 5 platforms and other ways of analyzing it, that this music is very popular, is a ton of really talented people, making it. Its a highly creative people like the kingsman who are getting ready to record their own original attributes. Music. The blend of different gera is, you know, hip hop. So pop, classical music, jazz. But let me blend that and now have um the african sounds surrounding that. You know, its, its, thats kind of half or beat so you know, a sound that news no borders, kristen salumi algae or a new york to 878 days in space encountering thats the back of a breaking mission for russian astronaut on a kind of 9 cause made history off to accumulating maybe 2 and a half years and all that that breaks the previous reco also set by a russian. The 59 year old has been to the International Space station. 5 times have spent 1000. 00 days and space when june, the 5th comes around and thats set for me, elizabeth per on them. So this will this, and you can always find more information on what are the stories were covering on our website that said out a 0 dot com to stay with us studio be is coming up next. Thank you for watching. The President Biden says once a 2 state solution for palestinians and israelis, but does anybody believe its doable . What this is real for . Im gonna say it back to us foreign policy. And what are the long term consequences for the region and the world . A quizzical look at us politics, the bottom line for the Artificial Intelligence, a is already transforming our world and is expected to bring some of the most profound changes in human history. Think of me is a friendly companion can provide helpful insights from the way we work to the way wars are. The very fabric of our societies. It seems set to bring huge opportunities, but other words it could lead to our own destruction. One or 2 extra central risk perspective, we face a potentially the most pressing need, some of the risk and im a journalist from the philippines through our investigations. I became a target of a harassment and this Information Campaign receiving thousands of Death Threats online. I received the Nobel Peace Prize in 2021 and acknowledgement of how did because it is for journalists to do our job. Today. I saw 1st hand the dangers of the tech and its threat to democracy. The design of the systems of social media prioritizes the spread of lies, least with anger in, in the special series of studio be on Artificial Intelligence. Ill be meeting some of the brightest minds working in the field today. My guest this week is Professor Mike port rich, and im working in a research for over 35 years in the oxford and at the prestigious island touring institute prolific officer. Hes written 9 books and over 400 scientific articles on this subject. So what exactly is Artificial Intelligence . How did we get here . And is it really a threat to our very existence the might, it is so good to see you. And you know, you had been studying Artificial Intelligence for 35 years. But something changed, right . It grew exponential, exponential right after november 2022, after check jeeps, p was launched. How did we get to where we are 1st . How do you define where we are . What does the science tell us . So obviously intelligence despite appearances is no. A new field has been studied very, very active, a since since since the 1950s. But the truth is the actually progress in a while. I was really glad seely slow until this century of computers in the past. Just one powerful enough and we didnt have the data and we are in the world of big data and a only is nothing with that data. You absolutely need data to, to, to train a i, to use the terminology. Then every time you upload a picture of yourselves to social media and you helpfully label it with your name or your kids, do what they are doing is providing training day training, the social social media companies. Thats literally what the role is in, in doing that. So you need data need lots and lots of computer power to be able to build Neural Networks that were big enough. So around about 2012 i was that tell us about alex net. So alex, it was a Computer Program in a, on a program to do basically image analysis. And it was entered into a competition and increasing this competition with judge to how well they could interpret pictures in in images. And the point about alex net was the in one year we source that change in capabilities and this call everybodys attention. Yeah. And it became clear at that point that we really were in that kind of a new era of a i. And that was the point i have to say that the Big Tech Companies noticed and started to get really, really, really interested. Can i ask you some sort of very geeky, you talked about Training Data, but machine learning, Artificial Intelligence, Neural Networks, large language models. How do these all fit together . Okay . The way to launch language models like chat g p to work is really bizarrely simple. Its just doing exactly what your smart phone does when you do or to complete. So if you open up your smart phone and you start sending a text message, say for example, i start sending a text message to my kids and i type have you, it will suggest completion so that the completion is, might be tied into a room. I hope the dog and i thought this might be the likeliest completion to that. So how is it doing that . Its been trained on all of the text messages. Ive sent my kids and learned that the likely as completion of have you. Is it going to be will to adult or, or tied to during . So try g p t is doing nothing more than that. The difference is the scale, right . Chat g p t is built with a i see for computers that run for months and costs tens of millions of dollars to be able to do that training. And the Training Data is basically, its not your small phone messages. Its all the digital Data Available in the world and the standard way that you build days is to stop by downloading the whole of the World Wide Web, right, the entire city of the world wide way. So wikipedia makes up just 3 percent of the Training Data for, for these large language models. So the scale of the data is incredible. There people will say that this will solve humanities worse problems like climate change, deep mind, which is behind Google Search now because they bought, it also does synthetic biology. Right. And maybe we can use fido planked and stuck and pull carbon out of the air like, can you give me your best Case Scenario . Okay, so i use quite remarkable that the discussion or die, either these to be extremely disturbing. Yeah, its going to be the end of humanity only extremely utopian. And theres no actually a lot between those to the reality is going to be between those to the id. I mean, i think a little mosque was on record recently a, suggesting that a, i was going to take a lot of jobs. That seems very unlikely to me. Know, and the lifetime of anybody in this room. And yeah, i will become a tool that most people use in that jobs, but its not going to replace people. I mean, for example, theyre going to be lots of applications of a i in education, which is going to be really wonderful. But what teachers do is a very human thing. Its not going to replace all of humanity and allow us to spend all lives writing poetry or whatever it is that we would do if we did, didnt ask for jobs, i think. So that scenario is, is, is extremely unlikely. The distorted p in scenarios ive been waiting, hotly discussed. People talk about exist central risk and that literally means the end of humanity that f a all i could become so powerful, but somehow the n c mt and see if it can program itself. If they can get the resources, it can continue do it without human supervision, right . So theres a scenario called the singularity. Yes. And its a beautiful scenario which makes the great Science Fiction, the idea only section. So the idea is at some point in the future, some point we dont know when a guy is going to be as small as we are. And at that point it can start to improve itself, can literally we, right, its product code. And then at that point, its smarter than we are and that improved a item then improve its code to get itself. And it just continues that process. And the fair is at that point, the eye is out of off and i saw this on black mirror. And actually of all of the content present sections as black mirror, i think hes obviously the by far the best. Its very thought provoking stuff. So is it so i mean in all of this discussion, ive never seen a single genuinely plausible scenario for existential threat. And it really has been discussed endlessly, we some very, very, very smart people, thinking about it. The biggest risks right now, the ai is a powerful tool that enables bad people to do bad things. Bad things that they couldnt previously have done that enables whole category of risk, snow, existential risks, but risk slight Cyber Security attacks, which would just know up in feasible. Yeah. Previously that i think focusing our attention on those issues, i think would be much more productive. That on Science Fiction before a weapon. So for right, like a i drones, which theyve used in ukraine are being used in moscow. There again are no boundaries set on this. And yet, the scientist with the profit motive are rushing ahead and we are like, pub loves dogs in real time. How can we protect ourselves in this . Cuz if you look it in the nobel lecture in 2021, i actually said that we had data that showed that were being insidiously manipulated to have so much of our data that it cuts into our emotions, information warfare, changes the way we feel changes the way we think and then the way we act. Electro integrity for example. I mean, i dont, i dont think its a coincidence that you have now according to the dumb, 72 percent of the world under authoritarian rule. Right . So these are some of the impact of this, mr. Scientist tell me, right . Because you dont have a profit motive, right . Youre studying the science. How do we ring them in . So i think this is 2 sets of issues. The 1st is mean were pretty confident, right . Now that social media, one of the unintended consequences, social media was a Mental Health crisis and teenagers, and we didnt see that coming right. But thats just one of the unintended consequences. I think what youre saying is, what are the unintended consequences of a im going to be. So for example, what if we end up with some future launch language model, which just completely inadvertently makes us more aggressive or more depressive, for example. And what seemed packed would that have globe, like, for example, a widely used a i tool that made us more aggressive, might lead to more conflict with, isnt that happening since they took all of the big data, the unstructured big data of social media full of fear anger hate great, isnt that happening now . Okay. As we already mentioned, the way this technology is configured is youve done like the hold of the World Wide Web. You dont have to look very hard on the World Wide Web to find all sorts of unpleasant s. I mean, if you go on and theres some social media platforms, they have types of unpleasantness that we could scarcely imagine. Right, right. So all of that has been observed by a large language model. Then its a saving cauldron of unpleasantness. Now, i think genuinely and irresponsible a uh, companies have no intention whatsoever of unleashing that on the world. So what we do is that building called rails. And so they try to intercept query is that uh, how do i build a pipe bump . Yeah, that will try to intercept such a good area. And also they will look at the outputs of a large language model and try to intercept anything which is inadvertently come out with which is it appropriate at the moment i just called rails, i think all the technological equivalent of gaffer type is it just being that being plastered homes in the spring. Yeah, exactly. Theres no deep fixes. Yeah. That one of the worries is if this technology is owned by a small group of actors. And so in developing this Technology Behind closed doors, we dont get to see the Training Data. So right, you have no idea what this has been trained on you and your public figure. That would have been a great deal of content about you. And some of it would have been very nice. Thats a safe bet. So this is, i think, a real concern, and this issue of transparency i think is, is, is really a concern which needs to be taken very good. Talked about bar grails, right. Theres no incentive for them to put gar grills in. I mean, they, the only incentive is that they wont be attacked by people, right . Its a reputational thing, but if they can get by without it, they have us, they have what social media, we still havent done anything. Well, i think here we are in a situation which is very old, which we go a i, which is gone viral. Its the 1st last language models of the 1st general purpose and im choosing those words very carefully, right . General purpose a i tools that have reached the mass market and the very powerful. Yeah. And the Tech Companies see impious. They seem nice and, and pause and they want to, they want to state the claim on those. And plus, they want to be those and pies, they want to be the google, they want to be that they want to be the amazon off the january to a wells. And a very big risk is that what they doing to try to get an advantage on the competitive. You rush ahead with this Technology Without thinking about, for example, whether its really fit for prime time. And that really is a worry. But these worries are not, you know, the all unknown, i mean the, the u. K. Government company and an International Ai safety summit. And i have to tell you that was that there was some skepticism about what it was going to achieve. But actually the, the debate was, was a sensible debate and it got it on the international agenda. So i think whats going to be challenging is the extent to which government can really hold the Rich Companies in the world to account. And the irony, of course, is that if they get it out to all of us, they get all of our data. We train their large language models, they gain more power, even as the very nation states that are going to try to put regulations in place to control them, lose power because the technology is already impacting society all around the world, right. Its a tough one. I guess, you know, and i have a big picture of this as you can tell because having been attacked and to hear them say, you know, well, we didnt intend, that doesnt really matter what the intent was. And so how do we, what can we do right now . Right, its moving to slowly government move at the pace of years, while the tech evolves in every 2 weeks. Ad Job Development means theyre rolling out code every 2 weeks. Right . So is there anything anyone watching can do . Well, i think there is, there is concretely something we can do. So were heading to elections in the u. K. And us, us and the, one of the very prominent risks with this technology is the possibility of industrialized during the production of this information and missed information on a massive scale on an unprecedented scale. Yeah. And personalizing it down to the level of individuals. Yes. So the a, i can look up my social media feed and pick up on the sentiments i express in my social media feed, pick up on my political stands, which is going to be in place that within sometimes explicit within my social media fee and then feed me personally tailored very high quality misinformation the Sentiment Analysis exists to do that. The genuine survey only makes it possible to do that. And im and the cost of launching a disInformation Campaign in an election because of january today. Ive come down massively. Yeah, i must be on a store, people in the world with a huge interest in disrupting elections in the us. So the u. K. Or india and so on. And it could be people just with an interesting vandalizing the process or it could be state level actors that really want to disrupt whats going on. So what can we do concretely . I think we absolutely need to be a look to that issue. I think trusted new sources are going to become so valuable. The difficulty with that goal is, is that we end up in a world where, you know, were all completely paranoid and dont believe anything but trusted new sources. I think youre going to be essential and understanding how we can be that minute. The latest i think is really, really important. There is so much more we can talk about this because in 2024, one in 3 people around the world are going to vote. And this is the Tipping Point for both Electoral Systems are democracies, but were getting to the 2 and a. So let me toss it to you. The gentleman in the back was the 1st hand up leading him from what you were saying, michael. About news and trusted news. Give them t