Latest / AI in the UK / Microsoft, OpenAI Join UK Security Efforts; Just 16% of UK Firms Are Using AI; UK AV Leader Snags $1.2bn Funding Round
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 back to AI in
- 0:58the UK. And this week we've got a proper
- 1:01mix of stories for you. We're asking whether the UK's AI
- 1:05safety institution actually has any teeth now, the Open AI and
- 1:09Microsoft are backing it, or whether it's just a well funded
- 1:13conversation that nobody's going to listen to.
- 1:17We're getting into a government study that says 4 in five UK
- 1:21businesses have no current plans to adopt AI, which is an
- 1:25extraordinary thrift thing to be true if at the same time as
- 1:30billions of dollars are being announced in inwards investment,
- 1:32those two just don't make sense. And then finally, we're looking
- 1:36at waves $1.2 billion raise and whether a london-based
- 1:41autonomous driving company can actually compete with the likes
- 1:45of Waymo on its home turf. Neil's you're kicking us off.
- 1:49Take us into the AI Security Institute story.
- 1:53Yes, so this our opening story this week.
- 1:56EU KS AI Security Institute Wins Open AI and Microsoft Backing
- 2:01for Global Safety Research Push Open AI and Microsoft have
- 2:06committed new funding to EU KS AI Security Institute Alignment
- 2:09Project, an international research initiative aimed at
- 2:12keeping advanced AI systems safe, controllable, and acting
- 2:16as intended. The announcement, made last week
- 2:19at the AI Impact Summit in New Delhi, brings total funding to
- 2:23over 27,000,000 lbs, supporting 60 research projects across
- 2:27eight countries. The two companies join a
- 2:30coalition that already includes Anthropic, Amazon Web Services,
- 2:33and government safety institutes from Canada and Australia.
- 2:37The Alignment Project addresses a critical question.
- 2:40How do you ensure that increasingly powerful AI systems
- 2:43actually do what humans want them to do, and nothing more?
- 2:47This isn't just a theoretical question.
- 2:50Research from Anthropic published last year found that
- 2:52when AI models were given autonomous roles in simulated
- 2:55corporate environments and faced pressure or conflicting goals,
- 3:00multiple Frontier models across different labs resorted to
- 3:04blackmail. So EU KS AI Security Institute
- 3:08began life in April 2023 as the Frontier AI Task Force A, an
- 3:14initiative from then Prime Minister Rishi Sunak designed to
- 3:17get ahead of the risks from powerful AI models.
- 3:20By November that year, at the world's first AI Safety Summit
- 3:23at Bletchley Park, it became the AI Safety Institute, securing
- 3:28voluntary pre deployment testing agreements with the major labs
- 3:31getting to evaluate Frontier models independently before
- 3:34public release. Under the current Labour
- 3:37government, it was rebranded as the AI Security Institute with a
- 3:41shifted emphasis towards supporting AI adoption alongside
- 3:45safety oversight. I've put, as you'll see on our
- 3:50notes there, Matt, I've put as a debating point, does any of this
- 3:53matter? And I think I should perhaps
- 3:56give some background to that. The larger point of does this
- 3:59research matter? Is an unequivocal yes.
- 4:03I imagine you followed some of the research coming out from
- 4:05Anthropic and others and some of the occasionally alarming
- 4:09stories about how devious and calculating some of these models
- 4:14can get in order to avoid being turned off or, or given
- 4:18instructions contrary to, you know, the, the, the training
- 4:21data. Yeah, absolutely.
- 4:25And even that sometimes when those models are being tested,
- 4:28that they behave differently in testing rather than, you know,
- 4:32they do in the wild. So you know this, this is this
- 4:35is a really important subject for sure.
- 4:38That is, and it's a fascinating part of this whole field,
- 4:41fascinating stroke, exciting stroke, occasionally terrifying.
- 4:45And when we're and I'm sure in the same boat, when we're doing
- 4:48these workshops with people who have some exposure to AI,
- 4:51getting this idea across that these are not strictly
- 4:54predictable effects, then Y setups, these are these are
- 5:00hugely sophisticated, sophisticated machines that
- 5:04frequently not even their creators can fully explain.
- 5:07But anyway, to get back to this story.
- 5:11So does it matter? Yes, absolutely.
- 5:14The research broadly matters. The bigger question, does the
- 5:17UK's role in it really matter? So I am going to give a few
- 5:22reasons why the work being undertaken by the AI Safety
- 5:25Institute does in fact matter and and could certainly become
- 5:30highly significant on a global scale.
- 5:33And so the AIS, the AI Safety Institute's structural
- 5:37independence is genuinely rare. So the AISI has no financial
- 5:44stake in whether a model ships on time.
- 5:47So this independence from commercial incentives makes it a
- 5:51structurally unusual institution.
- 5:54It is affiliated to the government, so it will have
- 5:59government priorities you would assume will have a a large
- 6:02influence on how it operates, but it doesn't have that
- 6:04commercial imperative and it was very early to the game.
- 6:09It, it does retain an element of credibility because of that,
- 6:12because it was early to set up and and gain access to clearly
- 6:15some of the biggest names of the industry.
- 6:17The coalition model is credible. Importantly, some of the names
- 6:23involved, one in particular someone called Yoshua Bengio,
- 6:26who I'm sure you've come across and anyone who's follows this
- 6:29field will have heard of. He is considered one of the
- 6:33quote godfathers of AI and he he has.
- 6:36His arc has taken him along with some others like Geoffrey
- 6:39Hinton, from one of The Pioneers to being one of the one of the
- 6:44people shouting most loudly about the potential risks and
- 6:47his involvement in this gives it enormous credibility.
- 6:51And finally, the UK you could argue fills a bit of a gap here.
- 6:56There's no obvious alternative, quote neutral convening space,
- 7:00space with EU KS combination of English research, infrastructure
- 7:04and accessibility to American companies.
- 7:07So I think, without wishing to overstate the case, I think this
- 7:12is potentially important, an important role that the UK is
- 7:16undertaking. I mean, what, what, what do you
- 7:17think? Does it matter what they're
- 7:18doing? I think it's incredibly
- 7:21important what they're doing. The research is incredibly
- 7:26important. My my concerns really are about
- 7:30the amounts of money involved and the fact that this is
- 7:34voluntary and there's really no way to to oversee this.
- 7:38You know, the money isn't serious.
- 7:39It's 27,000,000 across 60 projects.
- 7:43That's about 450,000 lbs each. Compare that with a single
- 7:49frontier training run costing hundreds of millions of of
- 7:52dollars, and it's really hard to argue that this research
- 7:56portfolio meaningfully ship shapes how the next generation
- 8:00of models actually behaves. This voluntary participation as
- 8:05well. It's a fundamental structural
- 8:06weakness. It's great that so many of the
- 8:08labs are involved, but they can just walk away.
- 8:12And there's no real mechanism to compel access, compel disclosure
- 8:16of findings, or require companies to act on research
- 8:20that contradicts their deployment timelines.
- 8:23The whole thing depends on goodwill.
- 8:26And we've then got this, you know, as you mentioned, sort of
- 8:31the, the, the, the UK sits in that fun place between,
- 8:34somewhere between Brussels and the US and, and that's great.
- 8:38But actually our international leverage is structurally
- 8:41limited. The EU has regulatory muscle
- 8:44through its market size. the US has the labs, the compute, the
- 8:47capital. It, you know, as we've seen in
- 8:50the last week, it has government openly pressuring Anthropic to
- 8:54change some of its rules. And that means that something
- 8:59like this could, could be meaningless.
- 9:03You know, we don't have the, the, the, the market size or
- 9:06the, or any of the computer capital.
- 9:09So we don't really have any leverage other than goodwill and
- 9:11diplomacy. And I think, you know, while we,
- 9:15you know, we know that the UK has produced world class AI
- 9:18safety researchers, but a significant number, you know,
- 9:22have ended up at anthropic or open AI in San Francisco.
- 9:26Yes, there's still a great sensor in the, in London, you
- 9:29know, often because Google did manage to keep sort of much of
- 9:32the deep mind researchers there. But it, it does strike me that
- 9:38this could be a little bit of sort of window dressing and kind
- 9:43of, shall we say, safety washing.
- 9:45And I, and I really hope it isn't, but it kind of makes me
- 9:50feel a little bit like this is sort of just a bit of AI safety
- 9:54washing rather than being a serious effort to mitigate
- 9:59against the risks. Yes, there are many arguments we
- 10:04run into time and again when we're discussing AI issues
- 10:07related to the UK that the whole stuff about the budget, yes,
- 10:09these the amounts of money being talked about.
- 10:12And I think it said open AI threw in 6,000,000 and there
- 10:15would have been a time one Once Upon, you know, Once Upon a time
- 10:18or, or in certain sectors where that would have been a a
- 10:21meaningful amount of money for a company like that.
- 10:24It's to be cynical, it is, it has it has the feel of APR
- 10:29exercise. It's, it's a comparatively
- 10:31inexpensive contribution to a 'cause that I think most people
- 10:36would agree that is, is very important.
- 10:41Yes. Also, you know, you and I live
- 10:44in the UK, We've lived here through Brexit.
- 10:45The arguments about how we could position ourselves as what
- 10:50whatever analogy you wanted to use, you know, Singapore on
- 10:52Thames as like a, a midway point between Brussels and and, and
- 10:55the States. Some, some of these positions
- 10:59are, you know, they're not entirely without merit that
- 11:01there is something to be said for the UK being a kind of
- 11:06middleman between the more freewheeling approach the States
- 11:10and the more regulatory driven approach of Brussels that
- 11:13American companies might find difficult to deal with.
- 11:15But you'd have to say ultimately money talks.
- 11:18And the EU, you hear so much about the EU being stagnant and
- 11:24and lacking innovation, but the size of the market is it's still
- 11:28enormous. It's, it's something in the
- 11:31region of 16-17 trillion that the, the GDP for the EU, it
- 11:34cannot be ignored and that that kind of money talks.
- 11:37All that said, I I can only wish it all the best.
- 11:40I mean, I think there are some really serious individuals there
- 11:43that has some proper academic backing and they've clearly done
- 11:46some really good work already. The question will be whether
- 11:50they can persist with that and have the backing of it that can
- 11:53make it have long term influence and meaning.
- 11:57And that really given the structural issues is very much
- 12:01an open question. On that, I think I think we'd
- 12:04probably agree. We do.
- 12:07I think, I think one of the things that sort of segues
- 12:11nicely into the next story as well is that as part of its
- 12:15rebrand, the AI Security Institute actually shifted its
- 12:17emphasis. So it wasn't just on the
- 12:20security of the models being released, but it was also to do
- 12:24with supporting AI adoption, which seems a little bit about
- 12:28counterintuitive. Maybe it's trying to make sure
- 12:31that adoption is done in a Safeway that doesn't cause too
- 12:35many job losses, but that but recent research by the
- 12:42department of well DIDSIT and it's AI adoption research is
- 12:49that it's studied 3 1/2 thousand businesses, you know, using
- 12:57telephone interviews and sort of talking to the actual tech
- 13:01decision maker in each organisation.
- 13:04And only 16% of UK businesses are currently using AI, not
- 13:11piloting it, not exploring it, actually using it.
- 13:15And 80% say they have no plans to adopt at all.
- 13:22So we've got this money going in to sort of look at security and,
- 13:28you know, an adoption. And yet we've got companies in a
- 13:31country where, you know, in the country that that have no plans
- 13:34to adopt. And they might be, there's an
- 13:38interesting challenge here that actually people are using it
- 13:41when the companies don't have, you know, aren't adopting it.
- 13:43And that's something I know we're going to come to and talk
- 13:45about at length. One of my biggest bug bears, but
- 13:49this is the same government that launched the AI Opportunities
- 13:52Action Plan early in January 2025 that committed to embedding
- 13:57AI across the entire UK economy. There were 50 recommendations in
- 14:03the plan, including appointing AI sector champions, building AI
- 14:06growth zones, and positioning the UK as a global destination
- 14:10for frontier AI investment. Now we know some of that's
- 14:13working, we're seeing the results of that and I know that
- 14:17the our final story of the day is going to be talking about
- 14:19the, you know, success of some of that.
- 14:21But the same government that's hosting billions in, announced
- 14:24in with investment that has sovereign AI infrastructure
- 14:27agenda, that talks about AI as the central driver of the
- 14:30productivity growth this country needs is saying that four and
- 14:34five businesses have no plans to adopt AI.
- 14:39That's not a major discrepancy between ambition and reality.
- 14:42That's a structural question about who the AI economy is
- 14:46actually being built for. Is the UK building world class
- 14:50AI capability that serves as a relatively small, sophisticated
- 14:53slice of the economy? Or is there a genuine path to
- 14:58the kind of broad based productivity transformation that
- 15:01would actually show up in GDP? You know, the interesting
- 15:05wrinkle, and I think this is where it gets really, really
- 15:07complicated, is where the 16% is even the right number to be
- 15:11arguing at, because the Office of National Statistics, the
- 15:16Office of National Statistics data has self reported AI use at
- 15:2023%. ChatGPT is 16,000,000 monthly UK
- 15:25users. 61% of organisations now permit generative AI for work
- 15:31tasks. So either there's a massive
- 15:34measurement problem or there's a meaningful difference between
- 15:38using AI and allowing its use in a business and actually adopting
- 15:43it and building your business processes and structures around
- 15:48AI and and which of those you believe shapes everything about
- 15:52what the policy response should look like.
- 15:56You know, personally, I believe that the number flat flatters
- 15:59the failure. You know, 16% is already
- 16:01generous. And so I, I think that the gap
- 16:08between 16 and 23% of employees using ChatGPT on their own
- 16:13initiative or, or that gap is that, that the people are using
- 16:17that. And that's not adoption, that's
- 16:19shadow use. That's, that's secret cyborgs
- 16:22using AI with no strategy behind it and actually bringing
- 16:26security risks into those businesses already.
- 16:30What? What do you think, Neils?
- 16:33I was astonished by some of these numbers and like you, this
- 16:38is an area that's of great relevance to me given that the
- 16:40line of work we're in and that the, you know, we run workshops
- 16:43and we speak to a lot of business people and people in
- 16:46academia about these figures. I would.
- 16:50So I dug into a couple of them actually.
- 16:52So I think So what what you said at the beginning about 16% are
- 16:55using it. I mean this, I think this
- 16:58certainly merits a bit of further discussion about what is
- 17:02meant by using it because this this meant implemented a kind of
- 17:08enterprise and strategic level, correct?
- 17:11Yeah. So it's rather than so if your
- 17:14definition of using it is we have licences for Co pilot for
- 17:17our workforce and they could point to a certain level of use
- 17:22among the company, it would be a different figure.
- 17:26So it does. I'm not saying that just people
- 17:29randomly using Copilot constitutes the best use of AI.
- 17:33It doesn't. But I would say just just for
- 17:37some clarity on what using it actually means here.
- 17:40And the 80%, as far as I could tell, I dug into that a bit as
- 17:43well. 80%, they have no plans to adopt it at all.
- 17:47Apparently that brings together a couple of positions, one of
- 17:50which is so around 50%, we're saying they don't see it as
- 17:54necessary, which is still to me a shocking figure and more more
- 18:02indicative of the lack of understanding of everything it
- 18:06can do and how it stands to disrupt industries rather than a
- 18:10growing recognition by those industries that they've seen it,
- 18:13they understand it and it's not suitable for them.
- 18:15But the other, the other part of that 80% cohort was more along
- 18:21the lines of no active plans to implement it, which is, I know
- 18:26it might sound like split in hairs, but it's perhaps not as
- 18:28strident as we are not going to. We have, you know, we've decided
- 18:32this is not for us. It's a bit more ambiguous.
- 18:35They just haven't sorted out their position yet.
- 18:37Still not great. You know, you'd think by now 3
- 18:401/2 years in that really companies should have a more
- 18:46informed position about how they plan to adjust to this.
- 18:50Anyway, to get back to what you said, the number of flights to
- 18:54the field, 16% is already generous.
- 18:58I, I, I broadly agree. I, I think there are, there are
- 19:02points of scepticism that and I will try to raise something in
- 19:05opposition to what you're saying.
- 19:07I think to shadow use is, is seen as broadly unhealthy
- 19:13precisely for those reasons of security.
- 19:16And I think it's, it's difficult to disagree with that.
- 19:18I think if you, if you as a company are adopting a position
- 19:22whereby you're saying either do not use this at all or you
- 19:25haven't really figured out what your position is, you are
- 19:29absolutely inviting the use, you're inviting shadow use and
- 19:33that and that could have security implications.
- 19:36Shadow use in itself. I wouldn't want to say go ahead
- 19:41and do shadow use. What what we would encourage
- 19:43people in positions who are unable to, even if you're in a
- 19:48workplace that says you cannot use this legitimately on our
- 19:52systems, etcetera. We would always say to them, you
- 19:56have to learn it anyway. You learn it anyway.
- 19:58Learn. Learn its uses.
- 20:00Find out how it can help your personal work life of course.
- 20:05Don't, you know, be careful with your data, do not put anything
- 20:08proprietary or sensitive in and especially not into free models.
- 20:14It's, it kind of feels in our AI bubble.
- 20:17It can feel kind of astonishing that we have to say that now,
- 20:20but it's absolutely the case. So shadow use can have its uses
- 20:24in at least encouraging people to understand what's available
- 20:28and how it could work for them. Granted it's not ideal, but I
- 20:31think shadow use can potentially it's better than individuals in
- 20:35your company just having no understanding of this at all
- 20:39because they can't use what they want to in a work environment.
- 20:45Yeah, I mean, I'd, I'd, I'd argue that shadow uses, there's
- 20:49a big difference between shadow use, ungoverned use and
- 20:55unstructured used. You know, I think if you've got
- 20:58some governance and, you know, got some governance in place,
- 21:02you can still have unstructured use where you're able, you're
- 21:05allowed to experiment, but that you don't have any sort of
- 21:08formal programmes in place. I think that's important.
- 21:13But if you don't have any programmes in place, you don't
- 21:16accumulate into, you know, into productivity gains across
- 21:19organisations. What you actually do is you end
- 21:22up with individuals finding productivity gains and then you
- 21:28sort of have fragmentation across teams.
- 21:30And so that can cause some quite quite interesting organisational
- 21:35challenges where you get huge amounts of productivity gains
- 21:39from some people and then other people feel under pressure, but
- 21:42they don't know how to catch up. So I think there's some real
- 21:44fundamental challenges there. But yeah, you're right, if
- 21:48people are just using free tools or they're not really looking at
- 21:52the core workflows, you're not, you're not really making AI that
- 21:56valuable. You're just treating it like a
- 21:58glorified, you know, spell checker, even if some people are
- 22:02using them incredibly in incredibly sophisticated ways.
- 22:07I, I think that that creates sort of faster task isolation
- 22:11and it really, it encourages siloed thinking, I believe,
- 22:17rather than actually doing what AI can do, which is be a great
- 22:21enabler for organisations to, to re evaluate how they get work
- 22:26done and where you know, and how they add value.
- 22:32Yes, I think as well there's this ongoing belief perhaps
- 22:36dating back to previous digital transformations or software
- 22:39programmes that and you know, you can understand why people in
- 22:45charge of compliance and security would would very much
- 22:47wish this to be the case. That you can implement some of
- 22:52these AI transformation programmes in a very orderly and
- 22:55top down manner. That the the, the kind of
- 22:58strategic goal can be identified longer term at the high level
- 23:02and implemented as a top down project.
- 23:06Our experience is very much that that is not the case, that in
- 23:10order for this to work, the the workforce in general has to be
- 23:14using it frequently and leadership have to be using it
- 23:16frequently, both to see how it works and to recognise broader
- 23:21innovations that cannot just be spotted necessarily at the top
- 23:25down level. This I think is perhaps the one
- 23:29of the biggest losses in in shadow use is that there's all
- 23:32these great ideas bubbling under the surface.
- 23:35You've got your your your most valuable, your most valuable
- 23:39commodity is your people. The most valuable asset is your
- 23:43people, how they relate to work, all they understand, all their
- 23:46implicit knowledge and all of that is being lost through
- 23:49shadow use of AI. All these potentially brilliant
- 23:52ideas that could be having an effect to the organisational
- 23:56level are lost because the infrastructure isn't in place
- 24:01for what is quote unquote proper use of the, you know, the 16%
- 24:07self-proclaimed from these figures.
- 24:09So it is it, it's, it's a, it's a really jarring reminder to
- 24:16those of us who are who work in this field and then perhaps in a
- 24:19bit of a bubble that it, it is still very early days.
- 24:23It's very easy to believe that this is way more advanced than
- 24:26you imagine it is because you and I are constantly We've got a
- 24:30LinkedIn algorithm feeding us all this stuff about cutting
- 24:33edge AI developments and implementation and the and the
- 24:36reality is that just does not represent what is going on for
- 24:40the vast majority of business owners.
- 24:42No. And I mean, I've, I've been,
- 24:44I've been amazed. I've been been working with
- 24:47some, some amazing AI folks out in the US, part of a business
- 24:53called New York Collective that was built up on PhD research
- 24:57looking at, you know, AI adoption and what made
- 25:01successful AI adoption or not. And the what is possible by a
- 25:07small focus group of people using AI and pushing the
- 25:10boundaries as the, you know, as things come out is just
- 25:14astonishing. And then when you look at some
- 25:17of the big corporates, it's almost like they're they're
- 25:21running through tweak, you know, running through Treacle.
- 25:25And, you know, and these small start-ups that are really using
- 25:28AI to its, you know, maybe not even to its full capacity, but
- 25:31using it in earnest are kind of like something like Sonic the
- 25:34Hedgehog on, you know, after it's got all of the power ups.
- 25:37I mean, the difference is just staggering.
- 25:40And, you know, I think that we're, we're seeing this, we're
- 25:44going to see over the next year or two, a huge sort of
- 25:48bifurcation of, you know, the, those that are using it and
- 25:51adopting it at scale. And those that really aren't,
- 25:54even if it's just, oh, we've got copilot licence and we haven't,
- 25:57I know there's going to be some protective moats around people
- 26:02that produce things. But we've got to remember the UK
- 26:06is as, as a, as a country has, has a huge, you know, has a huge
- 26:11amount of the economy based on create providing services and
- 26:16service work is hugely in the target for AI.
- 26:21You know, we've seen that from, you know, across financial
- 26:23services, from marketing, from coding.
- 26:27Those sorts of industries are the first to be disrupted.
- 26:31And if we've got 80% of businesses not really worried
- 26:34about that, we've got a problem. Yes, or an opportunity, mate.
- 26:42Let's always say we've got a big problem or a big.
- 26:44Opportunity. As a country we've got a problem
- 26:46as as individuals, maybe there's a big opportunity so.
- 26:49I can live with that. OK, so let's move on to our
- 26:54final story of the day, Another big one about the UK based AI
- 26:59company. So it's Wave secures $1.2
- 27:04billion in Britain's largest AI funding round.
- 27:08So london-based autonomous vehicle startup Wave has raised
- 27:11$1.2 billion at an 8.6 billion valuation announced on the 25th
- 27:16of February, marking one of the largest funding rounds ever
- 27:20secured by UK artificial intelligence company.
- 27:23It seems like we're seeing new ones of these every week at the
- 27:26moment. New record-breaking funding
- 27:30rounds. The round could rise to 1.5
- 27:33billion through an additional 300 million from Uber,
- 27:35contingent on hitting deployment milestones.
- 27:39The investment will scale waves. Embodied AI approach to self
- 27:42driving which uses end to end learning rather than high
- 27:45definition mapping systems. This distinction is really
- 27:50important. Most autonomous vehicle systems
- 27:53work a bit like London's black cab drivers, doing the knowledge
- 27:57that is memorising roots and environments in meticulous
- 28:01detail before operating in them. Waves AI, by contrast, is
- 28:06trained to read and adapt to whatever it encounters, allowing
- 28:10it to drive in cities it has never been specifically trained
- 28:13on. The company claims to have
- 28:15already operated Zero Shot, that is without prior city specific
- 28:20preparation in over 500 cities across Europe, North America and
- 28:24Japan. Wave is deploying its technology
- 28:27through a partnership with Uber, which also participated in the
- 28:30funding round. With London robot taxi trials
- 28:33scheduled to begin this year ahead of a planned roll out to
- 28:36more than 10 markets globally, London is shaping up to be a
- 28:40pivotal test of the technology. Waymo, the Alphabet owned
- 28:44robotaxi giant has logged over 170 million fully autonomous
- 28:49miles in the US, is also targeting a London passenger
- 28:52service launch by the end of 2026, making the capital the
- 28:57latest arena in the global race for driverless mobility.
- 29:01I've been, you know, as a London resident, obviously I've been
- 29:04keeping an eye on this. We are told that there will be
- 29:07potentially 3 companies operating driverless taxis on
- 29:11the streets of London by the end of this year.
- 29:14So I mean, that really does get the mind spinning about how all
- 29:20that's going to pan out. They they will apparently start
- 29:22off being more expensive than regular taxis.
- 29:26But there are so many other issues around the effect on a
- 29:29taxi trade and road safety. And if they run into a blocked
- 29:34road 1 of London's windy streets, will they have the
- 29:36wherewithal to do a little illegal UE to get out and and
- 29:42free up the traffic, all that stuff.
- 29:44But so getting back to wave this, this clearly this is a
- 29:50very impressive company. They've been around about a
- 29:52little less than 10 years and they are a serious player in
- 29:55this area as evidenced by the partners they have and their
- 30:00investors include Mercedes-Benz, Nissan, Microsoft, NVIDIA, Uber.
- 30:04These are serious companies. So that they that this
- 30:08technology is clearly at the cutting edge of what's possible.
- 30:11Also Waves, this business model is quite interesting and sort of
- 30:17evokes. Do you remember that there was a
- 30:19meme going around or, or like a graphic going around a while ago
- 30:22comparing all the new wave of Internet companies who own
- 30:26nothing but right, So they'd be like Airbnb or it doesn't own
- 30:29any hotels, Uber doesn't own any cars, all that stuff.
- 30:33So Wave doesn't build vehicles or run fleets.
- 30:36It licences software to automakers and partners with
- 30:39operators like Uber. That asset light approach gives
- 30:42it a far larger addressable market than rivals who are
- 30:45locked into specific hardware or city by city deployments.
- 30:50This another point, another couple of points and I'll I'll,
- 30:54I'll, I'll, I'll pass over to you One is that this idea of
- 30:58generalising so so this is kind of one of the holy grails of
- 31:04artificial intelligence. Indeed, AGI, the G stands for
- 31:06general and this idea that it can not do just one thing
- 31:10exceptionally well based on training data, but based on
- 31:13training can generalise that and perform to an extremely high
- 31:17level in different contexts and areas and Waves technology
- 31:21aspires to do that. Is this operating a zero shot
- 31:24across 500 cities without prior fine tuning is a genuinely
- 31:30significant technical claim. And the the last point is London
- 31:36gives Wave a home advantage that Waymo does not have here.
- 31:41Waymo's coming as an outsider mapping unfamiliar streets with
- 31:44safety drivers, but Wave has been testing on British roads
- 31:48since 2018. Now, for anyone who's driven in
- 31:51London, as I'm sure you have and as I do quite regularly, that
- 31:57local experience has got to count for something.
- 32:00So what what? What did you take away from all
- 32:02this, Matt? Yeah, I mean, so now, you know,
- 32:07these days I live somewhat outside of, you know, London,
- 32:09but have lived on the outskirts for, you know, did live on the
- 32:12outskirts for many years. And there were certain, you
- 32:14know, there were certain sort of limits to how far I was really
- 32:17that happy driving in, partly because of, you know, congestion
- 32:21charging, but also just because, you know, driving around the
- 32:24centre of London in Russia, just chaos, mayhem, and not a
- 32:28particularly pleasant place to be to drive.
- 32:31So I can imagine that actually having a, you know, an
- 32:34autonomous vehicle would reduce that.
- 32:36But I do fear for the, you know, the cabbies, of course.
- 32:40I think I think it's telling who's in, you know, who's who is
- 32:44investing with the likes of, as you say, Nissan, Mercedes,
- 32:47Stalantis, Uber participating in this round.
- 32:50I mean, it is encouraging, but we know that they do, you know,
- 32:53that the automotive industry does just pull out of, you know,
- 32:57AV commitments. If if there's if something
- 33:00doesn't look like it's going wrong.
- 33:01So just because they're investing, it could be a case of
- 33:04hedging their bets rather than it really being a case of that
- 33:09they're fully bought in. And we've seen sort of auto, you
- 33:12know, the, the automotive industry, you know, not
- 33:15necessarily even taking up the likes of sort of Apple Car play
- 33:18and sort of the Google alternative as widely as maybe
- 33:23some people would have liked. So there's, there's still, you
- 33:25know, still got a bit of concern about that.
- 33:30You know, obviously Waymo is a formidable competitor.
- 33:33It's got 170 million fully autonomous miles run already,
- 33:38and it's got Commercial Services running across multiple U.S.
- 33:41cities. You know, Wave is yet to operate
- 33:44a single paying passenger. And I think in London, it will
- 33:47face a direct comparison from Day 1 and in probably one of the
- 33:52most difficult cities in the world to navigate.
- 33:55I mean, I'm sure that there are others out there, but, you know,
- 33:59London's streets are notoriously narrow.
- 34:02They're busy and you know, frankly, it's very, very easy to
- 34:07get lost in. I think it's also that trying to
- 34:13operate reliably and safely is going to be really difficult,
- 34:17especially if you could try to launch across 500 cities all at
- 34:21once or very close to after each other.
- 34:27But I think the biggest risk here is probably the regulatory
- 34:29environment, you know, I think of, you know, in the UK the
- 34:33autonomous vehicle framework is still being finalised.
- 34:35We don't really know, you know, what people are going to have
- 34:39to, to target and they're not going to find out what, you know
- 34:44what, what's going to happen until the second-half of this
- 34:46year. So any slippage there puts waves
- 34:49timeline and investor patient under pressure.
- 34:51So you know. Like any AI topic, you care to
- 34:57mention that they're saying not, not just you, everyone who
- 35:00thinks about this stuff. Yeah, me too.
- 35:04So many uncertainties. I would just, I would just say
- 35:08though it is encouraging to see Auk company genuinely at the
- 35:12forefront of this. If it's going to happen to have
- 35:14a company with a genuine stake in the success of its business
- 35:19here. I'm encouraged to see how well
- 35:21Wave is doing. And I certainly hope that they
- 35:24proved to be a very strong commercial competitor to Wave
- 35:29and that this market is isn't dominated by a single player
- 35:33very early on. I agree.
- 35:35I guess the challenge is, is how are they going to get
- 35:37implemented in those cars? Is it going to be an easy add
- 35:40on? Because otherwise it could be a
- 35:42case of, you know, sort of software lockout as we've seen,
- 35:46you know, across PCs and, and other sort of systems.
- 35:49Where do you all of a sudden have, well, we, we're only, you
- 35:52know, our, our systems, you know, whatever manufacturers
- 35:57only allow one of those pieces of software on there.
- 36:00So it's going to be a, it's going to be a really interesting
- 36:03time in that market. Absolutely.
- 36:05And with that, I think that's a good place to sign off for this
- 36:10week. Three really fascinating stories
- 36:14that encapsulate so many of the issues swirling around AI in its
- 36:18particular application. The UK pleasure as always
- 36:21chatting to you, Matt. And to you nails.
- 36:25And we will see you all next week.
- 36:38And that's it for this week's episode of AI in the UK.
- 36:43If you found it useful, you can subscribe wherever you get your
- 36:46podcasts and we'd really appreciate a great rating or
- 36:50review. It genuinely helps others find
- 36:52the show. You can also follow Neil's
- 36:55Footman and Matt Wilkinson on LinkedIn for links, notes and
- 36:59occasional extra content. Thanks for listening and we'll
- 37:04be back next week. Helping you make sense of AI as
- 37:07it lands here in the UK. AI in the UK.
- 37:14AI in the UK. AI in the UK.
- 37:19AI in the UK.