inaudible thank you, senators, today's meeting is historic, _ thank you, senators, today's meeting is historic, i. — thank you, senators, today's meeting is historic, i, someone who has founded — is historic, i, someone who has founded al _ is historic, i, someone who has founded ai companies and someone who genuinely— founded ai companies and someone who genuinely loves ai but is increasingly worried. there are benefits — increasingly worried. there are benefits but we do not yet know whether— benefits but we do not yet know whether they will outweigh the risks — whether they will outweigh the risks. fundamentally, these systems are going _ risks. fundamentally, these systems are going to be destabilising, they can and _ are going to be destabilising, they can and will create persuasive lies at a scale — can and will create persuasive lies at a scale humanity has never seen before _ at a scale humanity has never seen before. outsiders will use them to affect— before. outsiders will use them to affect elections, insiders to manipulate markets and political systems — manipulate markets and political systems. democracy itself is threatened. chat bots will clandestinely shape opinion, potentially exceeding what social media _ potentially exceeding what social media can do. dataset choices ai companies — media can do. dataset choices ai companies use will have enormous unseen _ companies use will have enormous unseen influence. those who choose the data _ unseen influence. those who choose the data will make the rules, shaping _ the data will make the rules, shaping society and subtle but powerful ways. there are other risks, a law professor— there are other risks, a law professor was accused by chat bot of sexual—

professor was accused by chat bot of sexual harassment, and tree. appointed to a washington post article _ appointed to a washington post article that didn't exist. the more that happens, the more anybody can deny anything, as one prominent lawyer— deny anything, as one prominent lawyer on— deny anything, as one prominent lawyer on friday, defendants are starting — lawyer on friday, defendants are starting to claim complaints and making — starting to claim complaints and making up legitimate evidence. it can contribute to the undermining of democracy — can contribute to the undermining of democracy. medical device could have serious _ democracy. medical device could have serious consequences, and open source _ serious consequences, and open source large language model seems to have played a role in a person's decision— have played a role in a person's decision to _ have played a role in a person's decision to take their own life. it asked _ decision to take their own life. it asked the — decision to take their own life. it asked the human, if you wanted to die, why— asked the human, if you wanted to die, why didn't you do it earlier? and then — die, why didn't you do it earlier? and then up _ die, why didn't you do it earlier? and then up with, what you're thinking — and then up with, what you're thinking of me when you overdosed? without— thinking of me when you overdosed? without referring the human to help that was _ without referring the human to help that was of the sea needed. another system _ that was of the sea needed. another system rushed out to meet available to children. — system rushed out to meet available to children, told a person closing as a 13—year—old told parents how to lie about _ as a 13—year—old told parents how to lie about a _ as a 13—year—old told parents how to lie about a trip with a 31—year—old man _ lie about a trip with a 31—year—old man a_ lie about a trip with a 31—year—old man. a month after gpt four was

released. — man. a month after gpt four was released, plug—ins were released which _ released, plug—ins were released which lead others to develop something called auto gpt with direct— something called auto gpt with direct access to the internet, the ability— direct access to the internet, the ability to— direct access to the internet, the ability to write source code and increased — ability to write source code and increased powers of automation. this may have _ increased powers of automation. this may have drastic and difficult to predict — may have drastic and difficult to predict a — may have drastic and difficult to predict a security consequences. what _ predict a security consequences. what criminals are going to do is create _ what criminals are going to do is create cattle people. it is hard to envision — create cattle people. it is hard to envision the consequences of that. —— counterfeit people. powerful, recktess — —— counterfeit people. powerful, reckless and difficult to control. we agree — reckless and difficult to control. we agree on the values we would like for our— we agree on the values we would like for our ai _ we agree on the values we would like for our ai systems to honour. we want _ for our ai systems to honour. we want our— for our ai systems to honour. we want our systems to be transparent, to protect — want our systems to be transparent, to protect our privacy, to be free of bias— to protect our privacy, to be free of bias and — to protect our privacy, to be free of bias and to be safe. current systems — of bias and to be safe. current systems are not in line with these values _ systems are not in line with these values. they are not transparent, do not protect— values. they are not transparent, do not protect our privacy and continue to perpetuate bias. even their makers — to perpetuate bias. even their makers do not entirely understand how they— makers do not entirely understand how they work. most of we cannot remotety — how they work. most of we cannot remotely guarantee they are safe. hope _ remotely guarantee they are safe. hope is— remotely guarantee they are safe. hope is not enough. studio: we are going to come away

from those events. it is a busy afternoon, we are juggling two from those events. it is a busy afternoon, we arejuggling two main stories. we will return and here the questioning when it begins. i want to come away from washington because we want to return to ukraine where the capital keep has been targeted by an intense wave of russian air attacks. ukraine said all 80 missiles were shot down, footage showing their defences in action in the city of night. russia says its attacks which use drones and missiles had hit all of its targets. moscow stepped up its campaign in recent weeks ahead of expected ukrainian offensive. let's head to kyiv. live now to kyiv — and petro poroshenko, who served as the president of ukraine from 2014 to 2019. we saw a large number of those missiles targeting the capital of night, you must be relieved that your defence system is held up and

worked last night. first your defence system is held up and worked last night.— worked last night. first of all i want to thank _ worked last night. first of all i want to thank you _ worked last night. first of all i want to thank you and - worked last night. first of all i want to thank you and the - worked last night. first of all i i want to thank you and the british people, the british government and united kingdom for the great leadership, demonstrating how you support democracy, freedom, support ukraine in ourfighting against russian protein. point number two, you are right that this morning, by the work of russian missiles, the ukrainian defence system, i am proud that for me have an empty treasury, no money in reserve, it was my work to create in your defence of kyiv. i am proud that putin planned to attack and destroy kyiv government and everything, it was halted by the

ukrainian defence soldiers. but now they are much stronger because of they are much stronger because of the supply of our partners, including great britain, e united states, germany, france. we demonstrate the improvement but the nato air defence, nato standard weapons is dominating and russia even using their most modern missile cannot do anything with the nato weapons. cannot do anything with the nato wea ons. ., , y cannot do anything with the nato weaons. ., , y y cannot do anything with the nato weaons. ., , , , . . . weapons. previously they had claimed these missiles _ weapons. previously they had claimed these missiles couldn't _ weapons. previously they had claimed these missiles couldn't be _ weapons. previously they had claimed these missiles couldn't be shot - weapons. previously they had claimed these missiles couldn't be shot down | these missiles couldn't be shot down by defence systems. how did ukrainian authorities managed to do it? taste ukrainian authorities managed to do it? . , , , ukrainian authorities managed to do it? , _ ., , ., ukrainian authorities managed to do in , _ ukrainian authorities managed to do it? , _ it? we simply demonstrated it was a re ort it? we simply demonstrated it was a report of- -- — it? we simply demonstrated it was a report of. .. i know— it? we simply demonstrated it was a report of... i know that _ it? we simply demonstrated it was a report of... i know that the - it? we simply demonstrated it was a report of... i know that the six - report of... i know that the six missiles which wheresoever chased by

putin that this is absolutely no defence against that, was hit. let's imagine that we do it or so with the participation of the american system. but this is not only because of the patriot, but because of the level of professionalism of ukrainian armed forces. you mention to the patriot — ukrainian armed forces. you mention to the patriot missile _ ukrainian armed forces. you mention to the patriot missile system, - ukrainian armed forces. you mention to the patriot missile system, the - to the patriot missile system, the american system. moscow says they took at the patriot defence systems. is that true propaganda?— is that true propaganda? please never trust _ is that true propaganda? please never trust russian _ is that true propaganda? please | never trust russian propaganda. please don't be afraid to defeat putin. please don't be afraid of ukrainejoining nato. please don't be afraid of ourjoint victory on

russia. this is the convulsion of putin and he uses the propaganda as a separate weapon. we do not trust put in and that is why this weapon is useless. in put in and that is why this weapon is useless. , ., put in and that is why this weapon is useless— is useless. in terms of the patriot s stem, is useless. in terms of the patriot system. in _ is useless. in terms of the patriot system, in terms _ is useless. in terms of the patriot system, in terms of _ is useless. in terms of the patriot system, in terms of clarity - is useless. in terms of the patriot system, in terms of clarity on - is useless. in terms of the patriot system, in terms of clarity on the i system, in terms of clarity on the numbers, how many patriot systems are now on the ground, in theatre, working? it are now on the ground, in theatre, workin: ? , , ., working? it is definitely not enou~h, working? it is definitely not enough. point _ working? it is definitely not enough, point number - working? it is definitely noti enough, point number one. working? it is definitely not - enough, point number one. but it is in the professional hands of ukrainian soldiers, it is sufficient. in ukrainian soldiers, it is sufficient.— ukrainian soldiers, it is sufficient. , ., , sufficient. in terms of kyiv, it has been targeted — sufficient. in terms of kyiv, it has been targeted eight _ sufficient. in terms of kyiv, it has been targeted eight times - sufficient. in terms of kyiv, it has been targeted eight times this i been targeted eight times this month. a real step up in terms of the russian offensive. does it feel to you, you are out there, like a step change after relatives come over the last 50 days or so? this is

the biggest — over the last 50 days or so? this is the biggest attack _ over the last 50 days or so? this is the biggest attack in _ over the last 50 days or so? this is the biggest attack in may - over the last 50 days or so? this is the biggest attack in may be - over the last 50 days or so? this is the biggest attack in may be the i the biggest attack in may be the last six or seven months. definitely, maybe from april 2022, when kyiv was arms encircled and me, together with my team, together with the battalion, with weapons in their hands, protected key. keith. but yes, ukrainian air defence demonstrate it a miracle. but again we need to stop russian dominance in ukrainian air. we need two things, airi6 and long—range ukrainian air. we need two things, air 16 and long—range missile. from all of my heart, i want to thank the british government and british people for delivering, for us, there are a storm long—range missile and even more, i can confirm you, that some of these missiles is already efficiently used against russian.

those missiles have been used, you mention the f—i6s. president zelensky was here yesterday, talking about the coalition ofjets he wants. you just repeated it. do you feel they are in ukraine that you are getting close to that moment from european countries delivering that coalition ofjets? i from european countries delivering that coalition of jets?— that coalition of jets? i was the first president _ that coalition of jets? i was the first president of _ that coalition of jets? i was the first president of ukraine, i that coalition of jets? i was the i first president of ukraine, fighting forfive years and first president of ukraine, fighting for five years and it is impossible to beat the president of the country in a state of war, not to be an optimist. we have first a coalition of the tanks and our personal carrier, then we have our coalition, and now a coalition of long—range missile thanks to great britain. next coalition would be the coalition of the jet fighter and i

think you do important steps, you declare that you are ready to launch a training programme for ukrainian pilots and the french president emmanuel macron do the same, inviting ukrainian pilots. this is exactly the right time, right things to be done. min exactly the right time, right things to be done. ~ exactly the right time, right things to be done-— exactly the right time, right things to be done. ~ u, ., to be done. will it come in time for the s-urin to be done. will it come in time for the spring offensive, _ to be done. will it come in time for the spring offensive, and _ to be done. will it come in time for the spring offensive, and for i to be done. will it come in time for the spring offensive, and for that i the spring offensive, and for that offensive, in the spring, in the summer, what is the realistic target? russia occupies about a fifth of ukrainian territory. how much of that do you think you can reclaim in the coming months if things go well?— things go well? first of all, the sto and things go well? first of all, the stop and destroy _ things go well? first of all, the stop and destroy russian i things go well? first of all, the | stop and destroy russian winter offensive operation after the mobilisation in russia. they claim they will attack and for ten months,

they will attack and for ten months, the ukrainian fortress demonstrated that russia was stopped by ukrainian forces. i can tell you, this is not only the counter offensive approach, ukrainian offensive operation... ukrainian offensive operation... ukrainian membership in nato. this is the only instrument of the sustainable security system on the continent, and i have no doubt ukraine are not going to trade our serenity, our territorial integrity and our soil. serenity, our territorial integrity and our soil-— and our soil. that was what president _ and our soil. that was what president zelensky - and our soil. that was what president zelensky were i and our soil. that was what i president zelensky were single yesterday. in terms of more pressure being put on moscow, do you worry the sort of tactics you might see if that pressure increases over the next few months? and you make territorial gains, in terms of what

you might get in response from moscow and what does it feel like being in the second year of this war, just on a human level? putin do maximum he — war, just on a human level? putin do maximum he can. _ war, just on a human level? putin do maximum he can. if— war, just on a human level? putin do maximum he can. if you _ war, just on a human level? putin do maximum he can. if you have - war, just on a human level? putin do maximum he can. if you have any. maximum he can. if you have any instruments in his hands, he already use it. second, ukraine and ukrainian armed forces ready for offensive operation now. we have, as we said, more than 95% of the armed personnel, we have a sufficient number of tanks waiting after these weak negotiations, and we think any moment, at any part of the battlefield, ukraine is ready to

start and offensive operation. in russia definitely know how we do that. that would not be easy, that would be disastrous. we understand how difficult is our aim but we are fighting for our soil, we are for our people and at the end of the day, we are fighting for you. thank ou so day, we are fighting for you. thank you so much _ day, we are fighting for you. thank you so much for— day, we are fighting for you. thank you so much forjoining _ day, we are fighting for you. thank you so much forjoining us - day, we are fighting for you. thank you so much forjoining us live i day, we are fighting for you. thank| you so much forjoining us live here on bbc news, live from keith. —— kyiv. a very busy afternoon, we broke away to have that important interview from proceedings on capitol hill. let's return, the questioning has started in terms of those tech bosses so let's put the microphones up. we heard a little earlier from microphones up. we heard a little earlierfrom sam microphones up. we heard a little earlier from sam altman,

microphones up. we heard a little earlierfrom sam altman, singh regulation is critical. i earlier from sam altman, singh regulation is critical.— regulation is critical. i don't think anybody _ regulation is critical. i don't think anybody knows i regulation is critical. i don't think anybody knows the i regulation is critical. i don't i think anybody knows the answer regulation is critical. i don't - think anybody knows the answer to that question, in the long run, so called artificial general intelligence really will replace a large fraction of human jobs. we are not that close to artificial general intelligence despite all the media hype. i would say what we have right now is a small sampling of ai that we were built. in 20 years people will laugh at this, as i think it was senator whole or a senator durban made the example about cell phones. when we look back at the ai of the day 20 years ago, we thought that stuff was unreliable. they couldn't do planning, its reasoning ability is quite limited. but van may get to artificial general intelligence, maybe 50 years, that will have profound effects on labour and there is no way around that. last, i don't know if i am allowed to do this, but sam's worst fear is

not employment, he never told us what his worst fear is and i think it is to remained find out.- it is to remained find out. thank ou, i it is to remained find out. thank you. i am _ it is to remained find out. thank you. i am going _ it is to remained find out. thank you. i am going to _ it is to remained find out. thank you, i am going to ask - it is to remained find out. thank you, i am going to ask mr- it is to remained find out. thank. you, i am going to ask mr altman it is to remained find out. thank i you, i am going to ask mr altman if he cares _ you, i am going to ask mr altman if he cares to— you, i am going to ask mr altman if he cares to respond. we you, i am going to ask mr altman if he cares to respond.— he cares to respond. we have tried to be very — he cares to respond. we have tried to be very clear— he cares to respond. we have tried to be very clear about _ he cares to respond. we have tried to be very clear about the - he cares to respond. we have tried l to be very clear about the magnitude of the _ to be very clear about the magnitude of the risks— to be very clear about the magnitude of the risks here. _ to be very clear about the magnitude of the risks here. i— to be very clear about the magnitude of the risks here. i think— to be very clear about the magnitude of the risks here. i thinkjobs - to be very clear about the magnitude of the risks here. i thinkjobs and i of the risks here. i thinkjobs and employment _ of the risks here. i thinkjobs and employment and _ of the risks here. i thinkjobs and employment and what _ of the risks here. i thinkjobs and employment and what we're i of the risks here. i thinkjobs and i employment and what we're going to do with— employment and what we're going to do with a _ employment and what we're going to do with a time — employment and what we're going to do with a time when _ employment and what we're going to do with a time when it _ employment and what we're going to do with a time when it matters. i- do with a time when it matters. i agree _ do with a time when it matters. i agree that — do with a time when it matters. i agree that when _ do with a time when it matters. i agree that when we _ do with a time when it matters. i agree that when we get - do with a time when it matters. i agree that when we get to - do with a time when it matters. i agree that when we get to a i do with a time when it matters. i- agree that when we get to a powerful system _ agree that when we get to a powerful system is _ agree that when we get to a powerful system is the — agree that when we get to a powerful system is the landscape _ agree that when we get to a powerful system is the landscape for— agree that when we get to a powerful system is the landscape for change, i system is the landscape for change, i am system is the landscape for change, i am just— system is the landscape for change, i am just more — system is the landscape for change, i am just more optimistic— system is the landscape for change, i am just more optimistic that - system is the landscape for change, i am just more optimistic that we i i am just more optimistic that we are incredibly— i am just more optimistic that we are incredibly creative _ i am just more optimistic that we are incredibly creative and - i am just more optimistic that we are incredibly creative and we i i am just more optimistic that we i are incredibly creative and we find new things — are incredibly creative and we find new things to _ are incredibly creative and we find new things to do _ are incredibly creative and we find new things to do with _ are incredibly creative and we find new things to do with tools. i are incredibly creative and we find new things to do with tools. my i new things to do with tools. my worst— new things to do with tools. my worst fears _ new things to do with tools. my worst fears are _ new things to do with tools. my worst fears are that _ new things to do with tools. my worst fears are that because i worst fears are that because significant~~~ _ worst fears are that because significant... the, _ worst fears are that because significant... the, the - worst fears are that because significant... the, the field, | worst fears are that because i significant... the, the field, the technology— significant... the, the field, the technology industry, _ significant... the, the field, the technology industry, because i technology industry, because significant _ technology industry, because significant harm _ technology industry, because significant harm to _ technology industry, because significant harm to wild. i technology industry, because significant harm to wild. thatj technology industry, because i significant harm to wild. that can come _ significant harm to wild. that can come in — significant harm to wild. that can come in a — significant harm to wild. that can come in a tot— significant harm to wild. that can come in a lot of— significant harm to wild. that can come in a lot of different - significant harm to wild. that can come in a lot of different ways, l significant harm to wild. that canj come in a lot of different ways, it is why— come in a lot of different ways, it is why we — come in a lot of different ways, it is why we started _ come in a lot of different ways, it is why we started the _ come in a lot of different ways, it is why we started the company. l come in a lot of different ways, it| is why we started the company. it come in a lot of different ways, it i is why we started the company. it is why we _ is why we started the company. it is why we are — is why we started the company. it is why we are here _ is why we started the company. it is why we are here today _ is why we started the company. it is why we are here today and - is why we started the company. it is why we are here today and have i is why we started the company. it is. why we are here today and have been here in _ why we are here today and have been here in the _ why we are here today and have been here in the past _ why we are here today and have been here in the past. if— why we are here today and have been here in the past. if this _ why we are here today and have been here in the past. if this technology. here in the past. if this technology .oes here in the past. if this technology goes wrong — here in the past. if this technology goes wrong it _ here in the past. if this technology goes wrong it can _ here in the past. if this technology goes wrong it can go _ here in the past. if this technology goes wrong it can go quite - here in the past. if this technology goes wrong it can go quite wrong i here in the past. if this technology. goes wrong it can go quite wrong and we want _

goes wrong it can go quite wrong and we want to— goes wrong it can go quite wrong and we want to be — goes wrong it can go quite wrong and we want to be vocal— goes wrong it can go quite wrong and we want to be vocal about _ goes wrong it can go quite wrong and we want to be vocal about that - goes wrong it can go quite wrong and we want to be vocal about that and i we want to be vocal about that and work _ we want to be vocal about that and work with — we want to be vocal about that and work with the — we want to be vocal about that and work with the government - we want to be vocal about that and work with the government to i we want to be vocal about that and i work with the government to prevent that from _ work with the government to prevent that from happening, _ work with the government to prevent that from happening, but _ work with the government to prevent that from happening, but we - work with the government to prevent that from happening, but we try i work with the government to prevent that from happening, but we try to i that from happening, but we try to be very— that from happening, but we try to be very clear — that from happening, but we try to be very clear on _ that from happening, but we try to be very clear on it, _ that from happening, but we try to be very clear on it, but— that from happening, but we try to be very clear on it, but what - that from happening, but we try to be very clear on it, but what the i be very clear on it, but what the downside — be very clear on it, but what the downside case _ be very clear on it, but what the downside case is _ be very clear on it, but what the downside case is and _ be very clear on it, but what the downside case is and what i be very clear on it, but what the downside case is and what theyl be very clear on it, but what the i downside case is and what they work we have _ downside case is and what they work we have to _ downside case is and what they work we have to do — downside case is and what they work we have to do is _ downside case is and what they work we have to do is to— downside case is and what they work we have to do is to mitigate - downside case is and what they work we have to do is to mitigate that. i we have to do is to mitigate that. our hope — we have to do is to mitigate that. our hope is — we have to do is to mitigate that. our hope is that _ we have to do is to mitigate that. our hope is that the _ we have to do is to mitigate that. our hope is that the rest - we have to do is to mitigate that. our hope is that the rest of- we have to do is to mitigate that. our hope is that the rest of the i our hope is that the rest of the industry will follow the example that you and ibm, miss montgomery have set by coming today and meeting with us, as you have done, privately, and helping to guide what we are going to do so we can target the harms and avoid unintended consequences to the good. this thank you, thank you to the witnesses for being here. i you, thank you to the witnesses for being here-— being here. i think you grew up in saint lewis- _ being here. i think you grew up in saint lewis. it _ being here. i think you grew up in saint lewis. it is _ being here. i think you grew up in saint lewis. it is great _ being here. i think you grew up in saint lewis. it is great to - being here. i think you grew up in saint lewis. it is great to see i being here. i think you grew up in saint lewis. it is great to see a i saint lewis. it is great to see a fellow— saint lewis. it is great to see a fellow missourian here. i want are underlined — fellow missourian here. i want are underlined in the record, missouri is a great — underlined in the record, missouri is a great place. let me ask you. i

will preface — is a great place. let me ask you. i will preface this by saying my question— will preface this by saying my question here are a question about getting _ question here are a question about getting our— question here are a question about getting our heads around what this generative ai, the large language models, — generative ai, the large language models, can do. i am trying to understand its capacities and significance. i am looking at a paper— significance. i am looking at a paper here entitled large language models _ paper here entitled large language models trained on media diets can predict _ models trained on media diets can predict public opinion. this was posted — predict public opinion. this was posted about a month ago, and the conclusion, — posted about a month ago, and the conclusion, this work was done at mit, _ conclusion, this work was done at mit. the — conclusion, this work was done at mit, the conclusion is that large lun- mit, the conclusion is that large lung which models can indeed predict public— lung which models can indeed predict public opinion and they go through and modelled by this is the case. they— and modelled by this is the case. they conclude ultimately that an ai system _ they conclude ultimately that an ai system can predict human survey responses — system can predict human survey responses by adapting a pre—trained tanguage _ responses by adapting a pre—trained language model to subpopulation specific— language model to subpopulation specific media diets. you can feed the model— specific media diets. you can feed the model sets of media inputs and it can _ the model sets of media inputs and it can predict with remarkable

accuracy— it can predict with remarkable accuracy what people's opinions will be. i accuracy what people's opinions will be iwant— accuracy what people's opinions will be. i want to think about this in the context of elections. these large _ the context of elections. these large lung which models can even now, _ large lung which models can even now, based on the information we put into them, _ now, based on the information we put into them, quite accurately predict public— into them, quite accurately predict public opinion, ahead of time, predicts— public opinion, ahead of time, predicts before you even ask the public— predicts before you even ask the public these questions, what will happen— public these questions, what will happen when entities, whether it is corporate _ happen when entities, whether it is corporate or governmental entities or campaigns of foreign actors, take this survey— or campaigns of foreign actors, take this survey information, these predictions about public opinion and fine tune _ predictions about public opinion and fine tune strategies to enlist that certain— fine tune strategies to enlist that certain responses and behavioural responses? —— illicit behavioural responses — responses? —— illicit behavioural responses. we know about the effect of something as prosaic, as google search, _ of something as prosaic, as google search, the — of something as prosaic, as google search, the effect this has on voters — search, the effect this has on voters and elections, especially undecided voters who might try to -et undecided voters who might try to get information from google search

and what _ get information from google search and what enormous effect the rankings— and what enormous effect the rankings of google search has on an undecided _ rankings of google search has on an undecided voter. this is orders of magnitude. — undecided voter. this is orders of magnitude, powerfuland undecided voter. this is orders of magnitude, powerful and significant and directed. maybe you can help me understand _ and directed. maybe you can help me understand what some of the significance of this is. should we be concerned about models, large tanguage — be concerned about models, large language models, that can predict survey— language models, that can predict survey opinion and can help organisations fine tune strategies to itticit _ organisations fine tune strategies to illicit behaviours from voters? shoutd _ to illicit behaviours from voters? should we — to illicit behaviours from voters? should we be worried about this for our elections? it is should we be worried about this for our elections?— our elections? it is one of my areas of greatest — our elections? it is one of my areas of greatest concerns, _ our elections? it is one of my areas of greatest concerns, the _ our elections? it is one of my areas of greatest concerns, the more i of greatest concerns, the more general ability of these models to manipulate, persuade and provide one—on—one interactive disinformation. that is a broader version of what you are talking about but given that we are going to face an election next year and these models are getting better, i think this is a significant area of concern. there is a lot of policies that companies can voluntarily adopt and i am happy to talk about what we

do there. i do think some regulation would be quite wise on this topic, someone mentioned earlier, something we agree with, people need to know if they are talking to an ai, if content is generated or not. it is a great thing to do, to make that clear. we also will need rules, guidelines about what is expected in terms of disclosure from a company providing a model that could have these sorts of abilities that we talked about. i am nervous about it, i think people are able to adapt quite quickly when photoshop came onto the scene a long time ago. for a while people were really quite fooled by it, photoshopped images, and then developed an understanding that images might be photoshopped. this will be like that but on steroids. the interactivity, the

ability to really model predict humans, as he talked about, is going to require a combination of companies doing the right thing, regulation. companies doing the right thing, reuulation. ~ ., i. companies doing the right thing, reuulation. ~ ., ~' companies doing the right thing, reuulation. ~ ., ~ ., regulation. would you like to address this? _ regulation. would you like to address this? in _ regulation. would you like to address this? in the - regulation. would you like to| address this? in the appendix regulation. would you like to . address this? in the appendix to regulation. would you like to - address this? in the appendix to my remarks i have _ address this? in the appendix to my remarks i have two _ address this? in the appendix to my remarks i have two papers - address this? in the appendix to my remarks i have two papers to - address this? in the appendix to my remarks i have two papers to make | address this? in the appendix to my. remarks i have two papers to make it even more _ remarks i have two papers to make it even more concerned. _ remarks i have two papers to make it even more concerned. one _ remarks i have two papers to make it even more concerned. one is - remarks i have two papers to make it even more concerned. one is in - remarks i have two papers to make it even more concerned. one is in the i even more concerned. one is in the wall street — even more concerned. one is in the wall street journal— even more concerned. one is in the wall street journal saying _ even more concerned. one is in the wall street journal saying help, - even more concerned. one is in the wall street journal saying help, my| wall street journal saying help, my political _ wall street journal saying help, my political beliefs _ wall street journal saying help, my political beliefs were _ wall street journal saying help, my political beliefs were altered - wall street journal saying help, my political beliefs were altered by - wall street journal saying help, my political beliefs were altered by a l political beliefs were altered by a chat bot — political beliefs were altered by a chat hot the _ political beliefs were altered by a chat bot. the scenario— political beliefs were altered by a chat bot. the scenario you - political beliefs were altered by a chat bot. the scenario you raisedi political beliefs were altered by a . chat bot. the scenario you raised as we might— chat bot. the scenario you raised as we might basically— chat bot. the scenario you raised as we might basically observe - chat bot. the scenario you raised as we might basically observe people i we might basically observe people and use _ we might basically observe people and use surveys— we might basically observe people and use surveys to _ we might basically observe people and use surveys to figure - we might basically observe people and use surveys to figure out - we might basically observe people and use surveys to figure out what they are _ and use surveys to figure out what they are saying. _ and use surveys to figure out what they are saying, but _ and use surveys to figure out what they are saying, but some - they are saying, but some acknowledged _ they are saying, but some acknowledged the - they are saying, but some acknowledged the risk - they are saying, but some acknowledged the risk is i they are saying, but some - acknowledged the risk is worse, the systems— acknowledged the risk is worse, the systems will — acknowledged the risk is worse, the systems will directly, _ acknowledged the risk is worse, the systems will directly, maybe - acknowledged the risk is worse, the systems will directly, maybe not. systems will directly, maybe not even _ systems will directly, maybe not even intentionally, _ systems will directly, maybe not even intentionally, manipulate l even intentionally, manipulate people — even intentionally, manipulate peorrte and _ even intentionally, manipulate people and that _ even intentionally, manipulate people and that was _ even intentionally, manipulate people and that was the - even intentionally, manipulate people and that was the thrust even intentionally, manipulate . people and that was the thrust of the walt— people and that was the thrust of the wall street _ people and that was the thrust of the wall streetjournal_ people and that was the thrust of the wall street journal article. it| the wall street journal article. it links_ the wall street journal article. it links to — the wall street journal article. it links to an — the wall street journal article. it links to an article _ the wall street journal article. it links to an article i— the wall street journal article. it links to an article i have - the wall street journal article. it links to an article i have linked, i links to an article i have linked, not yet — links to an article i have linked, not yet peer—reviewed, - links to an article i have linked, not yet peer—reviewed, says . not yet peer—reviewed, says interacting _ not yet peer—reviewed, says interacting with _ not yet peer—reviewed, says interacting with models - not yet peer—reviewed, says - interacting with models changes users _ interacting with models changes users views _ interacting with models changes users views. this _ interacting with models changes users views. this comes - interacting with models changes

users views. this comes back. interacting with models changes users views. this comes back toi users views. this comes back to data _ users views. this comes back to data one of— users views. this comes back to data. one of the _ users views. this comes back to data. one of the things - users views. this comes back to data. one of the things i'm - users views. this comes back to| data. one of the things i'm most concerned — data. one of the things i'm most concerned about— data. one of the things i'm most concerned about with _ data. one of the things i'm most concerned about with gpt- data. one of the things i'm most concerned about with gpt for. data. one of the things i'm most concerned about with gpt for as| data. one of the things i'm most. concerned about with gpt for as we do not _ concerned about with gpt for as we do not know— concerned about with gpt for as we do not know what _ concerned about with gpt for as we do not know what it _ concerned about with gpt for as we do not know what it is _ concerned about with gpt for as we do not know what it is trained - concerned about with gpt for as we do not know what it is trained on. l do not know what it is trained on. sam _ do not know what it is trained on. sam mitts — do not know what it is trained on. sam mitts for— do not know what it is trained on. sam mills for the _ do not know what it is trained on. sam mills for the rest _ do not know what it is trained on. sam mills for the rest of- do not know what it is trained on. sam mills for the rest of us - do not know what it is trained on. sam mills for the rest of us do i do not know what it is trained on. i sam mills for the rest of us do not. what _ sam mills for the rest of us do not. what it— sam mills for the rest of us do not. what it is_ sam mills for the rest of us do not. what it is trained _ sam mills for the rest of us do not. what it is trained on— sam mills for the rest of us do not. what it is trained on has— what it is trained on has consequences _ what it is trained on has consequences for- what it is trained on has consequences for the i what it is trained on has. consequences for the bias what it is trained on has- consequences for the bias of the system — consequences for the bias of the system how— consequences for the bias of the system. how these _ consequences for the bias of the system. how these systems - consequences for the bias of the . system. how these systems might consequences for the bias of the - system. how these systems might lead to people _ system. how these systems might lead to people about— system. how these systems might lead to people about depends _ system. how these systems might lead to people about depends very- system. how these systems might lead to people about depends very heavily . to people about depends very heavily on what _ to people about depends very heavily on what data — to people about depends very heavily on what data is — to people about depends very heavily on what data is trained _ to people about depends very heavily on what data is trained on _ to people about depends very heavily on what data is trained on and - to people about depends very heavily on what data is trained on and so - to people about depends very heavily on what data is trained on and so wel on what data is trained on and so we need _ on what data is trained on and so we need transparency— on what data is trained on and so we need transparency about _ on what data is trained on and so we need transparency about that - on what data is trained on and so we need transparency about that and i on what data is trained on and so we| need transparency about that and we probably— need transparency about that and we probably need — need transparency about that and we probably need scientists _ need transparency about that and we probably need scientists in _ need transparency about that and we probably need scientists in their- probably need scientists in their doing _ probably need scientists in their doing analysis _ probably need scientists in their doing analysis in _ probably need scientists in their doing analysis in order- probably need scientists in their doing analysis in order to - doing analysis in order to understand _ doing analysis in order to understand what- doing analysis in order to understand what the - doing analysis in order to - understand what the political influences _ understand what the political influences of _ understand what the political influences of the _ understand what the political influences of the systems - understand what the political. influences of the systems might understand what the political- influences of the systems might be. it is influences of the systems might be. it is not _ influences of the systems might be. it is notiust— influences of the systems might be. it is not just about _ influences of the systems might be. it is notjust about politics, - influences of the systems might be. it is notjust about politics, it - it is notjust about politics, it can be — it is notjust about politics, it can be about _ it is notjust about politics, it can be about health, - it is notjust about politics, it can be about health, about . it is notjust about politics, it - can be about health, about anything. the systems— can be about health, about anything. the systems absorb _ can be about health, about anything. the systems absorb a _ can be about health, about anything. the systems absorb a lot _ can be about health, about anything. the systems absorb a lot of - can be about health, about anything. the systems absorb a lot of data - can be about health, about anything. the systems absorb a lot of data andi the systems absorb a lot of data and what they— the systems absorb a lot of data and what they say — the systems absorb a lot of data and what they say reflect _ the systems absorb a lot of data and what they say reflect that _ the systems absorb a lot of data and what they say reflect that data - the systems absorb a lot of data and what they say reflect that data and l what they say reflect that data and they are _ what they say reflect that data and they are going _ what they say reflect that data and they are going to _ what they say reflect that data and they are going to do _ what they say reflect that data and they are going to do it— what they say reflect that data and they are going to do it differently. they are going to do it differently depending — they are going to do it differently depending on— they are going to do it differently depending on what _ they are going to do it differently depending on what is _ they are going to do it differently depending on what is in- they are going to do it differently depending on what is in that - they are going to do it differentlyl depending on what is in that data. it depending on what is in that data. it makes _ depending on what is in that data. it makes a — depending on what is in that data. it makes a difference _ depending on what is in that data. it makes a difference of— depending on what is in that data. it makes a difference of the - depending on what is in that data. i it makes a difference of the change on the _ it makes a difference of the change on the wall— it makes a difference of the change on the wall street _ it makes a difference of the change on the wall street journal- it makes a difference of the change on the wall street journal as - on the wall street journal as opposed _ on the wall street journal as opposed to _ on the wall street journal as opposed to the _ on the wall street journal as opposed to the new- on the wall street journal as opposed to the new york - on the wall street journal as i opposed to the new york times on the wall street journal as - opposed to the new york times or reddit _ opposed to the new york times or reddit they— opposed to the new york times or reddit they are _ opposed to the new york times or reddit. they are largely— opposed to the new york times or reddit. they are largely trained i opposed to the new york times ori reddit. they are largely trained on all of— reddit. they are largely trained on all of the _ reddit. they are largely trained on all of the stuff _ reddit. they are largely trained on all of the stuff but _ reddit. they are largely trained on all of the stuff but we _ reddit. they are largely trained on all of the stuff but we did - reddit. they are largely trained on all of the stuff but we did not - all of the stuff but we did not understand _ all of the stuff but we did not understand the _ all of the stuff but we did not understand the composition. all of the stuff but we did not. understand the composition of all of the stuff but we did not - understand the composition of that so we _ understand the composition of that so we have — understand the composition of that so we have this _ understand the composition of that so we have this issue _ understand the composition of that so we have this issue of— understand the composition of that so we have this issue of potential. so we have this issue of potential manipulation _ so we have this issue of potential manipulation. it _ so we have this issue of potential manipulation. it is _ so we have this issue of potential manipulation. it is even - so we have this issue of potential manipulation. it is even more - manipulation. it is even more complex— manipulation. it is even more complex than _ manipulation. it is even more complex than that _ manipulation. it is even more complex than that because i manipulation. it is even more complex than that because it| manipulation. it is even more i complex than that because it is settlement— complex than that because it is settlement relation, _ complex than that because it is settlement relation, people - complex than that because it is. settlement relation, people may complex than that because it is - settlement relation, people may not

be aware _ settlement relation, people may not be aware of — settlement relation, people may not be aware of what _ settlement relation, people may not be aware of what is _ settlement relation, people may not be aware of what is going _ settlement relation, people may not be aware of what is going on. - settlement relation, people may not be aware of what is going on. that. be aware of what is going on. that is the _ be aware of what is going on. that is the point — be aware of what is going on. that is the point of— be aware of what is going on. that is the point of both _ be aware of what is going on. that is the point of both the _ be aware of what is going on. that is the point of both the wall - be aware of what is going on. that| is the point of both the wall street journal— is the point of both the wall street journal article _ is the point of both the wall street journal article and _ is the point of both the wall street journal article and the _ is the point of both the wall street journal article and the other- journal article and the other article — journal article and the other article culture _ journal article and the other article culture attention - journal article and the other article culture attention to. i journal article and the other. article culture attention to. let journal article and the other article culture attention to. let me ask ou article culture attention to. let me ask you about _ article culture attention to. let me ask you about al _ article culture attention to. let me ask you about al systems - article culture attention to. let me ask you about al systems trained l article culture attention to. let me. ask you about al systems trained on personal data, the kind of data that social media platforms collect and ask routinely. we have had many a chat office on this committee of many a year but the massive amounts of personal data that the companies have having each one of us and an ai system that is trained on that individual data that knows each of us better than ourselves and also knows the billions of data points but human behaviours and language and interaction generally, can't we foresee an ai system that is extraordinarily good at determining what will grab human attention and what will grab human attention and what will grab human attention and what will keep an individual�*s attention? the war for attention, the work for clicks but is currently going on on all of these platforms

try to make their money, i am imagining an ai system, these ai models supercharging that war for attention is that we now have technology... attention is that we now have technology- - -_ technology... and srumo: - technology... and srumo: will - technology... and | studio: will come technology... and - studio: will come away for technology... and _ studio: will come away for a few minutes. let's bring in the professor for artificial intelligence listening for the last hour. it has been fascinating, we heard sam elements representing us senators there was a need for regulation, it was critical in his view. he said things could go very wrong, there was talk of worries about jobs, wrong, there was talk of worries aboutjobs, employment, about fake stories and elections and the implications there. it is a long, long list of risks that go along with the potential benefits, isn't it? , , , with the potential benefits, isn't it? y ,, ., ., with the potential benefits, isn't it? ., ., ., with the potential benefits, isn't it? , , , ., ., ., ., it? they spent a lot of time on misinformation, _ it? they spent a lot of time on misinformation, not _ it? they spent a lot of time on i misinformation, not surprisingly, because it is the topic of the day perhaps and went through a lot about

it. however, we must mention that they are opposing things we know very well, for example miss montgomery proposed a regulation is newly identical to the european one stop it is pretty much the same solution. and proposed a system of licences to be able to operate, which has been discussed a lot. and professor stephen proposed an institution like certain where all bursitis can go together and assess —— for other scientists can get together and assess. what worries me is that we never sorted the questions that are coming from the previous generation of ai, manipulative powers. somebody said gpt could be the same but on steroids. i gpt could be the same but on steroids. ., ., , , ., steroids. i have to interrupt you because we _ steroids. i have to interrupt you because we are _ steroids. i have to interrupt you because we are coming - steroids. i have to interrupt you because we are coming to - steroids. i have to interrupt you because we are coming to the l steroids. i have to interrupt you i because we are coming to the end

steroids. i have to interrupt you - because we are coming to the end of the programme, apologies for that but we will return and get more from capitol hill on this in the next hour. stay with us here on bbc news. hello there. a fairly quiet weather day. a few showers around. it was a chilly old start this morning due to the largely clear skies overnight. but since then, we have seen more scenes like this, some patchy fairweather cloud bubble up. so through the rest of the week, then, temperatures gradually on the rise, we'll hold on to some largely dry conditions, too. so, the reason behind the settled conditions is this broad area of high pressure. the winds typically coming in from the northwest where we have had some exposure. for example, the northern isles, quite chilly here, but for most of us, it has felt a little bit warmer out there today because of those winds have been on the lighter side. so temperatures for many around the mid—teens. so, scattered showers mainly focused across parts

of scotland, northern england and northern ireland will gradually fade away through the evening hours. and then overnight, a legacy of some cloud in places. i think for most of us, it will be milder than the nightjust gone. still quite chilly, though, for the likes of northeast scotland and the northern isles. here just two—three celsius in places. then high pressure still on the scene. we do have this weather frontjust working around the top of the high that's going to bring a bit more cloud later to parts of northern ireland. and scotland will see some outbreaks of rain arriving too, together with some brisk winds. but actually for much of the uk, it will be a dry and fine day with some sunshine and a partial build up of cloud. and for many of us, temperatures getting up into the mid—teens in some of the warmest spots, 17, 18 celsius possible looking further out then on thursday, high pressure still sticking around. so when a cold front does head our way, it's not really going to bring any big changes. and also, here's a look at the air mass chart, these yellow and orange shades indicating some warmer

air on the way. so, thursday itself, then, any early mist or fog will soon lift. and then we're left with some spells of sunshine. meanwhile, further north, it will be a cloudier story at times, with a chance for a few spots of light, rain and drizzle and for more of us feeling warmer, with temperatures getting up to around 17—18 celsius more widely. looking at the outlook then, friday into the weekend, still some fairly lengthy dry spells in the forecast. temperatures on the rise, too. so in the sunshine, it should feel pleasantly warm. and that's your latest forecast.

live from london, this is bbc news. the rush to regulate artificial intelligence. the man behind chatgpt gives evidence to the us congress. has spoken about the benefits and dangers of ai, voicing his biggest fear. i dangers of ai, voicing his biggest fear. ~ , dangers of ai, voicing his biggest fear. ~' , . ., ., , fear. i think if this technology noes fear. i think if this technology goes wrong. _ fear. i think if this technology goes wrong. it _ fear. i think if this technology goes wrong, it can _ fear. i think if this technology goes wrong, it can go - fear. i think if this technology goes wrong, it can go quite . fear. i think if this technology - goes wrong, it can go quite wrong, and we want to be vocal about that. we want to work with the government to prevent that from happening, but we try to be very clear eyed about what the downside is that the work we have to do to mitigate that. this is the scene live from capitol hill where that hearing ukraine says they shot down six of russia's most advanced hypersonic missiles during a night of intense attacks on kyiv. we'll talk live to the former president, petro poroshenko.

a bbc investigation finds a uk conservative party donor