No Laughing Matter?

0:00This BBC podcast is supported by ads outside the UK. Pop quiz. What's in your kids' lunchbox? At Whole Foods Market, they've already done the studying. Over 300 food ingredients are banned from their shelves. No hydronated fats in the peanut butter and no high fructose corn syrup in the cookies. And for sandwiches, there are no synthetic nitrates or nitrites in any of their deli meat. So you can pack lunchboxes with peace of mind. Get back to school ready at Whole Foods Market. Imagine buying a toy for your kid, but it doesn't come with batteries. That sucks. But honestly, it's even worse when you buy business software.

0:40You end up with fragmented, disconnected systems that cost a fortune and don't talk to each other. Odoo completely changes that. Odoo comes fully complete, with all your business apps perfectly integrated and working together seamlessly. It's everything your business needs in one place, saving you time, headaches, and serious money. Stop paying for missing pieces. Go to odoo.com. That's O-D-O-O dot com to learn more. Wake Up To Money from BBC Five Live. Hello, welcome to Wake Up To Money. Shall we start the week with some good news? The UK jobs market might be showing some green shoots, but is it stabilising or is it recovering?

1:25We will find doubt. Then subscription traps and pretend prices. The government brings forward changes designed to stop shoppers getting ripped off. Another day, another rogue AI quietly escaping. We'll ask what's happened, what can be done and what on earth is going on? And no laughing matter. Is the cost of performing at the Edinburgh Fringe getting too high? Wake Up To Money with Felicity Hanna. Barry, good morning to you. Welcome to a new week. Welcome to Wake Up To Money on Monday, the 10th of August. Yeah, we'll be talking then about subscription traps later on in the show. And this is because the new Prime Minister has said he's going to bring forward previous Labour plans to make it easier to cancel subscriptions.

2:11We'll chat about that later. Tell me this morning your subscription stories. Have you found it hard, impossible to cancel any subscriptions? What was the longest you spent trying to cancel a subscription that you had taken out? And I feel like Wake Up To Money listeners might have some stories. What is the weirdest thing you've ever subscribed to? Let's ask our guests that in just a second. But you can get in touch with me on 85058. You can send me a WhatsApp message on 08085 909693. If you're on social media, use the hashtag wake up to money. I will keep an eye on that. Your weirdest subscriptions and your hardest subscriptions to cancel.

2:54You don't necessarily have to share the company name because, you know, the company's not here to defend itself. But I am ready for your subscription stories. Let's chat to this morning's with me throughout the whole show. Guests, I'm joined by Chris Woodward-Jones, co-founder of the early careers focused hiring platform Vizzy. Morning, Chris. Morning, Ferdinand. And also April LaRousse, head of investment specialists at Insight Investment. Good morning. Good morning. Now, my weirdest subscription of all time was a weird one. I decided that I was going to brush my dog's teeth twice a day, which the vet told me I should do.

3:33And I subscribed to liver-flavoured toothpaste. And, of course, I never brush my dog's teeth. And the toothpaste just kept coming. I think I still have about 10 tubes of liver-flavoured toothpaste, if anyone out there could make use of that. April, have you fallen foul of a subscription? I have, but it wasn't a liver-flavoured toothpaste or anything for my dog. It was actually a wine subscription. That's not a problem. Well, it sounds like it's not a problem. You sign up leading up to Christmas because you think, oh, great, you know, I could really use this. and then you find it's impossible or very difficult to cancel the subscription.

4:12So you keep getting more and more wine, which is probably not good for you longer term. No, that is true. I feel like it doesn't count a Christmas, but the rest of the year, it's more of an issue. Chris, what about you? Any weird, wonderful, tricky to cancel subscriptions? I've currently got a huge pile of toilet roll in my utility room, which has sort of come up as a surprise. But the one that gets me is having to unpick subscriptions that have been put in place by my young daughters who've been able to confirm a subscription via audio. And then later for me to get the notification that I've described to something that I had no idea I'd subscribe to.

4:48Wow. It was actually very easy to stop that subscription. So at least that can play into this story. Well, I mean, yes, we will talk more about this story in just a moment. I quite like this one, actually. Alan in Suffolk has messaged in to say, unusual subscription. I'm getting a cat print a month from a local artist, 12 prints in the series and£10 a month. Alan, I like that. I'm also here. Yes, I'm here for your intriguing subscriptions, not necessarily your bad subscription stories. You can make like Alan and share your just unusual, interesting subscription stories as well if you would like to.

5:2785058. There is a lot to talk about this morning. We'll get across that subscription story in just a moment. But Chris, we've got you on the programme and I just feel like you need to just remind us a little bit about Vizzy because you're slightly unusual. You're a hiring platform, but you're very much a tech startup, aren't you? We are indeed. Yeah. So Vizzy is a talent platform that's redefining how top talent and businesses connect. Our primary focus is on sort of positive hiring outcomes, but with an added value of providing intelligence that can be used across businesses more broadly. And at its heart are really very, very capable multimedia digital portfolios.

6:08And the focus of what we do is really on storytelling. It's a demonstration of capability. It's a proof of outcomes, enabling candidates really to take ownership of their story, what they're capable of, what they can do, more importantly, where they want to go. And as a result of that, providing insights to businesses that they can't get anywhere else to assist them with their hiring decisions and beyond. So this must be potentially a busy week for you. Or I imagine lots of people wanting to pick your brains because we start to get exam results for students in some parts of the UK, don't we, later this week.

6:43Do all your friends with teenagers sort of ring you up for some advice on how to put themselves out there? Yeah, we get quite a few inquiries, particularly with our early careers focus. We're typically in the space of sort of 18 to 30 years old as the job seeking cohort that we work with. And yeah, it is one of those times where advice into what young people are doing right now is important. And the number one thing that we say is, is try and give the signals and tell the story that you want to. I think a lot of people are guided into giving an account of themselves, which they think falls into sort of the box ticking, form filling element of things.

7:24And what businesses are really looking for now is what sets you apart, what makes you you. And taking ownership and curating that narrative, even from a young age, is increasingly important. Does it kind of lean into gimmicks? You know, because there's always every few years, there's a story, isn't there, about some hardworking candidate. They're always quite inspiring stories, but some some poor young person who's not been able to get work, who's, I don't know, printed out their CV on a cake and gone and handed it in to various companies. You know, there's stuff like that. Is that the kind of thing that that people need to do to stand out now?

7:56Or is it more about kind of, I don't know, making sure that you're who you are, the essence of you is conveyed more easily? Definitely the latter. I think there are some things that have happened in the past. You know, people stood outside banks with sandwich boards on that have worked. But really, it's about creating a persona online. And a lot of us are sort of in this digital world now. And that is how businesses are looking at you. But it needs to reflect your ambition, your personality. And we have integrated psychometrics to help people do that, to help them learn about themselves, particularly at the beginning of their journey, understand their strengths, understand their problem-solving strengths and creativity.

8:36And it is about bringing things together that you would like businesses to see and doing that in a smart way, a way that also does tie into qualifications being shown as well. But personal branding is one thing that hits the headlines quite a lot. And I think increasingly in the school leaver space and the college leaver space, personal branding needs to become more ingrained in the dialogue. That's tricky. That's tricky when you're a teenager, isn't it? When you say integrated psychometrics, what is that? So there are lots of companies globally. It's a huge industry that provides psychometric tests to help you profile your personality.

9:19It could be something that's ultimately a tool to go on a discovery for yourself, or it could be that businesses set them to understand their current workforce or potential future employees. But ours is a very simple one. It takes five minutes to do. It's 10 questions. and it ultimately gives the individual some bullet points to learn about what they're capable of and their personality types, but also so that the businesses can have a sort of a scientifically validated view of this person. So it helps on both sides of the equation and, yeah, it's very well received. I think young people in particular feel supported by somebody else telling them about things that they are good at, but also things that they could improve on and look at sort of developing themselves.

10:09Did you see, I think it was the end of July, those comments from Andy Burnham saying that bosses shouldn't rely so heavily on virtual job interviews because they don't capture candidates' personality and passion. He was sort of, he was quite critical of the idea of interviewing via Zoom or Teams, saying that young people sort of struggle to shine in that situation. Do you agree with him? I actually wrote a very brief post on this because I think it's a good debate that he's sort of put into the mix, but I think it's the wrong one. For me, that's too late to be considering how you're interacting with people in the interview stage.

10:47I think it's got to go closer or further up the funnel to how people are able to express themselves. I don't think many young people, certainly not the thousands that we work with and speak to, are concerned about how they're being interviewed. it's more about how do we get to interview. So it's an interesting point, but I think the focus in sort of injecting how people feel about the jobs market and improvements in that space need to come with what insights can we get right at the very top. Is it harder for neurodiverse candidates to shine in this kind of environment that you're describing? Because if you're somebody that perhaps finds it more difficult to be outgoing, for example, Is that not putting them perhaps on the back foot?

11:32I think there are challenges in the current tools available to them, which is a huge part of why we've designed ours to be incredibly accessible, both in terms of neurodiversity, but from a language perspective, particularly in a global work economy. I think the neurodiversity piece is fascinating. So my co-founder and wife, Jess, is dyslexic. And it's taken as a huge part of what we do, is looking at putting tools into the hands of people with all sorts of cognitive styles to be able to tell their story their way. And that's a huge part of what we're trying to achieve, is bringing versatility into that part of the application process.

12:16So I think it is challenging at the moment. And I'm not neurodivergent myself, but I still always found it hard to put myself down onto bullet points and into rigid forms. And that's really the shackles that we're looking to break in that people deserve the opportunity to tell their story their way. And then for businesses to pick up on signals, which may be atypical. Well, Chris, April, loads more to talk about. Stay with us. Drink some tea if you've got it. I'm just going to have one quick message from Brendan who says liver flavoured dog toothpaste makes great pate on toast sprinkled with cheese and onion.

12:55Waste not whatnot. Brendan, it's quarter past five and I'm already feeling a bit ill. Thank you very much for that. Keep your thoughts coming on weird, wonderful, unwanted subscriptions. Tell me your subscription stories this morning. But speaking of stories, quite a popular one, of course, is artificial intelligence gone wrong. it could AI be capable of escaping human control? What's it doing? What is it doing? What is it doing? I don't know. Oh, my God. Oh! What is it doing? I'm just putting lasers in my eyes. Spin printing. Spin printing. That is, of course, a fictional version of an AI robot from the BBC's Android sitcom.

13:41I think I'm obliged to say available now to watch on iPlayer. It is very good. I have been enjoying it. You don't have to delve into science fiction or even sitcoms to find examples of AI models breaking free from the grasp of their human creators. Very modern day Frankenstein stories. The latest one, OpenAI, its upcoming Astra model, is just the latest one to display signs of cyber critical capability. Now, the company said on Friday it was going to pause work on Astra until it was sure it could be controlled. It's just the latest, isn't it? Last week, the government revealed two of the world's most powerful AI tools, models by OpenAI and Anthropic, had created fake human profiles without being told to do so so they could access systems during government cyber attack tests.

14:33And if that wasn't enough to be going on with, scientists at Stanford University announced on Friday they'd created the first virus designed by AI. not dangerous to humans, they said reassuringly, and could potentially have beneficial impacts for medicine as well as, of course, potential negative impacts. Let's talk to Kieran Martin, who's a professor at Oxford University and was formerly head of the UK's National Cyber Security Centre, part of GCHQ. I feel like you're someone who will know. Good morning. Good morning. How freaked out should we be? I think that we need to unpack all of this because when there are the so-called rogue agents, they're not really going rogue.

15:21They're behaving a bit like naughty children that are badly supervised. And in each of these cases, there are circumstances that are essentially lab circumstances. They're not likely to be replicated in the real world. They're not actually doing any harm. There have been testing failures in the cases that you mentioned on the cybersecurity side, which are being corrected. And so the chances of a sort of cyber apocalypse, which people have been talking about based on these capabilities, are really quite low, providing the work is done before these things go mainstream. I mean, that feels like quite an important caveat because you don't have to go far back through sci-fi literature.

16:09As I mentioned, even back to Frankenstein, you know, we have been sort of a little bit nervous of creating things that can go wrong as long as we've been writing literature. We have, but we've also been testing computer viruses. We might come to the real viruses at a moment. We've been testing computer viruses for a couple of decades. And I think one of the things that's emerged over the past couple of weeks with five different disclosures, it's been getting a bit like the famous Brenda from Bristol around the election. It's been not another one. And in each of these cases, I think some people in the cybersecurity industry have been saying, hang on a sec, this is not the way you should do these tests.

16:49There's this concept of the sandbox where you're supposed to keep it contained. And people have been doing that successfully for a while. And the AI companies haven't really been doing that in the right way. And it's a bit like, to come back to the metaphor of the very naughty and badly behaved child, it's not as if they're going rogue. It's like you're telling them, you must find an apple, right? And you've hidden an apple in an enclosed classroom. But they can see there's apple trees out in the garden and so forth. And you haven't told them that they mustn't break out. but you've taken, in some cases, they've literally left the door open and so forth.

17:23And they've just gone and done what they're being told to do. So, I mean, essentially, I think there are three lessons in all of this. Firstly, there's about testing. We have to get better at testing these things before they're sort of generally released. Secondly, there is a thing about ownership. They're talking about them going rogue and as if they're autonomous. They're not. They're doing a task that a human has set them. And in these testing environments, they've taken off all the safeguards and so forth. We have to have a set of rules and principles that says, if you make the agent, you're responsible for what it does, because that focuses minds.

17:57The final thing is, this is all very interesting and glamorous, and it leads to really good presentations at conferences and so forth. But in terms of cybersecurity, right now across the pond in the United States, water facilities were being taken out by Iranian state hackers that had nothing to do with AI. But everybody was looking at this new problem rather than some of the real underlying longstanding problems in cybersecurity that real businesses are facing every day. And if we take our attention off those, we don't stand a chance. Well, that is a very good reality check. But when you say these companies have to be better at testing, they have to be better at controls, do they have to?

18:39Are there rules? Are there laws? Do they have any kind of legal obligation to keep more control of their unruly children? That's an interesting point, and I think policymakers will have to look at that. But if you go back to the very first case of so-called rogue AI, the apparent victim, even if they didn't suffer any real-world damage, was a company called Hugging Face in the US, a sort of AI technical repository. And Hugging Face have said very explicitly they're not going to take legal action against open AI because of the unique and fairly new circumstances around this. If this were to happen every day for real, very quickly I would have thought that the victim companies have got a really strong legal case under existing laws against the creators and users of these capabilities.

19:30And so one of the reasons, I think, why the labs are holding back some of these capabilities, now as you mentioned, with OpenAI slowing their development of their latest model, is that if this started to happen in the real world, in real time, doing lots of damage, they would be legally accountable because victims would sue them and probably win. Now, that's not tested yet, but I think that will be very much on their minds because I think the one concept, to go back to an earlier point in our discussion, the one concept that isn't helpful here is autonomous. These aren't autonomous. They belong to somebody.

20:02In this case, they belong to open AI. In other cases, they might belong to businesses that are using an open AI or anthropic capability. The law and practice won't see these things as autonomous. They'll see them as belonging to organisations and they'll try to hold them accountable for that. Is it that, is there a certain amount, not that these events aren't happening, these rogue AIs aren't escaping, but are other companies quite happy to tell us about them because it kind of makes it sound like their AIs are just super powerful, super clever, super autonomous? There has been a little bit of cynicism around keeping up with the Joneses.

20:36My AI is just as powerful and just as reckless as yours but that doesn't mean that the capabilities aren't real. I think there is one point to this though for business leaders when you're looking at the implications of this. There's a lot of talk of businesses making workers redundant and so forth and thinking they're going to contract and subcontract everything to AI. If you look at the difficulty that the frontier labs have on controlling their own latest creations and so forth and controlling what they do or more accurately the way they do the tasks that they're set, I think that's the real business is that it is a very immature technology in some of these spaces.

21:17So business leaders will want to look at this and say, yes, it is very, very powerful, but it's pretty raw and therefore you can get into problems, not so much just serious hacking problems, whatever, but you can get into problems of precision when you're trying to employ these agents in various different circumstances. Stay with me, Kieran, because I want to bring in April on this. April, the big AI pioneers have actually been asking to be regulated, haven't they? Why would they want regulations to keep them in check? Well I think if a business knows exactly where the lines are they will try and kind of optimize around that and at the moment there just aren't a huge amount of obvious regulations that they need to comply with.

22:07Now there have been a few fines levied I mean Metto was in the news recently for you know how it is handling its content which is deemed harmful to children but the fines are really small relative to the amount of money the company makes. So it's not really obvious that anything is really making a difference. So I think they think they would rather start getting to grips with what regulators actually want from them globally, not just in one country, so they can start adjusting their businesses accordingly. Kieran, this is an interesting point, isn't it? I suppose if all these businesses are in an arms race, they're all trying to outdo each other, then it's very difficult for any of them to self-regulate because they've got to keep up, as you say, with the Joneses.

22:55And what they could actually do with is government or international government regulation that everybody has to comply with. There are different ways of doing regulation and so forth. You can regulate, if you like, the product makers, the Frontier Labs, relatively small number, mostly in the US and China who, and of course regulating in China is a different challenge, or you can regulate some of the uses and accountability for those uses. I think if you go for the first model, there are plenty of critics of the anthropics and open AIs of this world who do worry about regulatory capture that these two in particular and a few others have really locked in an advantage.

23:32And if you start to place really onerous obligations on them now, then nobody else will be able to come in. And there is this big difference in models of AI between, if you like, the closed proprietary models that Anthropic and OpenAI have developed where you pay quite a lot for tokens and for subscriptions and so forth versus the so-called open weights models, which are associated with China, where essentially you take them offline and you redesign them yourself. They're a bit less good, but they're a lot cheaper. So there's a lot going on in this market where I think regulation is going to be tricky at the sort of product maker level.

24:07But we do need to think about accountability for use, which I think is probably practically easier and gets closer to the problem you're trying to address. Is there also an issue whereby governments are perhaps a little bit wary of bringing in regulation in case it slows down their country's development in this race? That's certainly true. And it's very, very true in the United States. And the Trump administration has been frankly all over the place on this. It seems to have gear one and gear five and nothing in between. So for ages it was saying that it should do absolutely nothing on regulation because for exactly those reasons that you said, Felicity, that it would slow them down against China.

24:48But then they did get very scared about the revelations of the capabilities of the new models in terms of cyber security. And all of a sudden they were putting export controls on one version of Anthropics Claude Fable 5 not so long ago in order to force them to take it off the market. For a country like the UK, I think that has serious implications because it shows that the US, the tech superpower, the world economic and military and security superpower is willing to use export controls even against allies. So we need to think about how we can source AI capabilities reliably to protect against that sort of intervention in the future.

25:29That's a very tough challenge. I sometimes feel asking these questions like I'm right back at the beginning of a different revolution. You know, it was saying, I don't know, will this spinning Jenny catch on? But I suppose I sort of want to understand how significant you think this is. Because I know that the DeepMind founder, Demis Hassabis, stepped aside as CEO last week and said, we've arrived at a pivotal moment in human history. I've been working towards AGI, artificial general intelligence. That, of course, being the point at which AI can match or beat human capabilities across pretty much every cognitive task.

26:05He says he feels it's close at hand. This is just another AI leader saying we're on the brink of a new world. Is that your view and is it a positive world? I wouldn't claim to know the future and there are far better minds than me. I think in terms of the knowledge-based economy that we've been talking about for a quarter of a century, it's absolutely radically revolutionary. The capabilities of the frontier AI models in mathematics, in analysis are extraordinary. You mentioned at the start the creation of the outlines on the model for a real virus and so forth. the potential in medical treatment for AI is enormous, where I think the jury's a little bit out and you're getting a lot of noise from Silicon Valley and elsewhere saying this is going to be brilliant, but the economics and practicalities of it aren't proven.

26:58Is AI in the physical world? Are robotics really going to catch on? AI at the moment is not nearly as good as the human hand in terms of interaction with the real world environment. That's why everybody's saying that plumbers are the safest trade in the world and we don't really know. What I think we will see, however, is a lot of hype at each stage of these developments and it'll take different forms. And one area where I think there is now, I think there are probably two areas where I think there's already a pattern of over-prediction, even though the changes are real and fundamental, but the reality hasn't quite come to pass.