appropriately, from the artificial intelligence summit at bletchley park. the other big story this week, though, has been the quite eye—popping testimony at the public inquiry into how the government handled various phases of the covid epidemic. and, i mean, it has massively, massively taken me back to those scary, uncertain days in early spring 2020. and i spent a lot of those days sat in this very studio and the person sat across from me and being very reassuring at all times was laura k, who's back in the studio. not in the studio. no, i'm not. but i'm logistically distanced today, not socially distanced. and i'm very pleased to hear that i was reassuring back in those early days of march 2020. you were one of the only people i saw! well, i'm pleased. well, it was very nice. we were a tiny little band of people, carrying on, trying to do ourjobs as everybody wondered what on earth was going on. and it's been a very strange experience watching this, hasn't it? because i was going to ask you, i mean, ask about your various jobs over the years. so you kind of did the first draft of all of this as political editor. then you made all these documentaries and you did podcasts and they're all on iplayer, and that's kind of like the second draft. what's it been like watching the latest draft unfold? i think this week has been a really important one, and i'm sure particularly for the families of those who lost people in the pandemic, it must have been very painful. and one comment sticks in my mind this week from one of them who said it was even worse in government than we had feared from the outside. i think for me, none of the testimony has been surprising. however, the act of hearing all of this pretty disturbing stuff actually being read into the public record is something that has real gravity and real meaning. and it's made me think also about the different phases of this, as we sort of discovered over time, actually, just how much government was struggling, perhaps even imploding with the pressures of the pandemic, that early phase, really, of such a national emergency when there was real political unity, real sense of unity in the country. but don't forget, you know, just a couple of months in, there was, you know, dominic cummings�*s trip to barnard castle blew out into the open. and then, a couple of months after that, it became really clear how terrible things were behind the famous black door in downing street. you had political division inside the tory party, as well as between the labour party and the government, over whether or not there was going to be a second lockdown. and then i think a year later we began to sort of have more comprehensive accounts of how much government had struggled, you know, through sort of 2020 and those early, really chaotic moments. and, two and a half years later, then we're past the stage of private briefings or off—the—record things, or somebody popping up to do an interview and to make an intervention here and there or giving evidence to mps. this is a comprehensive account of how bad it was, being given in the most formal of settings. so, even if the individual bits and pieces of evidence are not surprising, if you were close to it, the very act of them being read into the record like this is something that really, i think, has meaning and import and, you know, leaves no doubt either as to how bad things really were. and, i mean, what i keep thinking is that, at the time as a politicaljournalist, ifelt a bit like like pac—man going around the maze, gobbling up bits of information one at a time and trying to kind of get a picture that way. but actually, what you're getting from this is not only these people who were in the room, being on the witness stand, being professionally questioned by a professional question asker, as the kc is for hours and hours and hours and hours. but they also, at the end of the day, published the witness statements and dominic cummings�*s witness statement and helen macnamara, the deputy cabinet secretary's witness statement, more than 100 pages. it's like reading their autobiography of a six—month period, and that's just so detailed. and the fact that you can put the different versions together and look for the similarities and cross—reference them. and then the fact is that in the footnotes they then refer to other emails and other documents that haven't even been published yet. it is just such a different experience from kind of feeling your way through the maze, eating up the crumbs of information, trying to work it out. yeah, and that's the point, the purpose of having a big public inquiry, right? so the record becomes complete, although remember that there is some stuff... helen macnamara said this week that the cabinet office wouldn't give her her work phone from that time. so this isn't even complete. but you're right, compared to, you know, us trying to find out what was going on right at the time, it's like, here is an encyclopedia, you know, a big kind of brain dump. and from that today, two important things, a communication between mark sedwill, who was the cabinet secretary, the most senior civil servant at the time, and chris wormald, who was the boss of the department of health at the time, discussing in the beginning of march whether or not it was the right thing to let healthy people catch the virus, like chickenpox, agreeing on 12th march that, yes, that was part of the plan. now, to give it its posh name, that there in black and white is confirmation that the government did think that herd immunity was part of the plan at the beginning. of course, we knowjust days later that all got torn up and they completely changed approach. but there we have now formally in the record in black and white from those two senior people, that was the plan that was being discussed. we also had today chris wormald saying on the record, in his view, the first lockdown was a week too late. and remember, adam, just how vigorous and fierce and stressful the debates about herd immunity and the debates about whether or not the lockdown was too late went on for months and months and months and months and months on end. and now we've got people who were right in the heart of it confirming in this way that herd immunity was something that they had as part of the plan. and also chris wormald now giving his view that the first lockdown was too late. and, of course, people will take their own personal view as to what that really means in terms of the cost, the personal cost and the potential lives that may have turned out differently if the government had acted earlier. and let's hear how some of that sounded actually in the inquiry room itself. with hindsight, we were at least a week late at all points of the mpi decisions. i agreed with the decisions at the time and the timing but, looking back, we should have done each of the things onthe12th, 16th, 23rd. if we had got to the 23rd at least a week earlier, at least a week earlier. and i suppose, again, that's an interesting combination of stuff that seems eye—popping but is also maybe not that surprising because we knew that, for a large chunk of the build—up to the first lockdown, the government was operating on the assumption that covid was like flu. and basically there wasn't much you could do to stop people getting it. so what you needed to do was just mitigate how high the peak of infections was so that the country could cope. and it wasn't until a few weeks later that various groups of people at roughly the same time in different bits of the machine were like, hang on, even if we mitigate the peak and squash the sombrero, as boris johnson famously said, actually there's still a chance that the nhs gets overwhelmed and that's when the strategy shifted. and also i think a lot of people who've been giving evidence are thinking that one of the conclusions the inquiry will reach when it finally reaches its conclusions in years, years to come, is that the lockdown was too late. so i think maybe what we're seeing now is a sort of reverse auction of people saying, well, i was the first person to realise we needed to lock down. i realised on friday the 13th but we just couldn't do it for ten more days. and i just think that's in lots of people's minds. well, of course, reputation management is also always part of these things, although what is not clear to me, which will be another interesting thing and another really important thing and part of the conversation now about it is what were the other costs, the cost of kids�* education, the cost to other bits of health care that fell by the wayside because so much was being focused on covid. and i don't know if it will ever be possible, even if the inquiry has a plan to give a kind of ledger of, if they'd done this, this would have happened. if they'd done this, this would have happened. and, you know, ithink that is a conversation that will be impossible to conclude. people are always going to have their views, depending on what their own experience was. and i'm not clear the extent to which the inquiry itself is going to try to give a view in the round of the cost of the lockdown or the cost of the pause or the cost of it. had they acted earlier, this wouldn't have happened or that would have happened. and that's a debate that will be very, very hard to settle. now, some newscasters have been getting in touch with their questions on discord. and discord is the place where you can sign up on the discord app and then chat as much as you want about what we've been discussing on this podcast. and so lots of people this week have been making points and raising questions about the inquiry and about covid. and so i thought i'd pose some of them to laura. bev said via discord, has there been any evidence that brexit made the covid decision making worse? i think it's hard to come up with a specific slam dunk piece of info. what is certainly true is that when the pandemic arrived, it met a government and a political party and a machine in whitehall that was pretty battered, pretty bruised by a real period of angst and political carnage and disagreement and also genuinely difficult policy problems to solve. and i remember a former official from one of the devolved administrations saying to me, actually, in their view, like brexit had really damaged the civil service. so they were sort of walking wounded, if you like, and then suddenly roll straight into dealing with a problem that was even bigger, even harder to confront. so can you definitely say yes, because the uk was grappling with brexit, this particular thing didn't happen or went wrong because of that? i think it's probably hard to, but generally did it mean that, you know, the sort of machine hadn't been operating at a sort of rolls—royce level, shall we say? yeah, i think that probably is fair to suggest. and some witnesses have suggested that for sure. but equally on the flip side, some witnesses have said, oh, all the scenario planning for no deal meant that people were like, oh, actually, if we run out of medicines, this is what you do or you can't get that, this is what you do. so yeah, that's a great point. some of that ended up being quite helpful. so i should mention alsojust on that, some backers of borisjohnson would also say that although the uk could have gone on its own on the vaccine, if you're in the eu, it was much more likely to go on its own having left the eu. so, you know, the uk's huge success of the vaccine programme is not, you know, is maybe tangentially but probably linked to the fact that we had left the eu by that point. and then finally, here's a real brainteaserfromjim. i'd be interested to know your thoughts generally on how government can improve the quality of leadership. how can we improve the quality of people at the top? now, there's two different things there. because there's the leadership, which is actuallyjust the act of leading and the tools and machinery you've got at your disposal if you're a leader. and then there's actually the kind of person that ends up being a leader. so that's kind of two different things, isn't it? yes, i think lots of people who have done big jobs in and around westminster would say that the system, for want of a better word, hasn't changed enough to keep pace with how modern life is now, isn't fast enough, isn't responsive enough. i think a lot of evidence at the covid inquiry in the last couple of weeks has backed that up. i think there's a different question, as you say, as i'm there, jim, that how do you attract people to become politicians or to go into politics when it's a pretty tough business with huge responsibility and privilege, but also massive scrutiny, sometimes quite a poisonous environment. yes, people are very well paid if they choose to become mps, for example, compared to the average salary. but other very aspiring, ambitious people who think they're kind of brain boxes may well think, well, i can go in there and lots of money elsewhere. so there's two different debates there. one, is our government machine up to it? covid inquiry would suggest, well, yes, could do better. and the second is, why would you want to become an mp? jim for pm! is this a pitch? we have to remain impartial. well, the thing is though, it does sound really dweeby, though. ijust wonder, you, no, never! like all former prime ministers say, they wish they'd had like a bigger, bigger operation at the centre because so often they end up having to just like run the whole show, even though we actually do not have a presidential system. it's funny, that, isn't it? i don't think it's power—hungriness just for the sake of it. i think it is speaking from experience. and i just wonder if actually, as a population, if the majority of people then decided, actually, yeah, we do want a prime minister that's a bit more presidential, where it's like one person has more power and a less kind of parliamentary cabinet system, ijust don't know if we've got a system that would ever allow that to even happen. even if actually the person at the top and the people electing the government felt the same way. well, we do have that system. it's about size of the majority, isn't it? if someone gets an absolutely whopping, mahoosive, gigantic majority, in theory, they can do all sorts of things without very much resistance. if they have no majority at all they will find it a bit more challenging. but i also think that prime ministers, and this is to be very techie about it, but as a very smart official who once explained this to me in great detail, they said that they see it in whitehall, at the top of the machine they see it as their responsibility to kind of get like plasticine around the prime minister of the day in order to get the best out of them. because we don't have a written constitution, it's flexible and bendy and all the rest. prime ministers can therefore work in the way that they wish within a set of parameters. but what this person also said that theyjust never managed to get it right with borisjohnson. and whether he would have been able to be a successful prime minister with a different whitehall is entirely a matter of a completely different conversation. but it's an interesting, if thoroughly dweeby thing, that prime ministers can sort of set it up how they want, really. and we have had quite a lot of prime ministers in quite quick succession. we've had loads. loads and loads and loads and loads and loads. and then, you know, i mean, yeah, ithink, yes, lots. that's a whole other that could be a whole podcast series, behind the scenes with laura kuenssberg, you should think about doing that! i thought you were going to say the prime ministers! that would be another thing, you see, you'd have the prime ministers, i'd love to get four of them in a room. wouldn't that be good? i'm not sure they'd all accept it. oh, i'm sure they'd never do it. i'm looking at three empty chairs. maybe we could get three of them altogether, i don't know. in the newscast studio! amazing! the door�*s open! the invite always there. and i bet you'll get them to turn up on the saturday edition of newscast you'll be doing with paddy, and i'll still be sat here on my own! i just didn't want to tell you, i didn't want to embarrass you. right, laura, lovely to catch up and speak to you soon. bye. now, as promised, let's talk about this week's other big story, which is coming to you from bletchley park, where they cracked the enigma code in the second world war. it is, of course, the big international summit hosted by the uk about the safety of artificial intelligence. here's some genuine human intelligence, chris mason, hello! steady on, steady on. hello. i am in the back of a white van with a satellite dish on the top of it, and doing newscast from here, in this, as you say, quite a spot, really, the history of this place, really quite something. and also here in the studio with me face to face, the oldest tech of all is michael kratsios, who was the chief technology officer for the us government under president trump and now works for a company called scale ai. hello! hello, great to be here. now, before we dive into this summit, i always think when we talk about al, it's actually very useful to get in layman's terms, what bit of ai you actually do. so what does scale ai do as a company? yeah, we provide infrastructure for the large model builders, so folks like 0penai, and ai solutions for enterprises and for governments around the world. and that's interesting, because that sort of makes you realise that al is an idea, but it's also a quite complicated concept to just deliver as a piece of technology and a thing that works in a place. yeah, absolutely. i mean, chatgpt was a big breakthrough where most people who had never interacted with al had a chance for the first time ever to to touch it, had a chance for the first time ever to touch it, feel it, and understand, you know, what it can do. now, the next big step is how do governments and companies actually operationalise that to drive some sort of r0! or some sort of benefit. roi being return on investment. precisel . ,, .,~ ., we should also say that this summit, i mean, it's got two focuses. one is the idea of safety, so not about yourjob being lost because you've been replaced by an algorithm, and a type of ai called frontier ai, which is basically kind of like the super cutting edge that's not even really here yet. correct, yeah. so this particular event was very much focused on the risk posed by these next—generation artificial intelligence models. so i would consider these low—probability, high—impact events that could impact the world. and for a lot of people out there, this is extraordinarily important, you know, ai could, you know, take a life of its own and do something crazy. bad actors could potentially use some of these advanced systems to potentially create things like biological weapons. so the uk government thought it made sense to kind of bring the world together to think about these potential risks. and so, chris, you've had a couple of days at bletchley park now, so what's your take on how the summit went? so i think, first of all, it's been pretty well attended. there was quite a bit of scepticism. i remember the prime minister talking to us as a political press pack, i think when we flew back from a trip to washington, actually, about four or five months ago, and there was a bit of scepticism about who the prime minister, the british government, might be able to assemble around the table. why should the uk have this kind of convening role on the world stage? that was the kind of, if you like, sceptical reporters�* question. and actually when you see those who turned up here, particularly in the context of what's going on in the middle east and elsewhere, in other words, you know, leaders, significant political leaders who have got lots of other pressing demands on their diaries, to have the secretary general of the un, to have the vice president of the united states, the president of the european commission and others. it was quite a cast list, as well as plenty of people from the various technology companies. and then, i guess they're trying to start a potentially very long process of is trying to work out with this huge technology, with huge potential, but also huge potential risks and dangers, how you go about, at a global level, trying to shape some sort of basis of rules. and that, i mean, clearly is complicated, and can be controversial as well, because obviously, individual countries and individual companies, perfectly understandably, will want to sort of maximise advantages, either to their own bottom line or to their own national security or whatever it might be. and the argument that the government here made was, let's start a process about how you go about thinking about those rules, with an acceptance that it will take quite a while, probably, to get to something that that can work, if it can work. and here's how rishi sunak, the chair of the summit, summed it up at the end. this partnership is based i around a series of principles which set out _ the responsibilities we share, and it's made possible - by the decision i have taken, along with — vice president kamala harris for the british and american governments to establish i world—leading ai safety institutes, with the public—sector capability . to test the most advanced frontier models. - so, michael, put that into human speak for us. does that mean you're going to have to hand over more stuff to the british or american governments than you would have had to do last week? yeah, so for these large language model providers, i think what what i have found so fascinating and interesting over the past year is that they themselves are very concerned about ensuring that their models are safe for the long term. and in order to do those checks, they need the help of other private—sector companies and of governments to help do that testing and evaluation. so now, after this event, it seems like, you know, through these safety institutes, they will be partnering more closely with government actors to kind of help do some of this frontier testing and evaluation. but will that make any difference, then? will that be any different from what was what was already happening? yeah, you know, i think that the government may be a little bit more involved. i think what's very interesting and kind of an undertone or a conversation that was being had at the summit yesterday when i was there was this question about sort of open—source versus closed—source models. so the folks at the summit all agreed to sort of share some of the workings of their models with some of these regulators so they can do the testing and evaluation. but there's an entire community around the open—source development of these large language models. and what's interesting is these these llms, if you will, or large language models, are not just sort of the six folks that were at the table. take, for example, the united arab emirates. uae has a very high performing model. the chinese have a number of models themselves. and those are all models which, you know, haven't agreed to share their model weights with the uk or the us. so as this thing progresses over the years, you know, there's a lot of players out there who are not at the table today that are also developing them. oh, so the next alternative to chatgpt, the thing that grabs the entire world's imagination might come from dubai, and the british government and the american government, however huge and snazzy their institutes are, will have no idea how it works and how dangerous it could be and have no way to stop it? you never know, yeah. so hopefully you can create some sort of testing and evaluation standards or methodology that, wherever you are in the world, you can try to run those tests against any model globally. but what has sort of been proven over the last eight or nine months is that this technology is really just dependent on you having enough money to buy chips to do the compute necessary to train your model. 0na on a practical level, i was at a fringe event at this conference, not being a vip but a nice lanyard to get into bletchley, and they were talking about with these larger language models we have been talking about, you have to train them on the data and they scrape the internet for lots of examples of what they are looking at 11 of the things they talked about was facial recognition. the internet is awash with pictures of white men because throughout history they are the people that had their picture taken the most. how does your company or your clients make sure that they get enough pictures of black women, so the algorithm of recognising people's faces reflects the world at as is? you test and evaluate these models to ensure _ you test and evaluate these models to ensure the data used to train them _ to ensure the data used to train them is — to ensure the data used to train them is representative of what you ultimately — them is representative of what you ultimately want to do with the model and what _ ultimately want to do with the model and what interesting with these large _ and what interesting with these large language models it is a 2—step process— large language models it is a 2—step process to _ large language models it is a 2—step process to make the llm. the first step is _ process to make the llm. the first step is to _ process to make the llm. the first step is to go out and scrape the intemet — step is to go out and scrape the internet for the totality of human knowledge into a model. the second step is— knowledge into a model. the second step is something called reinforcement learning with human feedback — reinforcement learning with human feedback. what that means is that you essentially inject into the model— you essentially inject into the model expert created data, so humans themselves— model expert created data, so humans themselves create data of the best answers _ themselves create data of the best answers on very tough niche problem is the _ answers on very tough niche problem is the model isn't performing well on and _ is the model isn't performing well on and the — is the model isn't performing well on and the combination of human generated — on and the combination of human generated data plus the internet is able to _ generated data plus the internet is able to create a better model which answers _ able to create a better model which answers these questions more correctly _ answers these questions more correctly. i answers these questions more correctl . ., �* answers these questions more correctl. ., �* ,, ., ., answers these questions more correctl. ., �* ,, ., correctly. i don't know about you but when i _ correctly. i don't know about you but when i hear _ correctly. i don't know about you but when i hear that _ correctly. i don't know about you but when i hear that that - correctly. i don't know about you but when i hear that that is - correctly. i don't know about you but when i hear that that is kind | correctly. i don't know about you i but when i hear that that is kind of a bingo moment for me because actually you are like how smart are the systems if they are based on lots of people sat there telling them what to think in the first place? them what to think in the first lace? , , , ,, place? yes. i suppose it is like that old saying, _ place? yes. i suppose it is like that old saying, rubbish - place? yes. i suppose it is like that old saying, rubbish in, - place? yes. i suppose it is like - that old saying, rubbish in, rubbish out. ., ., . out. garbage in, garbage out. we have an american _ out. garbage in, garbage out. we have an american present. - out. garbage in, garbage out. we have an american present. is - out. garbage in, garbage out. wej have an american present. is that how it was? _ have an american present. is that how it was? i _ have an american present. is that how it was? i was _ have an american present. is that how it was? i was angus - have an american present. is that how it was? i was angus allison l have an american present. is that how it was? i was angus allison it have an american present. is that i how it was? i was angus allison it a bit. ., , ., ., ~ ., ., bit. that is how we talk to our customers. — bit. that is how we talk to our customers, whether - bit. that is how we talk to our customers, whether in - bit. that is how we talk to our - customers, whether in government or enterprise. _ customers, whether in government or enterprise, you have to have high quality— enterprise, you have to have high quality data before you can have hi-h quality data before you can have high quality ai. gn quality data before you can have high quality al— quality data before you can have hiuh auali ai. , high quality ai. on the next episode of newscast — high quality ai. on the next episode of newscast we _ high quality ai. on the next episode of newscast we will _ high quality ai. on the next episode of newscast we will be _ high quality ai. on the next episode of newscast we will be able - high quality ai. on the next episode of newscast we will be able to - high quality ai. on the next episode of newscast we will be able to talk | of newscast we will be able to talk about the real denouement of this summit which is rishi sunak doing a 0&a summit which is rishi sunak doing a q&a with elon musk of the most unpredictable tecpro that there is so i wonder what will emerge from that. chris, i hope you have had fun at bletchley park. bill! that. chris, i hope you have had fun at bletchley park.— at bletchley park. all been good, talk to you _ at bletchley park. all been good, talk to you soon. _ at bletchley park. all been good, talk to you soon. michael, - at bletchley park. all been good, talk to you soon. michael, thank| at bletchley park. all been good, . talk to you soon. michael, thank you for cominu talk to you soon. michael, thank you for coming into _ talk to you soon. michael, thank you for coming into our _ talk to you soon. michael, thank you for coming into our very _ talk to you soon. michael, thank you for coming into our very old - talk to you soon. michael, thank you for coming into our very old tech - for coming into our very old tech studio. that's all for this episode of newscast. we will be back with another one soon. bye—bye. of newscast. we will be back with another one soon. bye-bye. newscast. newscast from — another one soon. bye-bye. newscast. newscast from the _ another one soon. bye-bye. newscast. newscast from the bbc. _ hello there. tomorrow should be a drier day for more of the country with some sunshine around. today, though, still very unsettled again dominated by low pressure. this one is bringing some very strong winds through biscay into the west coast of france, northern parts of spain. we do around the base of that low see some stronger winds, though, running through the english channel and again, unfortunately affecting the channel islands. the winds this time gusting close to 50mph, much lighter winds further north, especially in scotland, northern ireland, where it's turning a bit drier, but quite cool there, particularly if you are stuck underneath this rain and there are some heavier showers running along the south coast of england heading towards the dover straits as well. now, if you are going to be seeing some fireworks this evening, looking a bit drier in scotland and northern ireland, a few breaks in the cloud, not such good weather across much of northern england and north wales. still got some rain here. and these showers that are coming all the way from hampshire to kent could be heavy, possibly thundery and there'll be some gusty winds in there as well. so not particularly ideal. those winds may tend to ease overnight as the low pressure that's been bringing all the wet weather tends to move away. so by the time we get to first thing tomorrow, these are the temperatures. it could be quite a chilly start in scotland and northern ireland. a few patches of mist and fog in scotland to clear away in the morning. quite a cloudy start for eastern scotland, eastern parts of england, a little bit of rain in the air. that should move away as the low moves away and we get more sunshine coming through. showers, though, getting blown into westernmost parts of scotland. not too many in northern ireland, a few coming over the irish sea into western parts of england and wales. again, the strongest of the winds are going to be across the south, across southern most parts of england and through the english channel. gusts close to 40mph or so. lighter winds further north. and those temperatures, 11 to 13, probably a little bit warmer than today across parts of northern england and north wales with the promise of some sunshine. bonfire night tomorrow evening, clearer skies across the east. showers continue out towards the west. now, the week ahead, not looking particularly stormy, not looking that windy nor indeed that wet. we are going to find a weather front coming in from the atlantic by the middle part of next week, bringing some rain and maybe some stronger winds. but ahead of that, the start of the week, we'll see some sunshine and some showers. again, most of the showers are going to be in the west. live from london. this is bbc news. foreign ministers from egypt and jordan call for an immediate ceasefire in gaza, but antony blinken says that would allow hamas to regroup. ceasefire now would simply leave hamas in place, able to regroup and repeat what it did on october the 7th. hamas says 15 people have been killed in an explosion at a un—run school. israel says it's investigating the circumstances. the us says up to 400,000 people remain in northern gaza, as israel's ground offensive there continues. the un says no part of gaza is safe. authorites at the rafah crossing say no foreign passport holders will leave gaza for egypt until the injured people in gaza's hospitals are allowed to travel through the crossing. hello. welcome to the programme. arab foreign ministers at a meeting in amman have told the united states' top diplomat antony blinken that there must be an immediate ceasefire in gaza. thejordanian foreign minister ayman safadi said the killing of civilians by israel could not be justified by its claim of self defence.