.png)
This Week in Leading AI
Imagine two mates at the bar. Thirty years of business between them. And all they want to talk about is AI.
That's "This Week in Leading AI". The podcast where Kieron and Neil cut through the hype, share what's really working in the world of Generative AI, and helping people figure out this AI thing without the techno-babble.
Just honest conversation, real stories from the AI coalface, and the kind of straight-talking advice you'd only get from people who've worked together for 30+ years, been there, done that, broken things, gone "Oh S***!, fixed it, and lived to tell the tale. They claim Leading AI is the best job they've ever had and are having a blast doing it. It shows.
Warning: may cause you to actually enjoy learning about AI
Pull up a stool. We'll get the beers in.
This Week in Leading AI
Will AI kill us all?
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Both Kieron and Neil have been gallivanting round the country all week, both are recording on a Friday afternoon with the pub in sight, and Neil has already had a text from a mate asking where he is.
Housing Community Summit — a full room, and one uncomfortable answer
Kieron was back in Liverpool, this time speaking solo to a breakout session of CEOs and chairs. The stats he opened with tell the whole story of AI in housing right now: 76% of housing staff say they use AI. 69% of the public don't know AI is being used in public services at all, and don't trust it. 87% rate their own AI skills as low. And 44% of organisations still have no AI policy. Enormous usage, by people who'll openly tell you they don't know much about it, with nothing written down about what they should or shouldn't be doing.
Then he asked the room how they know they can trust their AI's responses. The headline answer: we don't know.
So he sent them five questions to take back to their teams:
- Which documents does it read, and who keeps those current?
- Can it show its sources, so an answer can be checked rather than just trusted?
- Who built it, who signed it off, and who reviews it?
- What happens when a policy changes — does the agent change with it?
- Would you be comfortable if this answer reached a resident, the regulator, or the ombudsman?
If you can't answer those, you've got work to do.
The complaint that came back from its own reply
Neil was chatting to a friend (in a bar, obvs) who lives in a housing association property. He'd put in a complaint last week, ten pages, written with the help of AI. He got a response but couldn't tell whether it was AI or not. That's good because the housing association is a Leading AI customer so there's a reasonable chance KnowledgeFlow wrote the reply. Complaints in housing are up 90%, and one report this week says they've doubled since ChatGPT. The genuinely good side: people who'd struggle to write a compelling complaint can now represent themselves properly. The difficult side for housing thought is there's no change to statutory response times so they're having to deal with twice the volume and ten times the size.
"Which bit of stupid is that?"
A customer emailed Neil to explain they'd been busy so hadn't moved their AI adoption on, but mentioned in passing that one of the team had put a load of confidential information into a public AI tool to produce a three-page report which they thought was "great". This is an organisation that has a private, secure AI assistant built specifically so that wouldn't happen. Dumb. The formatting was lovely, apparently.
Are we all going to die? A 10% chance, apparently
Kieron got asked about the Anthropic researcher's resignation on a webinar with the Moving On Up team and gave the honest answer "You're about as qualified to assess that number as I am." The more useful point came from a Radio 4 interview which said you can't buy a sandwich, board a plane, or ride in a lift without a regulatory framework, and yet you can build superintelligence, and the only real constraint is investment. The companies themselves are asking to be regulated. Nobody's going to stop first in a trillion-dollar race.
A Singaporean government minister running his life on a $15 Raspberry Pi
He uses a Raspberry PI with 8GB of RAM, a small local model, and it manages his schedule, drafts his speeches, and briefs him before visits. Which raises the trillion-dollar question: if that's all most people actually need, has everybody placed their bets in the wrong place? Meanwhile Nvidia posted $96.2 billion of revenue in a single quarter, up 106% year on year. And Kieron's mum is currently a data centre protester in Devon, banners up outside the house and everything.
Watermarking, properly explained
Donald got under the skin of how Claude's watermarking actually works. It isn't a label, it's a pattern woven through the way tokens are assembled, detectable if you hold the right key. Very clever, but easy to bypass by pasting the text into a different model and asking it to rewrite while retaining everything. So much for the end of the AI detector industry.
Also: Daniel Susskind's book on teaching in an era of AI. PISA scores have fallen and haven't recovered post-pandemic, and his answer is to train children both with AI and without it so they understand the difference. Kieron tells the story of the 1890s New York horse manure crisis (he so full of **** 🤣). And the discovery that China's clampdown on AI companions applies only to Chinese citizens. Their companies are free to sell AI boyfriends and girlfriends abroad, though Neil doesn't think Mrs Watkins will allow him to have one and Kieron's worried about spending too much time on a screen without that kind of distraction.
It's been one of those weeks...
Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI.
Pull up a stool — we'll get the beers in. 🍺
All right, I'll take your advice. And we'll get the pantomime horse of a podcast underway for this week, Kieran. You're looking very cheerful for a man who's been running around the country all week.
SPEAKER_00I am cheerful because it's almost uh we're recording this on Friday, and uh it's the weekend coming up, and that is a very good thing for me. I've had a very busy week. Yeah, it's been great. Really enjoyed myself this week, been out and about talking to people, been on a webinar, um, and we'll share with we'll talk about it, I'm sure. But the Housing Community Summit for 2026, I was uh invited to speak there again, um, and I'll share with you what happened in that.
SPEAKER_01But yeah, really good. Cool. Well, I saw you on their um uh it's a wrap for the uh conference and you were in their little video at the end. That was nice, that was nice.
SPEAKER_00I saw the photographer coming in the room, but I didn't know he was uh filming as well. But I could kind of see him coming in over the back of uh the audience. So yeah, that was good to make it.
SPEAKER_01Well, I'm over here.
SPEAKER_00If I'd have known he was filming, then I could have made it even more I could have walked through. You could be more dramatic. That was really good. How about you? You've been uh gallivanting around the country.
SPEAKER_01I have been gallivanting, yes. I've been around the country this week, and uh I too am glad that it's Friday that we're recording this. And uh I've already had a text from a met saying you come into the pub, and I was like, I've got to do something, I've got to do this thing first. I can't serve it. So not yet, not yet. So soon enough, soon enough, we'll be in our respective locals at uh four o'clock, I would imagine.
SPEAKER_00Shortly after. Yeah, I hope so. That'll be very good. If if we survive the AI onslaught that's coming our way, apparently, according to according to a junior researcher, it appears to me, from Anthropic.
SPEAKER_01A guy who was really fed up with his uh bonus and uh decided to quit in the most dramatic style. I'm not sure that's the story.
SPEAKER_00No, he's certainly done himself uh a lot of uh uh I guess a lot of good from getting uh uh influencer profile.
SPEAKER_01He has, but he'll never get another job, he'll never get another job in AI though, will he?
SPEAKER_00I um I was asked about it on the webinar, and I my my first answer was well, you're about as qualified to answer it as I am, frankly. It's are we all going to be uh killed by AI systems? Um, and how you work out that it's about a 10% chance. I mean, I'd like to see the maths behind that equation. It's kind of finger in the air at best. Yeah. So um, yeah, but I interestingly, my reflection on it on the webinar was um, I should call out the webinar moving on up, uh, which is uh a network, a group of people who are passionate about getting care experienced people into apprenticeships, yeah, which is tremendous because they really struggle. Um, if you are a care lever at 18, which as we know is a pretty challenging time because I mean being in care is obviously not the greatest experience, and that's very underselling it. Um, but then you are suddenly left to your own devices and having talked to my son about getting an apprenticeship. I mean, it takes a it's a lot of admin, it's a lot of work to work out what's going on, and without a parent or a carer to kind of guide you through that, you're left on your own to do it. So massive challenge, and I'm delighted that they uh moving on up network are uh they work under the Amazing Apprenticeships team and with with them and their job is to try and focus on getting more apprentices into sorry, more care experienced people into apprenticeships. So good on them. I was delighted to join their webinar. The point I was making was about uh whether we're all going to be killed, and here's the challenge is it is an unregulated sector. Really. AI. I mean they've got the EU Act and a few other things coming in. But I saw I heard a guy on Radio 4 talking and he was he was saying you can't buy a sandwich without being subject to a regulatory framework, and yet and you can't go on a plane or go on a lift or but yet you can build super intelligence and the only thing stopping you is really investment, and they uh they're not short of that at the moment, so but it is potentially a challenge and probably needs a bit more regulation.
SPEAKER_01Well, we talked a little we talked a little bit about regulation before, especially when it comes to things like safety and and people and all of that good stuff, and I've got a couple of things I wanted to talk about on that, but just come back to your men's point about you know the 10% chance of AI killing the entirety of humanity in 10 years' time. So I kind of am a bit near about it because um uh actually my daughter worries about it, and she said to me, Don't why why aren't you worried? I was like, Well, there's absolutely nothing I can do about it. So it's one of those things which I just cannot control. Exactly. All I can do is try and use it for um things that I think are important and things are good, like the moving on up stuff, like our good friend Johnny Hoyle and and and all the good stuff that he and and the team are doing um uh for care leavers. I think that is important and and and but actually the the stuff I mean I I listened to a bunch of things on the uh FT and The Economist on the way up um from London the this morning and um and it was talking about um terrorists trying to um uh oh well depends on who you one man's terrorist, another man's freedom fighter is the old expression, isn't it? Yeah, yeah, trying to trying to to to to work out how to make ballistic missiles, which uh I mean if if they're anything like me, then that's gonna be you know wood and and sticky back plastic and gaffe tape channel, isn't it? That's that's gonna be rubbish. Um but the but there was a uh was it Claude um Anthropic who let out their report today on on state actors uh trying to use it for things like bioweapons, and that is genuinely, you know, yeah uh generally worrying. Anybody who's seen the James Bond movie, uh or the one where James Bond dies, which is the wrong thing. I mean, how can James Bond ever die? That's just the wrong thing, isn't it? Um but the old bioweapons is is is genuinely uh uh a concern. But um uh but they uh as it pointed out on the um uh on uh the article I was listening to, you know, they're just not going to stop. And as much as you know, Musk and Altman and others have all said, yeah, we need a pause, what they mean is I want a pause for other people so that I can crack on and and get get to the get ahead of everybody else and and be the best and be the first to super intelligent.
SPEAKER_00So the radio the Radio Ford dude said a thing which I thought was quite interesting. He said these companies are actually asking to be regulated. They because they kind of realize there is a challenge, and but it's a trillion dollar market that they're going for, and until someone tells them to stop, they're not going to. So there is a kind of angle. The challenge is global regulation here. There's no point in regulating just the US models or the UK models or any other, is there? Because you're gonna have other countries building them. Absolutely. But it is a concern about the the you know, designing a virus, design missiles, whatever, you know, just sort of uh in theory, people with without the resources to do that being able to. That is a concern. Uh that does worry me. That I in my mind, that is going to happen. Um you just hope that it is m minor enough that it's it's not a significant problem. But um, yeah, challenging.
SPEAKER_01It is really challenging, and and of course, um uh OpenAI re released uh Astra uh their new supermodel uh this week. And um, I know that Mark at our end has been testing that and thinks it's very good. Um he'd been pretty uh complimentary about it so far, and especially from an agentic perspective, it can run big old um uh workloads apparently. And and um uh I've yet to yet to put it, I've no chance this week. But yeah, that's but again, that's just another example of that that kind of escalating race. You know, one brings one out, the next one, the next one, the next one. And uh part of it is all that kind of uh um I think I mentioned this uh was it last time uh we talked, but uh in uh the US, the backlash against AI and data centers using uh power and water and the environment messing up local environments, etc. And uh there was a really fascinating thing I listened to this morning about um uh one of the ministers in the Singaporean government who runs his entire schedule on a Raspberry Pi. And uh he's gone. So he's downloaded a small language model, and um apparently this thing that costs like 15 bucks um uh has got eight gig of RAM and it just churns away, manages his schedule, produces his um speeches or briefs him on whenever he's doing visits, etc. And the and the question that was really in in the uh what the article was about was you know, have have all these people place their bets in the wrong place? You don't actually need all these huge data centres with all these huge um uh uh expensive chips because actually if you're running something uh small and what you're really interested in is um scheduling and and uh meetings and things, you don't need to know what hairy bikers in um California have for breakfast on a Thursday in September because it's not really relevant. So you could run around something really small and sensible on a on a tiny device. So uh I thought it was really, really interesting from a uh perspective of if all of that investment is wasted, what's gonna happen to the whole AI boom? What's gonna happen to the stock market? What's gonna happen to the loads of things? So yeah, yeah, uh, and and stuff we again, stuff we just can't control. All we can do is try and use it. But I I did I did think um it was a really interesting example, and maybe we should get uh our techie geeks to uh have a little play, get a little Raspberry Pi and have a little play and see what's possible.
SPEAKER_00Yeah, see if you can build a tiny model. Well, my mum would be very pleased with that because she is she lives in Devon. I know, I know how important her uh well-being and happiness is to you. But uh she lives in Devon, and there is a proposal to build a pretty large data center down there. They've identified it, they were going for planning, and uh the her and the local residents are furious about it for lots of reasons. And um yeah, so uh she'd be very pleased. Maybe she could put a proposal in to say, well, I've got a Raspberry Pi you can borrow instead. I can run it from the from the living room.
SPEAKER_01But you've talked before about your mum on here, I think, where you know you she was talking to uh uh being Pi, yeah, Pi the uh AI, uh about uh about art and things, and and it was engaging conversation about painting on plan air. So uh yeah, I thought that's where you were going. I thought that's where you were going with the story. But no, if she's if now she's a data center activist in Devon, then uh she definitely is.
SPEAKER_00I um she's got outside her house, she's got the big banners up. So any I know I was, but then there were quite a few of them in their town, so I I didn't feel quite so bad. But I did I was tempted and I said to her, but imagine the internet speed you're getting. Living next door to a data center.
SPEAKER_01You could run in a you could run an extension cable and get all your electricity for free. Get through the wire and just be.
SPEAKER_00But no, imagine it she could she could start herself a little uh AI business running with zero latency if you're right next door to the fastest AI in the country, brilliant. Anyway, on on on investment and costs in here, I've got a little guessing game for you. How much revenue do you think NVIDIA, their last quarterly reported data? How much how many dollars do you think they did in revenue?
SPEAKER_01In a quarter report.
SPEAKER_00In a quarter.
SPEAKER_01I don't I I if I was gonna send them I'm gonna go, I'm gonna make a wild guess. 1.2 billion dollars.
SPEAKER_00Right. 96.2 billion dollars in a quarter. Apparently that's up 106% on last year. So I mean they're somebody's very happy, and they presumably won't care too much whether it's on a Raspberry Pi or in a data center, as long as you've got an NVIDIA chip in it. Unbelievable, nearly a hundred billion dollars a quarter.
SPEAKER_01I don't know if this is true, but I read something that uh I've forgotten the chap's name now, is it Jason? Uh the the guy who runs it all, the rock star of he's always wearing his leather jacket and looking cool. Jensen it apparently he bought uh the majority of shares for $200 back in um 1997 or something. Can you imagine? Uh kind of unbelievable. I bet I bet he doesn't need to run up and do a podcast every Friday.
SPEAKER_00Well, we would do it for the love anyway, wouldn't we?
SPEAKER_01We'd do it for the crap if we were that rich, Kieran. But we might be doing it from uh boats or or more sunny climbs than the Caribbean somewhere.
SPEAKER_00Let me tell you about the housing thing, shall I?
SPEAKER_01Oh well, speaking of houses, yeah, let's let's talk about housing.
SPEAKER_00So yeah, really good, really good session. Um, I was uh there on the Wednesday. It's a two-day event. Uh they get loads of people from the housing community come together. And um really, really good fun to be there. I was as I mentioned last, I was there last year as well. This time I was there speaking just me, not with a client, which is always a bit alarming because I get it that if you you don't really want to go hear from a vendor these things because you get a very warped view of of the world, I think.
SPEAKER_01So um well, even even with even with a customer there, lots of them are are so rubbish because they're effectively selling the fact that they've bought these bloody things, and I think lots of those things are so skewed anyway. So yeah, anyway, you were you were you were up front and central on your aunt.
SPEAKER_00I was, and it was um I had almost a full house, so it was a restricted session, which was good. It was a breakout session, um aimed at CEOs and chairs, um, had a wider range of folk there. And it was firstly I kind of set up with what's going on now in housing, and the big message for most organisations, but the stats are very uh telling in housing, is big usage, massively grown. Now 76% of housing staff say they use AI, which is great. But when you go to the other side, tenants 69% of of the public don't know that AI is being used at all in public services and don't trust it to be used. And so you've got this kind of two sides of the you know, one side we're using it a lot, and then the tenants that don't know. The other challenge is 87% was this figure from uh from a report um from research. Happy to post those in the show notes um uh that said that people would felt their AI skills were very low. And they didn't, and only 40, I think it was 44% don't have a policy still, an AI policy for their organization. So what you've got is massive usage with people that will openly tell you, it turns out, that they don't really know much about it and no policy in what you should or shouldn't be using. So that was one of the things. So much. I mean, the the well, we know, I mean, they're using tools, so much shadow AI going on. I heard of one housing association only two weeks ago they put the firewall up on Chat GPT um because they're just uh you know aware so much of it's going on, so you can only use their uh prescribed tools. Yeah, but yeah, massive challenge. So that's one. Um we talked about.
SPEAKER_01And I just sorry, before you just go on to that, did you see? I posted it on LinkedIn that I've got a friend who lives in uh a housing association that will not be named, but uses AI um for handling complaints. And I told him this yesterday, and he said, I put in a complaint last week about something that's been going on for two years. I said, Do you use AI to create? Yeah, of course I did. And I said, How long was the letter? He said, Oh, it was about 10 pages. And I said, I said, Did you get a response? He said, Yeah, I did. I wonder if it was your AI that re replied. I was like, Well, there's no could you tell it was an AI response? And he said, No. And I said, Well, in that case, I've got no, I've no idea. I don't yeah, that was it was it was hilarious. He just well, I we were just chatting about AI in general, and then I mentioned the complaints thing and it being such a big deal. He's like, Yeah, I got a response to my complaint, and probably from your probably from your tools.
SPEAKER_00Yeah, but that is good. Well, complaints was the big thing that one of the other stats I talked about because it's up 90%. In fact, I just saw an article this morning from uh is it from anyway, some some publication saying it's a it's uh doubled. The number of complaints since ChatGPT has doubled. Uh and our blog this Sunday goes out is on that same subject. And there's a really good side to it, which is you're now enabling AI enables people that otherwise would find it very difficult to write a coherent, you know, compelling, strongly worded, or whatever they want to say. So uh, you know, to represent themselves, and in housing and in education, getting the voice of parents, the voice of residents, and really understanding what's going on for them is important. So good up to the point where you're then getting vexatious claims and indeed people going for the maximum kind of well, I'm entitled to all of this compensation, you know, and they start seeing the money as opposed to solving the problem. Yeah, um, so it's a huge challenge, but volume is the big thing, and and some housing associations are using a bit of AI to some to to try and understand the complaints because they're getting 10-page letters. Um, nobody's using it, well, nobody in that group was using it yet in the way that Knowledge Flow would would deliver an answer for you with evidence, with trust baked in. Um so that's uh the the big point that we keep making is your residents are are empowered with these tools. Your teams are now double, you've doubled the workload. There's no change to the duration of what your what the rule the code says about how quickly you've got to reply, and and long much, much longer responses, not only twice as many, but probably ten times as in depth. So you need AI on that side too. It's uh as you know, is the big thing. So we were talking about that. Um, and then we did a lot of table discussions, which was great, and let people kind of think through the pillars, the the strategy pillars. So the sort of three big things people and culture, data and governance, and use cases and roadmap are the three sort of pillars of your AI strategy.
SPEAKER_01Six pillars.
SPEAKER_00Well, you could kind of say that, couldn't you? But it's three is a much better number, I was told by a very learned, a very learned man told me you should always answer in three. So one, two, three, many is how people think. So um, and they they had uh we didn't have enough time to get properly into conversation, but it was nice to get you know, hear some really what sort of things came out? Um, so the big question was because one of the questions I posed was, How do you know you can trust your AI's responses? And the big headline was we don't know if we can trust it. That's like I mean, we just don't know. And there is you know the sort of whole point really that I wanted to kind of get across is you need to know, don't you? If you're if you're gonna use this to answer complaints and deal with sort of resident challenges, then you really gotta know what what it's using, how it came about its answer, and not allowing the kind of co-pilot approach of it found some answers. And they seem, you know, as we know, you read it, it looks amazing in its response, and you can you can't check all of it, so you just you know send it. So uh and I came up with some things in the report that I've sent them. I came up with some questions that they should ask about things they're currently running, uh, which is do you do you know? Here's five questions, right? Five questions that you should ask of your uh team about the AI. One, which documents does it read and who keeps that current? Two, can it show its sources so an answer can be checked and rather than just trusted? Yeah. Um who built it, who signed it off, and who reviews it? Yeah. Number four, what happens when a policy changes? Does the agent change with it? And then finally, would you be comfortable if this answer reached a resident, the regulator, or the ombudsman? And if you can't answer those, then you really need to go back and have a pretty priority look at what you're doing with your AI.
SPEAKER_01Yeah.
SPEAKER_00And that was one of the things that came up was about the idea of people creating copilot agents. A few people were were ahead of the game a bit and doing that. Yeah. Um, which is great, and but they they raise themselves. But how do we know they're right? Who's checked them? Who's uh ca team A have created one and now they use it, and maybe a few other people have got onto that. Yeah. But is it is it right even at the beginning? Never mind, is it right when the regulations updated next year?
SPEAKER_01And yeah, yeah.
SPEAKER_00So yeah, so that was the kind of the the nub of it all, but it was really good, huge engagement. They were a lovely audience, actually. For the you know, I've talked before about staring at an audience, and you see this sea of faces that just kind of only think, oh my god, they hate me. There was so much smile. Oh, I had it. I I mean I always get it's that voice in the back of your head, and when you're presenting, it's like that's different. Well, it but you see it in the room when they're just staring. But there were lots of nodding and smiling throughout, so I thank them because that really helps. It helps at the time, makes in my mind it makes me perform better, thinking, well, this is landing a bit. I can I can see that you know this is relevant, and I I think I do a better job for them rather than oh god, they hate it, I better move on quick. Yeah. So yeah, it was very good. I really liked it. And I met Kirsty Walk briefly as well. She was hosting the main event. That was quite fun. But she was pretty focused at that time, so I didn't really chat to her.
SPEAKER_01But you when she coming on to talk about things then.
SPEAKER_00Oh yeah, I still I forgot to invite her to talk about it.
SPEAKER_01Oh don't next time I see her. I'll share this with you though. Uh I um I was in a um I was in a certain establishment uh on Monday evening and I saw a friend who hadn't seen for absolutely ages, and he works in a different part of the world. He um and by that I mean he works in uh investment and banking and all that nonsense world. And uh and he said to me, he said, I haven't seen you for ages, and when am I coming on your bloody podcast? Then why don't you haven't got me on to talk about my AI stuff? And I was like, okay. So that was that was a genuine surprise. How did you know I had a podcast? Yeah, exactly.
SPEAKER_00How did you know? You're not our audience. Yeah, that's all right. Well, Johnny Hoyle could have a call out on the uh shout out on the web on the webinar about it. So he said, I love listening to it. Yeah, we should get him on.
SPEAKER_01Yeah, that's a great idea. Yeah, let's get Johnny on. He can talk. He can talk about it.
SPEAKER_00Johnny, here's your invitation to uh get in touch with us. And uh get in touch with our with our agents and he can give me a shout, Johnny.
SPEAKER_01I'll sort it out. Kieran's rubbish at the admin. It's true.
SPEAKER_00Sadly. Oh well it sounds like you had a great time in in Liverpool then. I did indeed. It was very, very good. I was delighted to be there. It was lovely. The feedback afterwards was really nice. Lots of people came up and were like, that was the best session I've been to here. And uh one guy said that he'd come from Spain specifically to see it, which was uh I was like, Well, that's amazing, but you should have just got in touch and I could have saved you a flight and done a Teams meeting instead.
SPEAKER_01Have you heard of Teams? Thinking we could have just done this by a video record. Yeah, that was no, that's a that's that's highly complimentary. That's really good. Well done. Well done. Thank you very much. Yeah, well, I can't I I've had a much um uh less busy on the LDI front uh um uh uh activity this week. Um but there was a couple of things that I wanted to pick up. Um I got uh I got an email from someone who uh uh we've been uh a customer we've been working with, and and they said, Oh yes, well blah blah blah, we've been very busy and we haven't quite got to this up. Anyway, uh one of the team has put a whole lot of um uh confidential information into Claude and produced this three-page report. I was like, which bit of stupid is that? It's like you know, we created a private chat GPT for you. We created tools specifically so that they um they didn't go outside and you're going, oh, look at this, it's lovely, it's formatted beautiful. We can fix the formatting, it's not the point. The point is you believe people are putting confidential data out. We've already discussed losing confidential data already, but yeah, nuts, absolutely nuts. So a big part of the adoption challenge. And I saw that um the adoption thing is coming back around again. Uh Tech UK, who um we are members of, um, are doing a uh uh a whole session on uh AI adoption uh later in the month, 22nd of September, I think. So uhsk one other team to go to that because I know that I can't, but um uh I think it's uh such an important topic for us, and as we discussed, you know, the whole disc pilot and learning from from colleges and universities about adoption in in their organizations and how do we take those lessons across to to to other sectors as well.
SPEAKER_00So and that's what that I mean that that three pillarslash six pillar strategy. I mean, the people and culture is first on purpose. Yeah, and it is about you know what are you teaching, what are you training, you know, giving a lunchtime webinar and prompting isn't good enough, frankly. You need a bit more than that, and who's in charge of this bit and that bit. So yeah, it's really really key to hold how you're gonna get that right. And data pillar two in all of that area is is the the big challenge, and getting your data organized so that AI can actually do the thing for you is where everybody needs to be, as we've talked about many times.
SPEAKER_01Yeah, one of the other things that um I did hear from one of the team was that um uh actually I I had a conversation with a customer uh who said, Oh, I can't get the perpetual memory working. So I asked uh I asked Ben if he'd have a chat with him, and and he it turns out he didn't want perpetual memory, he wanted context uh uh setting, and context is king, as we knew from the Gartner conference. But it was really interesting. I thought they'd sorted that stuff out ages ago, but it turns out that you know uh things like new people being added and new information. So, how do you keep that contextual stuff up to date for your organization? And and um uh given the the the areas that they're working in, that's just going to become even more important. And I know we've got some technical solutions, but we're back to the whole data and curation and keeping things up to date and making sure that it's all all live. It's not it's not kind of want and done, and I think that's part of the problem for lots of people, is because so much happens. I it it feels like so much is done in the background on the free models, you know, it's all upgraded, and you can pick which one you need for which task, and da-da-da-da. It's all done for you. Actually, if you're running your own private model, you there's actually quite a bit of effort in in making sure it's up to date and and people underestimate that um is my personal.
SPEAKER_00Exactly. Well, and they and the argument is that what the the the initial response to that is well, copilot 365's plugged into our SharePoint, so it's great. And you're like, is it? Yeah. I did share with the uh group on Wednesday that um if you ask Copilot 365 what redundancy conversations have happened, see what comes up. You know, you've got to be really confident that you haven't saved something in the wrong place because it does. I mean, copilot 365 respects the your own the access of your users, you're yeah, you know, they're what they're allowed to go into. But who's not saved something in the wrong folder?
SPEAKER_01Exactly, because they haven't checked when it comes when the serve comes up and they don't follow, you just check the file name and then you can't find the bloody thing afterwards. Exactly that, yeah.
SPEAKER_00So it's uh yeah, and not notwithstanding that, of course, it's scouring across your whole 365 is useful for some tasks, but it isn't useful for a response to a complaint where you need to know it's the right policy and the right information, not just a draft version that's sitting around somewhere, a half-written thing that was, you know, whatever. So yeah, it's um getting your data organized. As always, there's always something to do, isn't there?
SPEAKER_01There is always something to do. That's that's why we do this on a weekly basis, because there's always something new. It's not um it there was uh I was chatting to somebody else in a in a different sector, and they were like, Oh, things just don't change that fast in our sector. We've got nothing to talk about. Like, right, right, okay. That sounds like a boring place to work then.
SPEAKER_00It's all right. Well, this is the white, it's perfectly good for people with brains like ours that get bored easily and want to be on the next thing.
SPEAKER_01Oh, something's shiny.
SPEAKER_00Yeah, cool. Right. Have you what else is on your list? Well, I've only got um I was talking with Donald yesterday about Claude watermarking. So, as you might remember, we talked about it before, from the 1st of August, the EU Act requires AI-produced work to be tagged or in some way identified as such. And Claude have added watermarking into their stuff. Um, but Donald has got under the skin a bit more of how they're doing it, because it's not like they write this was produced by AI in text we're talking about specifically here. They're doing a really clever thing where it is the way that they uh the way that the tokens are put together in the response has a sort of thread through it, which is enables you if you are Claude and have the right unlo unlock tool, you can then check and say, Yeah, that's definitely written by Claude. So really interesting, but apparently entirely uh you can overcome it entirely by copying and pasting it to a different AI tool and then to saying rewrite this, but retain everything in it, and it will all that token, whatever they're doing, the little trails they're leaving in the tokens, is gone. So um it's it's yet another. I was thinking about for the AI detectors that run in education, so Turnitin is a very big one. There's a few others, um zero GPT, or is it GPT zero, one of the way around. Anyway, there's a bunch of them, and none of them are a hundred percent, let's say, being polite. Um, you couldn't rely on any of them to actually make a a decision. But um, yeah, this I was thinking they're gonna be gone if watermarks are clear and easy to spot, but it turns out all you gotta do is go, right, take that into ChatGPT, rewrite this for me. Yeah, and there you go, you'll get away with it until ChatGPT put them in as well, of course. Now you need to find one of these scurrless Chinese models. And uh I shouldn't call them that, should I? I'll I'll take that out. Find one of these Chinese models that will then uh redo it without a watermark. Yeah.
SPEAKER_01Um just talking about education. Um, one of the things I read this week was uh how to teach the next generation in an era of AI by a guy called Daniel Suskin. I don't know if he's the one related to Richard Suskin.
SPEAKER_00He's his son. I've got the I've just got his book on what what to s what to tell your children or something.
SPEAKER_01Yeah, I've just I've just got hold of it. Exactly the one. So um uh uh let me spoiler alert, I'm gonna tell you what he says. Oh dude, yeah, please do, yeah. Take me uh take me reading it. Well broadly yeah, he's if you master about the time. Uh so broadly speaking, he says uh it's a bit like when the calculators came along and um actually I'll I'll I'll set a bit of context first. Something called the PISA scores, which measure how well um uh education systems are working around the world, uh both literature and numeracy have dropped off massively in the last wee while and uh people were expecting them to recover after after the pandemic and they haven't, they've actually got worse. And part of the reason they believe is is around AI. And what what Daniel says is that um it's a bit like when calculators were introduced in uh to maths lessons, and I know that you've got a picture in your slide deck of teachers protesting against having calculators in like 1976 or 77 or something stupid. And um and he said it's it's a bit like that moment, but the way to deal with it is to uh train with AI and train without AI and so that children can understand the difference. And that's what the education system needs to do. Is it how to use the tools and understand them and but actually if you don't have the tools, what's the difference? And I should have been much more um thoughtful about the approach to teaching these things. So that that's that's the basic premise of what he's he's gonna push. So I don't know how many pages, 300 pages, that's what it was down to.
SPEAKER_00Right, there you go, as always. Oh well, I will I will have a listen because it is very relevant, particularly education. The future of education is obviously close to our heart to many of our customers. So it is a chance.
SPEAKER_01And to us for your for your children and my grandchildren. So yeah, no, it is I I actually thought it was a really interesting article, but I I hadn't clocked that it um he was uh uh related to to Richard Suskin. But actually I figured that there couldn't be that many Suskins on the old uh um uh uh media uh environment. So uh indeed.
SPEAKER_00He wrote I remember hearing a a little extract from one of his books, I think it was about future of work or something, um, and he wrote about the horse manure crisis in New York of 1896 or whatever. It's a fascinating story, and it's I don't know if we've talked about it on here, but it was if I try and pracy it, it's they used to have so much horse manure in New York City because of all the obviously horse-drawn everything. Um they were filling like fields with you know meters deep or feet deep, I guess, in uh in that time, and particularly in the US, uh, of this stuff. And then so they were putting out like loads of how do we solve this problem, and people were spending loads of time on innovative solutions for what they do. Um, and then someone invented a motor car, and all of that went away overnight. Yeah, and it's and it's really interesting because you know, all the kind of big challenge, probably the the biggest challenge in New York at that time, it was causing health crises and all kinds of stuff, literally went away over well, hardly overnight, but very quickly. With a new technology going very interesting, I thought.
SPEAKER_01Yeah. Well, uh, listen to uh Daniel's book and um uh yeah, I I thought it was a really interesting article that I I read on it. So that was good.
SPEAKER_00I certainly were. That's my weekend reading for you or listening. Is it more probably more likely?
SPEAKER_01Yeah, more likely. Um there was what there was one other thing on my on my list for this week, uh, and I mentioned this last week was about uh Chinese uh authorities clamping down on uh AI relationships. And um uh but what I hadn't realized until this week was um they're only doing that for Chinese citizens. They're very happy for their companies to sell AI uh relationship uh uh assistance, bots, whatever you want to call them, AI boyfriends and girlfriends to other countries. So anybody outside China, they can and and the main one uh has 10 million outside of China alone. Um so I'm assuming maybe wrongly, but I'm assuming that's kind of Southeast Asia. But um it really interested the Chinese uh authorities very keen on one rule for one rule for their citizens and and and a completely different rule for everybody else. So uh I don't know if that's if that's their um uh world takeover model is to reduce the uh birth rates of all of their neighbours by having everybody have such relationships and no and no actual children.
SPEAKER_00Wellzers. Well, there is a lookout for us all, and uh if you haven't got an AI boyfriend or girlfriend, then uh you know maybe now's the time to have a look. Ten million of them. That's quite a few, isn't it? Wells it.
SPEAKER_01Well, I don't think uh I'm not sure, um not sure Mrs. Watkins will allow me to have an AI boyfriend or girlfriend, so uh I'll put I'll steer clear of that if that's all right with you.
SPEAKER_00Indeed, yeah. No, indeed. Well I spend enough time at my screen without uh without having that as well.
SPEAKER_01I think that's why I've got my eye problems. I'm spending too much time at the screen and not looking at my AI boyfriend or what they say about that stuff.
SPEAKER_00So no, no, I don't what do they say kids? No, I can't put it on the podcast, but you know what will make you make you go blind.
SPEAKER_01On that note, I'm gonna do the other thing that's probably gonna make me go blind and go to the pub and have a few beers and chat to my mates about sheep and farming and pheasants and whatever else goes on around these here countryside parts.
SPEAKER_00Very nice. Well, and I will uh given we've finished a couple of minutes early, I'm gonna nip up and collect my son as a surprise because he I don't get to pick him up very often from school, so I'm gonna do that.
SPEAKER_01Excellent. You do that, fella, and you have a great weekend, and I'll see you next week.
SPEAKER_00Nice one, and you take care. See ya.