Latest / AI in the UK / Good Week, Bad Week, Genuinely Complicated Week
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.
- 0:55The UK this week announced a win on AI investment while quietly
- 1:00kneecapping early stage funding, watched open AI freeze its data
- 1:05centre plans and started drafting love letters to
- 1:08Anthropic. Also, Oxford built an AI that
- 1:11can see heart disease before your cardiologist can.
- 1:14Good week, bad week, genuinely complicated week.
- 1:17Welcome to AI in the UK and over to you Neil's for our first
- 1:22story. Yeah.
- 1:23This is a topic like many that we've discussed, that we touch
- 1:27on again and again and combines what I think is some really
- 1:30encouraging news with bits that are rather more complicated.
- 1:34So this week brought a signal that the UK's long term bet on
- 1:38patient capital may be starting to pay off.
- 1:41The British Growth Partnership, a new 200 million investment
- 1:45backed by UK pension funds including M&G and Aegon, has
- 1:49made its first investment, £8,000,000 in Wave, the British
- 1:55autonomous vehicle company that we've discussed recently, who
- 1:57will be rolling out autonomous taxis on the streets of London
- 2:01by the end of the year. The number itself is fairly
- 2:05modest, but the structure behind it matters.
- 2:08Over the past two decades, UK pension funds have steadily
- 2:11reduced their exposure to domestic companies.
- 2:14Currently less than 5% is invested in UK equities, making
- 2:18Britain a global outlier. The allocation to unlisted
- 2:22equities, where early stage tech companies sit, stands at just
- 2:260.6% of default pension funds, against a government target of
- 2:305% by 2030. Getting pension funds into a
- 2:35structured VC vehicle at all is a meaning meaningful shift, and
- 2:39not just a press release. It's the Mansion House
- 2:42framework, about which we've heard quite a lot.
- 2:44And that this actually stems back to Jeremy Hunt's time as
- 2:47Chancellor. I believe the government
- 2:49sustained push to redirect pension capital into UK growth
- 2:53companies is delivering something concrete.
- 2:56If it scales, this is the kind of patient domestic capital that
- 2:59could change the funding calculus for UKA companies at
- 3:03the growth stage, possibly reducing the pressure to seek
- 3:07U.S. investment and with it the risk of losing British companies
- 3:11to American ownership. The same week, however, the
- 3:14government's new entrepreneurship package came
- 3:16into force. It expands equity schemes for
- 3:19scale ups and introduces new relief for UK listings.
- 3:24But it also cuts tax relief for venture capital trust investors
- 3:28from 30% to 20%, one of the primary routes through which
- 3:32individuals back UK startups at the earliest stages.
- 3:37The last time this relief was cut in 2006?
- 3:40VCT funding fell, fundraising fell 65% year on year and took
- 3:47over a decade to recover. This matters particularly for AI
- 3:51companies, which typically require sustained capital across
- 3:54multiple funding rounds before generating meaningful revenue,
- 3:58high compute costs, long development cycles, and a market
- 4:02that hasn't yet matured for quick monetization.
- 4:05I, We seem to begin talking about all these stories by
- 4:08saying in a mixed bag of complicated lot to unpack, but
- 4:11it's very much the case here as well.
- 4:12I would say this thing about employing the, the potentially,
- 4:17I believe it's trillions that UK pension funds and companies
- 4:22manage and getting it invested in British companies in the tech
- 4:25sector is exactly what we need. But the on the, on the flip side
- 4:30that the restructuring of venture capital tax relief would
- 4:35appear to be less favourable. Matt, what was your take on all
- 4:39this? Yes, numbers are pretty sobering
- 4:42really. We're at .6% of default funds in
- 4:45unlisted equities against a target of 5% and a company
- 4:51that's already had quite significant investment wave
- 4:55teething a little bit more. It doesn't seem that it's it's
- 5:01even a drop in the ocean compared with the scale of
- 5:04capital flowing into USAI companies.
- 5:07We're talking orders of magnitude difference, and while
- 5:10it's a positive step in the right direction, it's like
- 5:13saying that a positive step when you're trying to cross the
- 5:17Atlantic is a step in the right direction when really all you're
- 5:20doing is getting your toes wet. I kind of feel that when we step
- 5:23forward, we're taking at least half a step backwards here.
- 5:26We don't seem to have an integrated strategy for really
- 5:29getting the money into the places where it's really going
- 5:32to make a difference to this country.
- 5:34That there is obviously some truth to what you say it is.
- 5:39You can't deny the amount of money and the overall funding
- 5:41for Wave is at this stage still modest. 8,000,000 out of it was
- 5:46an enormous round and backed I believe largely by NVIDIA and
- 5:49other companies of that scale. Of that I think it would be hard
- 5:53to debate, but I think what really matters is the direction
- 5:56of travel. I don't imagine any scheme like
- 6:00this would begin with an enormous investment upfront.
- 6:03This is still a big structural shift for how these funds have
- 6:07been managed and the mandate of pension fund managers actually
- 6:12having this mandate now to invest in what are considerably
- 6:17riskier endeavours. It's not something that you can
- 6:20just switch on and off. I know, like I said, that the
- 6:22government has been talking about this since towards the
- 6:25tail end of the the last Conservative government when
- 6:27Jeremy Hunt was pushing this very strongly.
- 6:30But it's taken a bit of time, Yes.
- 6:32But for me, this is tremendously encouraging that, that it's,
- 6:34it's started and you would feel now they've got there as, to use
- 6:39your metaphor, now they've got their their toes wet.
- 6:42It's I think the beginning of something much bigger.
- 6:45I mean, who knows for certain, but someone has to had to get
- 6:48the ball rolling and do. The amount for this specific
- 6:51investment is cited as 8 million, which is in these
- 6:55worlds a modest sum. Aegon itself, I believe, has
- 6:58committed 100 million to this fund, which is a far more
- 7:02serious token of intent. Yeah, I think you're right about
- 7:05that 100 million some from Aegon being committed already.
- 7:10I think it's also positive that we've got that EMI expansion
- 7:13listing relief, you know, they all benefit companies that are
- 7:16scaling and that's that's really important.
- 7:19But the VCT expect, you know, the, the VCT cut in the same
- 7:24announcement though just feels like it's a step backwards.
- 7:27As soon as we're taking one forward, you know, that's 65%
- 7:31being, you know, of drop in VCT fundraising straight after the
- 7:36last cut is really sobering. And so I'd be quite worried
- 7:40about making those two things and saying well, does this, does
- 7:44this one move on one hand outweigh the other?
- 7:47I know that only time will tell, but bit concerned about the sort
- 7:50of the financial. Impact here, fair enough, but
- 7:55this is kind of a measure of measure of 1 cynicism I expect
- 7:59about what what the actual intentions were behind that
- 8:01move. So I think the tempting
- 8:04narrative for this one is they cut that text relief in order to
- 8:07boost revenues in in an exceptionally fiscally difficult
- 8:12moment. And on the face of it, that
- 8:14would seem to be giving with one hand and taken away with the
- 8:17other. Now believe, I believe the
- 8:20government's response has been actually this is part of a, a
- 8:23broader plan that if you reduce the incentive for early stage
- 8:28developments among venture capitalists and you also are
- 8:31pushing this, also pushing the pension fund investments and,
- 8:35and getting them to contribute more, you actually are
- 8:39corralling investors more towards late stage capital.
- 8:43So by that reading, you know, by the government's position, it is
- 8:47part of a plan to move resources from early stage to more towards
- 8:54a late stage. So, so supporting this is, I
- 8:57mean, I think there is a legitimate argument that VCTs
- 9:03became relatively low risk by focusing on early stage
- 9:08investment and they were operating more as kind of tax
- 9:10back vehicles. We, we know, without wishing to
- 9:13get drone in acronyms, but the EIS and SEIS schemes, right?
- 9:17The Enterprise Investment scheme, I believe it stands for
- 9:21and for early stage investors. So obviously this is what it's
- 9:24designed to do. It's designed to provide
- 9:27attractive tax incentives to offset a big risk for early
- 9:30stage investment. That's what it's designed to do.
- 9:33The concern has been in some levels, including the
- 9:36government, that it's become such an incentive that people
- 9:40are reluctant to take any risk at all.
- 9:42They just say, right, this is an attractive investment for me
- 9:45because I'm getting very favourable tax conditions.
- 9:47And by making it somewhat less attractive on the tax front, the
- 9:50idea is that it'll push them more towards these higher
- 9:54return. Now the the evidence for this is
- 9:55that these start up investments in the UK are very heavily
- 9:58focused on certain certain areas.
- 10:00Fintechs, for example, are, are are enormously successful,
- 10:06successful investments in the UK.
- 10:08So I mean, I think as with all these things, it's very
- 10:11complicated and assigning a simple motive to one party or
- 10:15another is, is fiendishly difficult.
- 10:17I would just repeat, I think the direction of travel is
- 10:20encouraging, despite the reduction in tax relief.
- 10:25Yeah, the Mansion House money is definitely growth stage capital
- 10:28and that is incredibly exciting. We've spoken, you know, at
- 10:32length about the challenges of growing and keeping companies,
- 10:37you know, UK based and funded. The concern would be that do you
- 10:41then, you know, at a certain point of time, do you run out of
- 10:45good bets to put into that growth stage capital?
- 10:49You know, if you cut the VCT pool too much, then does the
- 10:53does the Mansion House money really kind of have anything to
- 10:56invest in? You know, it's pretty clear that
- 10:58AI funding across Europe is, is now outpacing that of fintech,
- 11:04understandably so. Although you could argue that
- 11:06maybe many of the Fintech investments are now being
- 11:09relabeled as AI investments, particularly those from the UK
- 11:12where Fintech is such a huge part of our industry because of
- 11:16the great work that the Bank of England and others have done to,
- 11:20to sort of really build a financial centre here in the UK.
- 11:23So it's, it's definitely complicated.
- 11:25Only time will tell how that financial modelling goes.
- 11:28I just really hope that it goes in the right direction.
- 11:31On that we can unequivocally agree and we'll be keeping a
- 11:34close eye on how these pension fund backed investment funds
- 11:38that go from here. Talking about the mismatch
- 11:40between capabilities and reality.
- 11:42We've got a couple of interesting stories about some
- 11:46of the really big USAI companies and how they are interacting
- 11:51with us here in the UK. Matt, over to you.
- 11:54Yeah, thanks, Neil's in the same week, 2 developments landed
- 11:58together almost at the same time to define that, you know, this
- 12:02define the tension inside the UK's AI ambition.
- 12:05The government has started actively courting Anthropic, the
- 12:09$300 billion Frontier Lab behind Claude Look, trying to get them
- 12:14to expand their London footprint and pursue a dual Stock Exchange
- 12:18listing, having identified A strategic opening after the
- 12:22Pentagon branded Anthropic a supply chain risk for refusing
- 12:26to strip safety guardrails from its models for military use.
- 12:30The dispute centres on Anthropic CEO Dario Amadei's refusal to
- 12:35allow Claude to be deployed for lethal autonomous weapons or
- 12:38mass domestic surveillance. President Trump directed federal
- 12:42agents to cease use of Anthropics technology.
- 12:47A federal judge has since temporarily blocked the
- 12:49blacklisting. With the appeal pending in the
- 12:529th Circuit. UK officials backed by Downing
- 12:55St and London Mayor Sadiq Khan have drafted proposals for
- 12:59Amadei ahead of a scheduled late May visit.
- 13:02So we're sending love letters to anthropic addicts almost exactly
- 13:06the same time as Open Air has decided to pause Stargate UK,
- 13:12it's data centre project that was announced with much fanfare
- 13:15alongside NVIDIA and N Scale last year last September as part
- 13:20of a $31 billion tech investment.
- 13:23The stated reasons? Regulatory uncertainty and
- 13:26industrial electricity costs running four times higher than
- 13:30in the US. Open AI's language is
- 13:32deliberate. It's a pause contingent on the
- 13:35right conditions, not an exit. But the symbolism is damaging
- 13:39for government. This stakes its growth agenda on
- 13:42AI leadership. Now it's interesting to note
- 13:44that Open AI has also had a similar sort of time pulled back
- 13:49and actually shut its sore to a video creation app and it's
- 13:53really refocusing. So this could just be a story
- 13:55about open AI and a refocus of where it allocates its capital
- 14:01and trying to become, you know, more focused on its core
- 14:04capabilities. But it could be could talk about
- 14:06right, a wider issue. Read together, these stories
- 14:09raise a harder question than any individual headline.
- 14:13UK is simultaneously seen as a safety friendly alternative to
- 14:17the Washington and as an infrastructure environment that
- 14:20is too expensive and regulatory with with too much regulatory
- 14:23uncertainty for large scale compute investment.
- 14:26You know, that's not obviously a contradiction, but it's
- 14:29genuinely difficult position to hold.
- 14:32Really curious to see what you think about this nails.
- 14:35I think when we saw the story about anthropic, there was a
- 14:39real ripple of excitement. I mean, the prospect of
- 14:43Anthropic and I know we're both massive fans of Claude, like so
- 14:45many people increasingly are these days moving to the UK or
- 14:50to London. Yeah, obviously very exciting.
- 14:51It was broken. That's that story was broken by
- 14:54the FTI believe, which clearly adds it some credibility.
- 14:58But I have to say I I've got to say when push comes to shove, I
- 15:02don't think it will happen. I think, I can't imagine I'm
- 15:05actually uprooting and moving here entirely.
- 15:09I can imagine them having bigger operations here perhaps, and
- 15:13moving some functions here. But the idea of them moving
- 15:16entirely, I think as pleasing it would be, both for the
- 15:20development of the UK tech scene and for the look on Donald
- 15:22Trump's face when he reacts to that news with incandescent
- 15:25rage, would be quite amusing. I think.
- 15:28To be honest, it's the actions of Open AI that are the more
- 15:32significant signal. We've discussed a few times
- 15:35about the government's grand plans and pronounces.
- 15:38In fact, I think we were talking on this very topic.
- 15:40Was it last week or the week before about.
- 15:44Yeah, about how done sort of quite a grand sounding joint
- 15:48announcement with Open AI, with all these very ambitious
- 15:51sounding projects and eight months on, nothing's really
- 15:53materialised. I totally agree that that there
- 15:56will be various things behind this energy costs.
- 16:00I mean, you'd have to say, yeah, sure, that's that's an entirely
- 16:03believable part of the equation. I think when he was talking
- 16:09about regulatory uncertainty, I don't know what his exact words
- 16:11were, but I gather he was alluding to the copyright issue.
- 16:16Another story we've covered a couple of times and how the UK
- 16:20has backed off from let me get this right and opt in has backed
- 16:26away for at least for now from an opt in option for copyright
- 16:29rules, which makes things a lot more complicated for them.
- 16:31And of course, yeah, open AI is, I would say if you're looking at
- 16:35the business side of it, it's a tough time for it.
- 16:38I mean, I think until fairly recently, there was a kind of
- 16:41inevitability that it was going to carry on being the flag
- 16:44bearer for AI in general. The other ones are getting
- 16:46bigger, but open AI was still so far ahead.
- 16:48They are really feeling the heat from anthropic.
- 16:51Anthropic has made enormous gains in in monthly annualised
- 16:55revenues. So for all those reasons it's on
- 16:59pause. And I think that just goes to
- 17:01show how delicate and prone to external forces any UK plan for
- 17:08any kind of AI sovereignty really is.
- 17:10Yeah. It does feel that the the
- 17:12Stargate pause is sort of part of that bigger open AIIPO story.
- 17:16And I think part of that is really sort of the
- 17:18infrastructure stories close. You know the company's managing
- 17:22its cost based ahead of public markets and it's pulling
- 17:25discretionary CapEx. That's a corporate finance
- 17:28decision, potentially not a verdict on the UK long term, but
- 17:32I think it does speak to, you know, that big a challenge
- 17:36within open AI where one they're struggling with to make the
- 17:39books look attractive ahead of the open, you know, the IPO
- 17:43itself. But I think that there's also
- 17:45they've they've been leaking leaders to, you know, in fact,
- 17:49it's no, you know, it's no secret that anthropic and was
- 17:52formed after people that left open AI.
- 17:55There's been a lot of commentary about Sam Altman and his
- 17:58trustworthiness in the recent. Commentary is doing a lot of
- 18:01work. Yeah, Commentary is doing a lot
- 18:04of work there, but there's a lot of stories out there about what
- 18:08Open AI is doing and how it operates.
- 18:10I do see that that organisation is sort of going through one of
- 18:13those challenging moments that organisations go through.
- 18:17But one of the things that Sam Altman has said, which I thought
- 18:20was really interesting, is that they've been really looking at
- 18:24these, you know, these data centre investments because they
- 18:27can see very quickly how organisations, including
- 18:30Anthropic are going to becoming very compute constrained as they
- 18:34look to build the next generations of models.
- 18:37Having said that, Anthropic has just announced the really and
- 18:43sort of confirmed the the existence of a new model called
- 18:47Mythos, which has been getting a lot of attention because just
- 18:51how good it is at coding. They didn't train it to be a
- 18:55cybersecurity tool, but it's actually able to.
- 18:59It's been finding vulnerabilities in software that
- 19:04have existed for decades and it's finding more in minutes
- 19:07than most organisations find in years.
- 19:10So it really is an incredibly powerful model that they've
- 19:13built. They've only opened it up to a
- 19:15very small number of organisations so they can try to
- 19:19protect the critical infrastructure on which so many
- 19:22of our software products are built before they release this
- 19:25in the world. So it's pretty scary how quickly
- 19:29even a a compute constrained organisation like Anthropic, as
- 19:35Sam Altman would call them, able to build models that are
- 19:38staggeringly powerful. On the subject of how this is
- 19:44panning out in the UKI mean, do you imagine the anthropic thing
- 19:47will actually happen? Were you concerned about what's
- 19:50going on with Open AI? I think that the open AI
- 19:52investment will happen when the time's right.
- 19:55I can see them wanting to use this partly as a play to make
- 19:59sure that. Their models are in government
- 20:02systems. That's that's the way I read it
- 20:05the first time. In terms of Anthropic, I could
- 20:07see a dual listing, although I think that that's unlikely, but
- 20:10I can see them. It would be easy to see them
- 20:13increasing the amount of research they do in the UK based
- 20:16off of all of the deep mind history that we've got here.
- 20:19After all, we've got a great history of building AI and
- 20:24helping to shape the field. So it would make sense that that
- 20:28that organisations at the forefront of this are trying to
- 20:32tap into that part of the knowledge base.
- 20:35Yeah, absolutely. The AI starts up scene across,
- 20:39certainly across London and you know moving into Cambridge.
- 20:41Notchwood, as we've said many times is very strong, deep mind
- 20:45as resulted in a very kind of robust and in depth ecosystem
- 20:49here in the Uki would just say, yeah, as far as a complete move,
- 20:54I still find it quite far fetched.
- 20:56And and just to sign up with a little dig at the government as
- 20:59a possibly question how they've handled this.
- 21:02I think that the whole issue around the copyright opt out was
- 21:07was a factor in attracting investment from some of these
- 21:10big companies. Now, whatever your whatever your
- 21:13position on the copyright laws in themselves, and I've got
- 21:16quite mixed feelings about them. I've come from a kind of
- 21:19creative market marketing bank and myself.
- 21:22That's the government was backed off from that position as
- 21:26quickly as it did that it wasn't prepared to face them down I
- 21:28think will have been noted in the boardrooms of these enormous
- 21:32companies, these enormous AI companies.
- 21:34So you may very well feel that's a good thing and that, you know,
- 21:38the government was responding to popular pressure from the
- 21:40creative industries and then there is a very legitimate
- 21:42argument to be made for that. But at the same time, these
- 21:47decisions as they relate to corporate governance, they, they
- 21:50have consequences and investors will make decisions accordingly.
- 21:54And I, I would guess on some level that's what's happened
- 21:57here. And if push comes to shove,
- 21:59it'll be something of an impediment to anthropic coming
- 22:02here as well. Yeah.
- 22:04And let's just hope that we get to a point where we build a, you
- 22:08know, a regulatory framework that is appealing to these
- 22:11companies being here because we've got such great talent.
- 22:14And, you know, the combination of the AI, you know, legacy as
- 22:18well as the life sciences really does bode well.
- 22:22And especially when you're looking at organisations like
- 22:25Anthropic, having built Claudeford Life Sciences, that
- 22:28seems to neatly move on to a story coming out of Oxford
- 22:32University, I think that you're about to tell us about nails.
- 22:35Yeah, it's always great to talk about these really important and
- 22:41potentially groundbreaking discoveries here in the UK So
- 22:44researchers at the University of Oxford have developed an AI tool
- 22:47that can identify patients at high risk of heart failure up to
- 22:50five years before the condition develops.
- 22:52The tool works by analysing changes in the texture of fat
- 22:55tissue surrounding the heart, changes visible on a standard
- 22:59cardiac CT scan but that the human eye typically misses.
- 23:03It was trained on anonymized scan data from more than 59,000
- 23:07patients across 9 NHS trusts, then validated on a further
- 23:1113,000, achieving 86% accuracy in predicting 5 year risk.
- 23:17Patients flagged as high risk were found to be 20 times more
- 23:20likely to develop heart failure than those in the lowest risk
- 23:23group, roughly one in four, going on to develop the
- 23:26condition within five years. With their own 350,000 cardiac
- 23:31CT scans carried on the UK each year, the researchers say the
- 23:34tool could act as an opportunistic screening layer,
- 23:37flagging at risk patients without any additional tests or
- 23:40appointments. The team is now seeking
- 23:43regulatory approval to roll it out across the NHS.
- 23:47This isn't the first time the same Oxford group has taken this
- 23:49kind of research to market. Lead researcher Professor Sheryl
- 23:53Ambos Antoniades cofounded Caristo Diagnostics in 2018, a
- 23:59spin out that has since raised over £23 million and gained FDA
- 24:02clearance for a related cardiac CD tool.
- 24:06Right. Much to be happy about here, I
- 24:07would say shows that innovation with real life and potentially
- 24:14very consequential tech is, is taking place here.
- 24:17But also this, this thing. We've frequently struggled with
- 24:20commercialising scientific discoveries.
- 24:23Now, Matt, as our resident science correspondent, I think
- 24:27you should be perhaps explaining this a bit more than I will.
- 24:31So let me kick off with a question for you.
- 24:33So in this tech, in this development, what is the AI
- 24:36actually doing and why can it see what humans can't?
- 24:40So taking a little bit of a step back, what happens is, is that
- 24:45the fat tissue surrounding the heart changes texture when the
- 24:48heart muscle beneath it becomes inflamed or is under stress.
- 24:51Those textual changes are invisible to the human AI, you
- 24:56know on ACT image. But what the AI is doing is
- 24:59picking up on subtle statistical patterns across thousands of
- 25:03data points in the image that have no visual signature.
- 25:06Cliche, a clinician could really be trained to spot.
- 25:09So it's not augmenting human perception.
- 25:11What it's doing is operating in a space that the humans can't
- 25:15access at all. So to put it on the way, would
- 25:19you say so rather than just AI helping doctors read scans
- 25:23faster, it's actually finding signals that were previously not
- 25:28detectable at all? That's certainly the more
- 25:31interesting claim. Yeah.
- 25:32And what makes the training data so important is that 59,000
- 25:35patients? Oh, sorry.
- 25:36Yeah, 59,000 patients is a large enough data set to find patterns
- 25:41that would be completely invisible in a study of just a
- 25:43few 100. So they've got a huge amount of
- 25:45data already and these are tests that are already run.
- 25:48So you can imagine that this capability would be pretty easy
- 25:53to be able to go and roll out because you're not actually
- 25:56asking for more time in the lab or more time with the doctors
- 25:59themselves. These scans could be automated
- 26:03or processed sort of automatically, giving people a
- 26:06risk score and then flagging them for clinical follow up.
- 26:11Medicine is one of the areas where AI is thought to have some
- 26:15of the greatest potential purely in terms of being able to look
- 26:19over enormous reams of data and, and, and address this kind of
- 26:24needle in the haystack in a haystack problem.
- 26:28So for this, so this announced technology, the team is already
- 26:32seeking NHS regulatory approval. But how far along are we?
- 26:38Would you say really, is this ready?
- 26:40Is this something that's going to be in general use anytime
- 26:43soon? Well, the 86% accuracy as long
- 26:46as that is something that enables that these additive.
- 26:49So we're looking at five year horizons and we're able to look
- 26:52at something that will increase our ability to diagnose disease
- 26:56ahead of time. Then I think that that's it.
- 26:58That's that's probably cut pretty close to something that
- 27:02we would want to roll out and look at.
- 27:05Typically what happens is you go through these regulatory
- 27:08processes and they may want to see more data.
- 27:11The likely next step is that you'll be having, you know,
- 27:15rather than retrospective trials, which is the what this
- 27:17was done. It was done.
- 27:18Only, you know, that the AI was trained and then validated
- 27:23using, you know, historical data or pre-existing data.
- 27:27What they'll now be looking to do is to look at this in a real
- 27:30world situation and see, you know, really measure the long
- 27:33term outcomes. But it does look like it's, it's
- 27:37something that's particularly interesting.
- 27:39And it's important to know that there are about 350,000 of these
- 27:43scans a year. So if we do run this, you know,
- 27:46roll this out across the country, then we've got a much
- 27:49bigger data set to even, you know, to start improving the
- 27:51model on. And that's where this gets
- 27:53really exciting. OK, and a question for you, but
- 27:57maybe the last one of the episode.
- 27:58So the so the tool is understanding, so it predicts
- 28:01who gets heart failure. And you know, this is something
- 28:03I've always there's always stuck with me as very much someone who
- 28:05doesn't who's never really studied this area.
- 28:07But when he's when you get news stories talking about some new
- 28:11technological development, in this case an AI1 makes it much
- 28:14easier to predict conditions you are likely to suffer from.
- 28:19I just wonder is that knowledge I would really want anyway.
- 28:21I mean, the the real question surely is what then happens with
- 28:25with that? Does does knowing this, with
- 28:27that time horizon actually help? Does it change that, change what
- 28:32people can do and affect the outcomes?
- 28:35Yeah. So cardiovac, you know,
- 28:37cardiovascular disease, which includes heart failure, is the
- 28:40world's leading cause of death. You know, it causes about 20
- 28:44million deaths a year. So a major subset of that is
- 28:48heart failure. So hopefully, you know, this cat
- 28:50could capture an awful lot of that.
- 28:53You know, in the UK over 1,000,000 people live with heart
- 28:56failure, which is a major cause of hospitalizations.
- 28:59So if we can look at reducing the risk of that and there are
- 29:05certain things that we can do to, you know, lifestyle changes
- 29:08that we can, we can put in place.
- 29:10There are certain drugs that we can prescribe as well, doctors
- 29:13can prescribe to help reduce the risks of heart failure.
- 29:17You know, simple, you know, key risk factors includes smoking,
- 29:20high blood pressure, high cholesterol, obesity, poor diet.
- 29:23So by making changes to some of those, you know, if we're
- 29:26flagging the risks to somebody and we're able to proactively
- 29:30make a change, we know these things are important to do
- 29:33anyway. But but having the same, you
- 29:35know, having the having the knowledge that you're now in a
- 29:37high risk group and you've got five years, you know,
- 29:40essentially that you've got five years to sort of turn this round
- 29:44or somewhere and you may need to make that change quickly.
- 29:46That's I think that's a pretty compelling argument for saying,
- 29:49yeah, done these scans, you're at risk, you need to make some
- 29:53changes. So if we can proactively manage
- 29:55health rather than and if we can turn sick care into health care,
- 30:00then I think we're in a much stronger place and hopefully
- 30:04reduce costs on the NHS and that, you know, that, you know,
- 30:07that burden of hospitalizations and you know that taxpayer
- 30:10burden on, you know, on the country.
- 30:12Yeah, yeah. It seems to me this is an area
- 30:14where AI could be an almost unequivocal good.
- 30:17If you could get something like that, that really does improve
- 30:20outcomes, reduces pressure on the NHS, reduces costs and
- 30:23especially reduces increasing costs on the NHS, that would be
- 30:27a huge step forward for us all. Yeah, I mean, obviously if
- 30:31obviously if obesity is one of those things, then this could
- 30:34also be good news for the all the companies selling GLP one
- 30:37the weight loss drugs as well, just as a bit of a fun side of.
- 30:41Sure, sure. And we all want that, don't we?
- 30:43Exactly. All right, on that note, it's a
- 30:47good place to sign off for this week.
- 30:49Thanks, as ever for talking us through the issues.
- 30:51Matt. Pleasure.
- 30:52As always, Neil, thank you as well.
- 30:54All right, bye for now. And that's it for this week's
- 31:07episode of AI in the UK. If you found it useful, you can
- 31:12subscribe wherever you get your podcasts and we'd really
- 31:15appreciate a great rating or review.
- 31:18It genuinely helps others find the show.
- 31:21You can also follow Neil's Footman and Matt Wilkinson on
- 31:24LinkedIn for links, notes and occasional extra content.
- 31:30Thanks for listening and we'll be back next week.
- 31:32Helping you make sense of AI as it lands here in the UK.