Latest / AI in the UK / Nice Talent, Shame About the Infrastructure
Transcript
- 0:00All right, your luck. Listen up.
- 0:15Hello and welcome to AI in the UK.
- 0:18I'm Neil Footman. And I'm Matt Wilkinson.
- 0:21Professionally, we help organisations, leaders and
- 0:24individuals learn about and implement this big scary
- 0:27behemoth called AI. There's some amazing stuff going
- 0:31on with AI in the UK that doesn't always get the attention
- 0:35it deserves. With AI in the UK, we want to
- 0:38change that. Hello and welcome to what has
- 0:57undoubtedly been another very colourful week in the world of
- 0:59AI in the UK. This week's stories.
- 1:03First off, an AI model so capable the Bank of England
- 1:06governor named it in a speech before the public can even use
- 1:10it. Open AI has walked away from a
- 1:13data centre that was part of a £31 billion investment package
- 1:18while signing a London office lease three days later.
- 1:22And the Bristol startup discovers that selling an AI as
- 1:25a woman who will never ask for a raise is not, in fact, the
- 1:28marketing win of 2026. So for the first of that lot,
- 1:33Matt, over to you. Thank you, Nils.
- 1:36Yeah. And the first of these stories
- 1:37is one of those stories that possibly isn't yet a story.
- 1:41Here's what actually happened. Anthropic has chosen not to
- 1:44release Claude Methos. It's a gated, invite only
- 1:48preview, limited to around 40 or 50 organisations under a
- 1:53programme named Project Glasswing.
- 1:56Named partners include Amazon Web Services, Microsoft, Apple,
- 2:00Google, NVIDIA, Cisco and JP Morgan Chase.
- 2:03All those that you could imagine.
- 2:05Although the addition of finance is interesting, but even in that
- 2:10controlled state the models demonstrated capabilities of
- 2:13enough to set regulators moving at speed we haven't seen before
- 2:17from AI. The Bank of England, the
- 2:19Financial Conduct Authority, Her Majesty's Treasury and the
- 2:22National Cybersecurity Centre are all in active talks with
- 2:26major UK banks, insurers and exchanges, with a formal
- 2:30briefing expected within a fortnight.
- 2:32Bank of England Governor Andrew Bailey named Mythos explicitly
- 2:37in a Columbia University speech on the 14th of April, suggesting
- 2:41it was a major cyber security concern and that cyber had
- 2:45climbed regulators risk rankings faster than any other category.
- 2:50What perhaps makes this feel different to previous AI safety
- 2:53moments is that the capability claims are independently
- 2:56verified. EU KS AI Security Institute
- 2:59evaluated Mythos and found it to be the first model to complete a
- 3:03full end to end simulated corporate network attack, a 32
- 3:09step chain from initial reconnaissance to full takeover
- 3:12that they estimate takes skilled humans about 20 hours to
- 3:15complete. Mythos completed it in three out
- 3:18of 10 attempts and Tropics own red Team documentation shows it
- 3:22autonomously finding and exploiting a 27 year old
- 3:26vulnerability in open BSD, a system with such a strong
- 3:30security reputation. It's, you know, it's, it's
- 3:33dragged people in to try and hack it and it's generated
- 3:37thousands of 0 day findings across every major operating
- 3:41system and browser. This is not extrapolation from
- 3:44Benchworks, it's operational results.
- 3:47The context that really matters for how we read this is that
- 3:50without Project Glasswing, the coordinated defensive deployment
- 3:54that gives vetted organisations across access before anyone
- 3:57else, this story looks exactly like a play from the Dario Amede
- 4:02Fear playbook. Create the risk, define the
- 4:05response, control the narrative. With it, the structure is at
- 4:08least coherent as a genuine attempt to put defenders ahead
- 4:11of attackers. Whether you believe that framing
- 4:14is the core tension here is worth noting.
- 4:17Also worth noting is that AI danger warnings have been
- 4:20overhyped for the past seven years, ever since GP2 was
- 4:24released, and it was released back when Amede was at Open AI
- 4:27prior to him leaving and forming Anthropic.
- 4:31However, this is the first time that a major jurisdictions
- 4:34financial stability apparatus has responded to a single model
- 4:39as if the future was at risk. Either.
- 4:42The boy has cried, Rolf. Either the boy has cried wolf
- 4:47correctly this time, or we're about to find out how much
- 4:49institutional credibility gets spent when the wolf doesn't
- 4:53show. What do you make of this Neil's?
- 4:55This has been absolutely all over the news everywhere is so
- 5:01on on the question of whether it's hype or whether it's real.
- 5:04That is, that's a global concern, isn't it really?
- 5:06That's to do with Anthropic and, and your views on Dario Amaday
- 5:10as, as an individual and, and, and their track record in
- 5:12dealing with this. Now you told me just before the
- 5:14show started, I was, I was aware that there had been a kind of
- 5:17big security risk announcement around the release of GPT 2.
- 5:21I was aware of that. I was not so aware of Amaday's
- 5:24direct involvement. Is that is that your
- 5:26understanding is that he was, he was one of the key players in in
- 5:29making that decision. Yeah, he was one of the key
- 5:31players at at Open AI before he left at a form Anthropic.
- 5:35And he obviously there was a famous falling out with with Sam
- 5:38Altman, which has been well documented.
- 5:40And we don't need to cover here. But that GP2 safety announcement
- 5:44happened when Amadei was there. And he's always been very driven
- 5:48by AI safety in a number of ways.
- 5:50And in fact, we've had the recent news about the guardrails
- 5:54that they've put around the usage of, of their models in the
- 5:58US government, and that's caused issues with the Department of
- 6:01War over there. They've even created an entire
- 6:07ethics framework for their, you know, for their new powerful
- 6:09models. So there's, there's a lot to be
- 6:11said about Amadei's, you know, stance towards safety.
- 6:14And sometimes maybe they, they see these results and maybe get
- 6:18a bit carried away. I know from friends of mine that
- 6:22that work in cybersecurity that a lot of these zero days
- 6:24vulnerabilities, they, they're like the AI slop of the
- 6:27cybersecurity world. And the thousands that are being
- 6:30reported are basically clogging up the ability of people to spot
- 6:35the really dangerous vulnerabilities from those that
- 6:37aren't. Before we drag this kicking and
- 6:40screaming back to the Uki mean, my my own view on Amede is that
- 6:44no one's no one is purely motivated by anything.
- 6:47Everyone is complicated. Is is he also subject to the
- 6:52pressures of trying to raise huge amounts of money in order
- 6:56to keep the anthropic show on the road?
- 6:57Of course, my own feeling is that his concerns about security
- 7:02are sincerely held. I mean, you've probably, I've
- 7:05read, I think I've read one of his essays and I've heard
- 7:07excerpts from interviewed on podcasts.
- 7:10I've got no inside track his his his concerns seem genuine, at
- 7:14least to a large degree. Does this mean this was entirely
- 7:16free of a marketing positioning? Perhaps not, but I mean from
- 7:21what I've read, it seems genuinely concerning.
- 7:23All that said and to drag it back to the the UK.
- 7:26So it was an eye opener that the AI safety Institute undertook
- 7:31research on this and as you mentioned in the story, the 32
- 7:34step corporate network attack. So that that was the first time
- 7:38they they verified this concerning yes, as it always is
- 7:43with these things. And as you can have suggested,
- 7:45there's been a fair bit of crying wolf on this over the
- 7:49last seven years. And the real danger is that when
- 7:53something really is a major and imminent threat, it's going to
- 7:57fall kind of on deaf ears. And I have to say, as someone
- 8:01who keeps up with this stuff, I hear the story.
- 8:04I'm really concerned by it. I do some research and look into
- 8:07it and then a few days pass and something else is at the top of
- 8:10the headlines. Yeah, it definitely feels that
- 8:13there's, you know, the hype train is real.
- 8:15But again, I think we have to separate the, you know, what
- 8:19makes the news media and, and the, the news day in, day out.
- 8:22And of course, once the story's broken, there's a, you know,
- 8:24there's a tidal wave of coverage.
- 8:26And then what else can you add to it so it dies away, but it's
- 8:29getting a lot of news. I think what's Oh, sorry.
- 8:32What's what's worth noting is that they didn't train this
- 8:35model to be good at cybersecurity or cyber attacks.
- 8:40They trained it to be good at coding.
- 8:41Clearly, clearly, absolutely to to, to go back to the issue of
- 8:45the UK regulators. I do think it is significant
- 8:47that they have so publicly staked a position on this.
- 8:51The whatever the other strengths and weaknesses of of EU KS
- 8:55position, AII think that they have a genuinely credible and
- 8:58strong background. The, the AI Safety Institute is
- 9:01a very serious institution. Bank of England too, I would
- 9:05say. I think there is that I'm
- 9:07reading into that, that it is a serious issue.
- 9:09The other thing, of course, and it's too early to say is that
- 9:12when and well, assuming these are mythos is potentially as
- 9:15dangerous as you say and it stirs this kind of institutional
- 9:18and industry response. And perhaps they, they, they
- 9:22managed to deal with these issues.
- 9:23I mean, one danger there is the Y2K phenomenon, as we all
- 9:26remember, no Y2K is remembered as this big damp squib.
- 9:30But you talk to or certainly I've read interviews with many
- 9:34IT experts at the time who said the reason it wasn't a disaster
- 9:37was partly because the measures that were taken.
- 9:39One question to ask regarding this is how big a deal will it
- 9:42seem five years down the line if it turned out to be another bit
- 9:47in the news cycle? It's things like this that could
- 9:50really harm future responses because people will shrug.
- 9:53So it's a bit of a paradox really.
- 9:55It is essentially you're damned if you do, you're damned if you
- 9:58don't. Either you don't take it
- 9:59seriously enough and something bad happens, or you do and
- 10:02people swing into action. Some, some kind of solution is,
- 10:07is, is put together and people move on to the next thing.
- 10:10Anyway, for, for, for what it's worth to bring back to I, I, I
- 10:13think this is seems to be materially different to these
- 10:17previous announcements in the past.
- 10:19And I think the involvement of the UK regulators in the bank
- 10:21is, is, is a, is a very clear signal for me.
- 10:24Yeah. Just to just to wrap this one
- 10:26up, I think they had to that there was so much noise coming
- 10:29out about this model that they had to take it seriously.
- 10:32So even if it is the boy crying wolf at this time, they had to
- 10:36take it seriously. You had to be shown to be taking
- 10:38it seriously to to make sure that they still had the
- 10:40confidence of the public. So there is a there is a
- 10:42challenge there. But you're right, this does feel
- 10:44a little bit different to me. And I think it's mainly because
- 10:47of that. Defenders first project, last
- 10:49wing in place, Yeah. Absolutely OK More news from
- 10:54Frontier AI Labs and their connection with the UK So 3
- 10:58announcements from Frontier AI Labs in the past week told a
- 11:02consistent story about Britain's place in the global AI
- 11:05infrastructure race. The UK is a home for talent and
- 11:09policy, but not for compute. Last Thursday, Open AI paused
- 11:14Stargate UK, it's planned data centre at Cobalt Park in North
- 11:18Tyneside, citing high energy costs and regulatory
- 11:21uncertainty. Announced last September
- 11:24alongside A wider 31 billion UK tech investment package,
- 11:28Stargate UK was developed with NVIDIA and British
- 11:31infrastructure firm N Scale. It was always a small cousin to
- 11:35Open A is $500 billion US Stargate programme.
- 11:39Three days later, Open AI signed its first permanent London
- 11:43office lease. 544 seats at Regent Quarter, Kings Cross
- 11:47opening in 2027. The company already employs
- 11:51around 200 people in London and last month named the city its
- 11:54largest international research hub outside San Francisco.
- 11:59Yesterday, Anthropic followed with an 800 seat in expansion in
- 12:03the Knowledge Quarter, the cluster around Kings Cross and
- 12:06Houston, home to DeepMind, Meta and Wave.
- 12:09The move comes after reported approaches from Desit, which is
- 12:13the Department for Science, Innovation and Technology,
- 12:16courting the company following its public dispute with the
- 12:19Trump administration over military use of Claude.
- 12:21A story we touched on last week. The UK government isn't letting
- 12:26opening eyes explanations stand unchallenged.
- 12:30Speaking at Bloomberg event in London, AI Minister Kanishka
- 12:34Narayan said very mixed bag here.
- 12:36On the one hand, some clear room for optimism, but the
- 12:40uncertainty around Stargate is less welcome.
- 12:44So let's let's take a first question around that.
- 12:48Who do we believe open AI or the government around the reasons
- 12:51for this delay of rolling this out?
- 12:53What, what, what side do you sit on, Matt?
- 12:55What do you think is going on? Well, I, I think it's a bit of
- 12:58both really, and I hate to take the, the middle ground on my
- 13:01first response, but U KS industrial electricity is among
- 13:04the most expensive in the developed world.
- 13:06And we've also yet, you know, unresolved that, you know, that
- 13:10copyright regime. So taken together, we've got
- 13:14expensive costs to build challenges getting, you know,
- 13:18getting things built. We've got predicted in inflation
- 13:22due to geopolitics around the world.
- 13:24I kind of can see why when a company is gearing up to IPO
- 13:28that they might pause at a big investment like this.
- 13:31Having said that, the, you know, our minister is just doing, you
- 13:34know, is doing politics. Nothing might have changed.
- 13:37But actually is it still as, is it still as attractive to build
- 13:41data centres in the UK? You know, surely you'd want to
- 13:44go and build them where land was cheap, where electricity was
- 13:47cheap, and that you didn't have to worry about getting sued to
- 13:51use a copyright. Yeah, I think that the given
- 13:55reasons are both a convenient pretext and also true.
- 14:00Yes, the energy costs are very high.
- 14:02The regulatory picture is uncertain around copyright.
- 14:05It's not new. And I think Open AI has sort of
- 14:09rather bigger concerns at the moment.
- 14:10Yes, they've got the forthcoming IPO.
- 14:13They want to be first, they want to be biggest, that they're in a
- 14:15sort of death match with Anthropic about that.
- 14:20So and not only on that, I mean, I think they're feeling the heat
- 14:24now as a company more than they ever have done.
- 14:26It was always kind of assumed they were the market leader and
- 14:28they're under all sorts of pressures now.
- 14:29So this would be a good time to get out of a potentially
- 14:34extremely expensive entanglement in the UK.
- 14:37So we can move on to another element of the argument and that
- 14:40is, is this. So assuming this is what's
- 14:43happening, assuming it is a big magnet for talent, but it's not
- 14:46a a magnet for these infrastructure projects, is that
- 14:50a problem? What do you think?
- 14:53Yeah, We know that frontier research and policy influence
- 14:56and high paid, high value jobs, you know of the value end of the
- 14:59AI value chain. We and we know that by having
- 15:03researchers here, we influence what happens to the model.
- 15:06So that's a really, really good thing.
- 15:09You know, we know also know that data centres are really just
- 15:12construction projects that create transient building jobs,
- 15:15modest ongoing headcount and consume enormous amounts of grid
- 15:18power. We don't have that electricity
- 15:21freely available. We know that if Open an AI and
- 15:24Anthropic between them employ over 1500 people in London on
- 15:28research, safety, policy and enterprise work, that might be a
- 15:31better outcome for British AI capability than just having a
- 15:35data centre, you know, running somewhere in Northumberland.
- 15:39You know, I think until recently I might have agreed with that
- 15:42too. And I think that there is more
- 15:46than it just being 1500 people now working for Open AI and
- 15:49Anthropic. We keep on going back to the
- 15:51Deep Mind example and and their their role in forming this
- 15:55Knowledge Quarter cannot be overstated that the decision by
- 15:58Demosus Abbas to insist on Deep Mind remaining in London is
- 16:02basically what one of the absolutely central tenets of why
- 16:06the Knowledge quarter was formed.
- 16:07Now add in 1500 staff for Open AI and Anthropic and you assume
- 16:12these are going to be on the whole very smart people.
- 16:15You've got a whole potential you've got a massive growth of
- 16:18that ecosystem. People from those companies,
- 16:20people who work there will and some of them are researchers as
- 16:24Open AI is saying. Some of these people will end up
- 16:27spinning out and starting companies of their own, and some
- 16:29of them might even do it here in London given the exposure to the
- 16:32university and so on. But on the question of data
- 16:36centres, so I recently heard a really fascinating interview
- 16:38with Tom Blomfield, who as you may know was one of the founders
- 16:42of Monzo. Now he worked, he now works for
- 16:45one of these accelerator, what do you call them, Accelerator
- 16:48labs in San Francisco. Y Combinator, that's it.
- 16:50Y Combinator I think is 1. He worked for or one of those
- 16:53similar companies. He's very much on the side of AI
- 16:57is going to take away most knowledge worker jobs in the
- 17:01next three to five years. He feels that very strongly and
- 17:04he was making a point. A fascinating 1.
- 17:06I, I, I don't know if you've wondered this yourself, but I
- 17:08have often thought that if, if this happens, if this big
- 17:12enormous hit on knowledge worker jobs takes place and suddenly
- 17:16there's loads of people out of work or underemployed and, and
- 17:19suddenly with far less disposable income, then who buys
- 17:23stuff or how is, how is the economy maintained?
- 17:26This pressure is becoming very acute in the States and it will
- 17:29become more so here. And you, you may have heard this
- 17:32is going in all sorts of directions, but you heard this
- 17:34thing about Sam Altman being attacked by this, it was like a
- 17:37fire bomber. And I think afterwards,
- 17:40partially in response, he, he, he talked a bit about the
- 17:44responsibility, the huge responsibility of trying to deal
- 17:47with societal and global issues that could come about as a
- 17:51result of AI. And one of the things he talked
- 17:54about was having this complete change of social contract
- 17:58whereby these enormous hyperscalers would essentially
- 18:03Americans would all have an individual stake in a company
- 18:06and they derive income from it. Now, it's a big, bold idea.
- 18:10And I dare say it's been thought of elsewhere.
- 18:12And it's really good if you're an American.
- 18:15How about if you're a Brit? And now we don't or not only a
- 18:18Brit French or well, they've got Mistral, whatever.
- 18:21Most, most countries do not host these huge LLMS, OK.
- 18:26And so they're not in a position to build AGI domestically in
- 18:30that case. So Tom Bloomfield's suggestion I
- 18:33thought was really interesting was that data centres could act
- 18:36as the engine of the economy. Revenue taken from data centres
- 18:40and, and all, the, all the economic activity they spur
- 18:43could become a, a critical set source of income to, for this
- 18:47new social contract. In which case having data
- 18:50centres becomes completely different in terms of value.
- 18:53I know that, I know I've gone on a bit of a tangent there, but I
- 18:55thought it was a fascinating idea.
- 18:57How feasible it is, who knows? Yeah, I mean it.
- 19:01It all looks towards moving us towards some form of universal
- 19:05basic income, which on one hand I'm not averse to having a
- 19:09minimum level that people are funded to in general, if
- 19:13economies can support it that way.
- 19:15The problem I have with it is that it neglects some key
- 19:21tenants of human nature, one of which is the need for purpose,
- 19:25the purpose that most people get from their roles, from their
- 19:29families, from the rest of it. It's important.
- 19:31So I sort of look at this as really as a challenge in if the
- 19:35need to go to work, the need to earn, to support ourselves is
- 19:39taken away from us, even if that's completely unrealistic in
- 19:43the future. I really worry about what that
- 19:44does for the very core nature of of human beings and our need for
- 19:48purpose. As do I I think about that a
- 19:51lot, particularly, you know, having a young daughter and just
- 19:54wonder what she's she's going to be doing.
- 19:55But any anyway that that is a rabbit hole that we could spend
- 19:58a lot of time going around and and unfortunately, we're not
- 20:01going to resolve today. Should we move on to our final
- 20:06story for the day? And let's back over to you for
- 20:08that one, Matt. Yeah, and I think that this one
- 20:11is actually sort of follows on quite nicely because two AI
- 20:14stories have landed this this week where and sort of explains
- 20:18why framing is everything. Narble Labs, a Bristol startup,
- 20:21ran a billboard campaign at Bristol airport for the its
- 20:25communications AI agent Deep Blue OS.
- 20:28The ads featured a woman along slide alongside slogans
- 20:32including Outworks, everyone and she'll never ask for a raise.
- 20:36Working 9:00 till 5. She works 24/7 and she starts
- 20:40for free. The ads were pulled after
- 20:42multiple complaints to the Advertising Standards Authority
- 20:45and an LSE gender studies researcher called it the
- 20:49deliberate commercial commercialization of patriarchy.
- 20:52The company. The company's CEO Luke Sartain
- 20:55defended the campaign by saying the real question is about
- 20:57defining the role of humans in a world where we are no longer the
- 21:01most efficient option. Unfortunately, like so many CEO
- 21:06statements on these sorts of things, the framing didn't help.
- 21:08Meanwhile, the BBFC announced it used a bespoke AI tool to
- 21:12classify the entire HBO Max UK catalogue, about 35,000 hours of
- 21:17content. And it did it in just six
- 21:19months, a task that would otherwise have required 1570
- 21:25compliance officer working days. The response was broadly
- 21:28positive. No ASA complaints, no protests.
- 21:32The BBFC was careful to emphasise that humans retain
- 21:35full responsibility for all final ratings.
- 21:38So what do you make of this nails?
- 21:40I think, first of all, that I think that would actually be
- 21:43quite an enviable job to classify the entire HBO Max UK
- 21:46catalogue. And I'm guessing that The Pit
- 21:49must have accounted for a good chunk of those hours.
- 21:51That is a Have you watched The Pit?
- 21:53I have not had time. Well, it, it does, it is an
- 21:57investment 15 hours per season, but it's great.
- 22:00So, so I would, I would heartily recommend you do it.
- 22:02And back onto the topic at hand, this defence.
- 22:07So I think the, the the CEO of what was the Narwhal Narwhal
- 22:14Labs perhaps attempted to frame this as trying to open a
- 22:17discussion about the true nature of AI in the future of work and
- 22:21so on. I colour me cynical.
- 22:25I do find it a little difficult to believe that that was a a pre
- 22:29planned move that this was this was always their intention.
- 22:32You've seen the ad to write the Billboard in Bristol.
- 22:35To me it looked a bit hasty and clumsy and and at some level it
- 22:42seems surprising that it was signed off.
- 22:44But hey, you know, young and fast moving companies make
- 22:47mistakes. What I would say though that
- 22:50what it's one of those examples of to use the expression saying
- 22:53the quiet part, part out loud. I am sure that many companies
- 22:58are talking in these terms, particularly publicly traded
- 23:00companies that are under enormous pressure to increase
- 23:04margins, to preserve and increase margins.
- 23:06And if they can do so by making big savings in the workforce,
- 23:11then unfortunately that is something that is generally
- 23:13rewarded on on Wall Street and in the city and on stock
- 23:16exchanges. I think that most companies are
- 23:19not saying this yet, and I would hope that many companies are
- 23:23still not really thinking in those terms.
- 23:26But to imagine that this is not a conversation that is going on
- 23:30at the corporate level at many companies around the world, I
- 23:34think is sadly fanciful. That's that's kind of how how I
- 23:38see it. Yeah.
- 23:39So I think that there's two separate pieces here. 1 is the
- 23:42fact that the the nature of work is going to change.
- 23:44I think we can both say that the nature of what we do for a
- 23:47living has changed dramatically thanks to AI, and hopefully
- 23:50we'll end up still having a living in the future.
- 23:53The other side to this, though, is that this was a marketing
- 23:56failure of epic proportions. You just have to look at the
- 24:00fact that the story that they're telling and who they're selling
- 24:02it, selling it to, nobody wants that narrative of hey, we're
- 24:06going to replace you. That's just epic failure of how
- 24:10to message it make you know, you can frame that story so
- 24:14differently. It could be your system that
- 24:17carries on working when you're not in the office.
- 24:20Even that small shift could make a big difference.
- 24:24But then also, why do you need to put a, you know, somebody's
- 24:27face on that? You don't.
- 24:29It's just a mess of epic proportions.
- 24:32There's so much wrong with all of that marketing message.
- 24:35The story where the ads were placed, the the execution.
- 24:38It feels like it was developed by a group of tech Bros that
- 24:42have asked an AI to help it put together some advertising.
- 24:46And that's what it looks like. That's what it feels like, and
- 24:48it's what it sounds like. And as a marketer that works in
- 24:51AI, I'm really disappointed in the execution of that.
- 24:54Felt right no one when you tell us how you really feel.
- 25:00Who knows it could be a mix of all of the above, but I mean for
- 25:03me, I think so so they could retort and and and other
- 25:06companies could retort again, we're just saying what people
- 25:10are thinking and you know there there will be elements of of
- 25:13truth to that. But I think I agree with you as
- 25:15well. I think that making this an
- 25:17acceptable messaging that the more it becomes commonplace, the
- 25:20more it becomes an acceptable idea.
- 25:22I think and, and for me, I think for me, it is absolutely
- 25:26inevitable that AI is going to have a profound and not positive
- 25:29effect on the jobs market, at least in the in the medium term.
- 25:32In the long term, who knows. But if it becomes a kind of
- 25:37common currency to describe these situations as getting rid
- 25:41of people is a virtue, this is get really great, get rid of all
- 25:45your workforce wonderful. Then it it kind of starts to
- 25:49starts to kind of desensitise people to what's actually
- 25:51happening. And I think that's a terrible
- 25:53thing. I mean, I think this is, this is
- 25:55going to be a really challenging situation for many people to
- 25:58deal with anyway. So to make light of it and to
- 26:00make a virtue out of it, I think it's something I'm certainly far
- 26:04from comfortable with. We've seen where that took the
- 26:07likes of Klarna when they got rid of, you know, a huge
- 26:10proportion of their developers only to need to rehire many of
- 26:13them. Yeah.
- 26:15It's not. The narrative isn't quite there
- 26:18yet. And I think that they missed the
- 26:19point on the entire way to to message that, you know, and
- 26:23that's in contrast to taking a job like classifying, you know,
- 26:27HBO Max's, you know, back, you know, catalogue.
- 26:30You know, that's a job that you and I might have enjoyed
- 26:33actually sitting down there watching them catalogue in those
- 26:37those films, but it's going to cost them a lot of money that
- 26:39they probably didn't have. I think it's a question of how
- 26:42do we use AI and how should we use AI?
- 26:44This is a tool that we have a choice about.
- 26:46And I think that that's something that needs really
- 26:48careful attention as we, as we move forward, you know, with the
- 26:51power of something like mythos with, with where we choose to do
- 26:54compute and where we, you know, and how we use it and how we
- 26:57communicate it. I think it's AI think we're in a
- 26:59really, really exciting time and one where we're going to see
- 27:03both a lot of big hits and a lot of big misses.
- 27:06On that note, I couldn't agree more.
- 27:08That seems like a good point to wrap up for this week.
- 27:11As ever, thanks for talking through the issues with me back.
- 27:15Thanks Nails, it was good fun. And we'll see everybody next
- 27:18week. And that's it for this week's
- 27:29episode of AI in the UK. If you found it useful, you can
- 27:35subscribe wherever you get your podcasts and we'd really
- 27:38appreciate a great rating or review.
- 27:41It genuinely helps others find the show.
- 27:44You can also follow Neil's Footman and Matt Wilkinson on
- 27:47LinkedIn for links, notes and occasional extra content.
- 27:52Thanks for listening and we'll be back next week.
- 27:55Helping you make sense of AI as it lands here in the UK.