Sweaty man-boobs

This week’s podcast starts with (what Neil thinks is) a hilarious story about Kieron’s brush with a makeup artist who had (what Neil thinks) is the worst job in the world. The clue is in the podcast title. Poor lady…

Kieron also talks fashion (I know… who knew) and how AI is not just disrupting that world, but an app that just a few years ago would have raised millions as a start-up is now just a prompt.

We talk about AEO being the new SEO because LLMs don’t care about the adverts. But why, rather than writing your website copy for SEO engines, you should be writing it for humans. Peter Went’s humorous book Brandw*nk is a great read about the boll*cks people write at work and why no-one says “leveraging synergies” in the pub.

It all comes down to trust in AI. Related to that, we talk about the future of newsletters and how Nate B. Jones (their favourite AI guru) has the fabulous superpower of being able to talk at both end of the Strategy/Techy-Detail spectrum in language that anyone can understand.

Big news that JISC, the UK’s education technology body is going to run a pilot of KnowledgeFlow for up to 10 universities and FE Colleges.

And does being nice to your AI make it want to work harder for you? There's two ways to find out. The easy one is to listen to this podcast.

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. 🍺

Transcript

SPEAKER_01 0:00

I can hear you now, which means we can get the pod this Pantomine Horsebell Podcast underwear. Right. Have you you've got your shirt on, Scruffy? Have you been through hair and makeup yet?

SPEAKER_05 0:09

I had a makeup artist one time and one of the things I did. Yeah, did not, yeah, for VMware. They they had me do some stuff for their CIO uh sort of VIP groups. Yeah, yeah. I had to do some filming on it was the hottest day at the time, a few years ago. And it was like something got 35 or something, and the and they said, Bring bring a variety of clothes, and then our creative director will choose what he wants you to wear. So he put me in the bloody, he put me in the bloody pullover, didn't he? So I'm in the jumper and sweating. And they had a makeup artist, and I was like, that's going to be weird. And so I had to see the makeup artist. And she was like, Well, you don't really, you're not really very sh shiny. I'll put a little bit of anti-shine on. Guess what though? She spent most of the afternoon drying the sweat underneath my man boobs with a hairdryer.

SPEAKER_03 0:59

I mean, that's that's a rubbish job, that isn't it? That's the worst job in the world, I could do it.

SPEAKER_05 1:04

She had to do it for Downton Abbey when they filmed their Christmas special in the summer and they were all in these massive coats. But I was like, ah, here I've got a makeup artist. I finally hit the big time, and her whole day job is holding uh as soon as they finished filming a scene, they then run me off, and I'd have to go and get dried again under my man boobs with the hairdryer. It was it was a humiliating day. That's brilliant. You've never told me that story. That's not well. There you go. Well, now I've shared it with you and Matt on the audio as well.

SPEAKER_00 1:37

Now three of us know, or um the poor makeup artist who's probably still in therapy as she's trying to recover from the audio experience.

SPEAKER_05 1:45

She was very good about it in the way that you would be going, Yeah, no, this happens all the time, sir.

SPEAKER_02 1:53

Dear me, poor, poor lady. Oh my goodness, crack it. I don't know where to go from there, Kiry. Should we try and talk about something else?

SPEAKER_05 2:00

Well, I can take you straight into a good segue because there's a bunch. Um, this is um I was reading this thing this week on 5.6 Sol, the chat GPT's latest model, or one of them. Um, they I saw this uh post from Sam Altman, the CEO of um uh OpenAI. Yeah, sorry, yeah, he he sends me posts. I think it's just for me. I think it's contacted me, yeah, yeah. I think it's just for me, yeah. I'm not sure. There seem to be some other people liking it as well. But anyway, anyway, this is a wardrobe uh app. So uh my other half was very excited about the potential of it. And you basically take a load of photos of you in your clothes, and then it creates like a kind of what looked like a kind of shop for you with all your clothes, and you can put them together in different ways, and it can suggest different colours of the same things that fit you well, and blah, blah, blah. Um, the point was Sam Altman's comment on it was this would have been a whole new startup not long ago, and now it's something you can just do. You just get your phone out, you know, spend a few minutes or probably a couple of hours taking photos, write a bit of prompt, and you've got one. It's kind of like, what does that mean for the world? I mean, this is the AI disruption of all those kind of things that yeah, that would have probably raised five million in funding and had a little crack out there and tried to sell a few of its uh of itself. And here it is, just do it yourself.

SPEAKER_01 3:25

Wow, wow, that is incredible, isn't it? I was thinking about the piss of um uh developments, uh, not just because of the the models and the new ones coming out and tools like that. I was thinking about it for ourselves, really, because um this last week, I I think it was last week we joked about um Ben winning a bet because uh Donald said he wasn't gonna put any any make make any new developments and he'd failed by Wednesday and uh Ben was was still still to collect his money on that bet. But uh again, similar thing this week, it's just like the whole kind of um uh like 10 past eight this morning. Are you are you busy? Can I borrow for a second? Can you just I've just I've created a way I've created a web app so that you can have knowledge flow directly into your desktop. So it's just like can you press that button? Oh yeah, it works perfect, right? Okay, crack on. Brilliant. I guess yeah, amazing. But the other thing that kind of made me think yesterday was I um were we were pitching to a uh healthcare organization, and um we've got limited um uh staff in healthcare currently, but good enough to to to definitely talk about. But they were talking about uh some features which I said I think I think we're doing similar things in um housing. So uh we we demoed um some of the stuff that we're doing in housing, and they could see immediately the crossover, the kind of relevance, the kind of how the application and it was uh but the but what struck me was I hadn't seen half of the bloody things. Uh it was all it was new to me. I was like, hang on a sec, how did I not know about stuff that we just do it ourselves? It's and so that whole kind of pacer development and working with customers and developing new stuff and and getting feedback, it's just uh it is incredible. And that whole kind of you know, that step-by-step stuff that he's done where you kind of yeah, do this, do step one and da da da, and then it tells you what to do on step two. So there's kind of there's there's no room for kind of um error in in the in the approach. And and what these health people were saying was we can't afford for there to be any deviation from the process because it's got to end up in looking like this, is that well? Yeah, just showing the step.

SPEAKER_05 5:23

And that's what those are as you know, that's what they're designed for, and also with the regulatory angle on that, with knowledge flow being able to capture the audit trail of what happened and each change and everything that went through that process, you've then got something that should you know, should it end up in a complaint or you know, something even more formal than that, then you have the full trail, which is the part that um other apps are not able to do, frankly. And of course, all of that is stored inside your own Microsoft Azure tenancy, all down there for you securely. So, yeah, really good. Yeah, I'm glad to hear that um they can see the reader cross. Often, as we've talked about before, customers find it sometimes very difficult to say so to see something in a different sector and see the application in theirs. So it's great if they're they've got that that innovation, that imagination. So good stuff. That sounds encouraging.

SPEAKER_01 6:11

It is interesting because uh Helen uh Mrs. Watkins will say to me, Oh, yeah, I can't imagine, you know, I struggle with the imagination piece, I can't do the imagination thing. And I I think we've got some customers like that. There was the it was the kind of the story we were talking about on stand up of you know the customer going, oh yeah, that's good, and then just moving on and not doing anything about it, just kind of not not seeing how things would develop. But these guys, these guys didn't lack for imagination. What they lacked for was common sense. It was like, oh, and can it do this? And oh, can it do that? And could you just it's like we could do all of those things, but yeah, the two things uh it takes time and effort and money, and uh I'll come back to the money piece in a second, yeah. And uh and you need to prioritize because you've just given me like 87 things you want doing. It's like well, where do we start? What's the last week?

SPEAKER_05 7:00

And what has happened is they can I get this a lot, is they want all these little things. Well, it's got to be like that, it's gotta do that, and then they'll come and tell you that's got to be really cheap, and it's that same old project management triangle or whatever it's really for. But I always think of it as that is scope, time, and quality uh and uh budget. If you move in one, another's gonna move. That's why I f I kind of not fell out, but if you remember, there was a customer which um I ended up um pricing high so that they would say no to us, and it was exactly that because they were kept on going, Well, we're a charity and we've got no money, oh, but it's gotta do this, and you're like, it doesn't need to do that, that's a preference. What it does, what knowledge flow does as a platform will do what you need really, really well. You are just adding these things because whatever reason leaders and managers add things, whatever, sometimes for right, and sometimes just because they thought of it just now. And it's yeah, I wish people would get kind of understand if they want that bespoke solution, then you're not gonna get it as a as a normal price.

SPEAKER_01 8:01

No, indeed. And we also suffer uh as we have done with a couple of customers recently, where somebody new comes in and you've got to start again, or a new IT person comes in, and here's my 30, here's my 30 questions, which I've just knocked out of chat GPT that you you now have to answer within 30 seconds, or or you or uh or I don't like you. It's just like absolute bonkers. Just uh yeah, this is a good thing.

SPEAKER_05 8:27

And the fact that they come to us on that is a bit annoying to me. I mean, it's inevitability that uh the one we were talking about on stand-up today is like, well, all of this stuff exists already, you've already asked all of this. We've you've got paperwork on all these things, you should be asking yourselves before you come to us and having meetings with us and asking all the same questions again.

SPEAKER_01 8:47

Yeah, yeah, it's uh it is your predecessor went through you made us jump through all these hoops, and uh we passed all the tests, and the contracts are there, and your purchasing people went through it and and and but you're just being lazy and just trying to make that you you've come in new, and you uh you need you feel like you've got to make an impression and and and you want to do something, and and it's it's feels like the wrong thing to do, but that's frustration with customers. But uh, but I'll show I'll show the the cost thing. I thought yeah, so yesterday we talked to our good good old friend Cameron Merzer, and then there's a great podcast episode coming out for those who uh uh uh uh are interested. But um one of the things he said that really struck me, so Cam, really fascinating chat, uh, but he he now works out in the Middle East um for uh an NGO uh supporting um uh higher education across uh Middle East and North Africa. Uh and he said there's a number of uh senior leaders in um uh education in in uh especially in Africa, who think AI is free, and you have to tell them that it's not free, and they're absolutely gobsmacked that it's not free. And uh and I was I I'm and I was and I was like, hang on that's really how how can you know anyway? I'm on this call yesterday with this healthcare group, and um and then the guy goes, uh and we start asked I mentioned cost, and he was like, What? He said, Oh, is it not free then?

SPEAKER_05 10:11

I was like what's just turning up and spending our days doing this stuff out of the goodwill of our hearts.

SPEAKER_01 10:20

It's not just that, it's because people are fixated on the kind of whole uh yeah, I can get I can get onto the free chat GPT and uh I can get onto these free models and I can use Gemini or whatever and I can use them. It's like, yeah, you can, but for all the reasons we've talked about, like security, like accuracy, like trust, like confidentiality, like GDPR, all the things that you say are important to you. Funny enough, there's a cost that comes with that stuff. So the the fact that they said it, I was absolutely are you serious?

SPEAKER_00 10:49

Just unbelievable, unbelievable.

SPEAKER_05 10:51

Crazy. They just keep shoving their data into the free tools instead and getting crap answers and uh muddling their way through.

SPEAKER_01 10:58

That's right, so even GDPR call.

SPEAKER_05 11:00

Yeah, exactly. So I had um uh Nate B. Jones, our favourite little um uh person we follow, he accepted my LinkedIn request at last. So I don't know what changed for him. I reckon he's obviously heard about us and uh is he listening to the podcast? Yeah, I imagine he's our only listener, perhaps now.

SPEAKER_00 11:24

And uh anyway, so I was very I was very excited. Um he's happy to take my money for his substack uh post because yeah, yeah, but you know, it's good.

SPEAKER_01 11:34

Actually, no, he he is really excellent, and I've said before, you know, he is our he's he's our guru, and uh we listen to him a lot, and he has some great things to say. Uh actually, both what I really like about him is right in the he's a bit like you, Kieran. Um uh he's yeah or so you're you're a bit like him. I wish I was like him. Yeah, yeah, yeah. So he comes at things from two completely different ends of the spectrum, uh, which makes him really interesting as a character because he goes right into the weeds, he's got his own benchmarking that tool that he's created, he he tests all the LLMs, he goes into some detail, he goes into real detail, real hard tests for for these. And and then he comes at it from a kind of uh strategically here's what you need to do as a CEO, and uh here's all the things you need to think about in in terms of your organization, your customers, and and the direction of child of your business, and his the ability to kind of uh both do both things but also bring them together is actually I think one of the things is quite unique because um uh there is a lot of nonsense talked about AI on the um uh on the interweb. And um uh and he's one of the few people uh that I kind of uh kind of he's one of the few go-to people that I've got. So I really I really like him. Um and that kind of links to something that else I wanted to talk about. Uh so there's a guy called Peter Wendt who uh we've worked with in the past, he's a really nice chap, had a really excellent, he's really good at the at the marketing stuff and and uh really good at the messaging piece. And uh he wrote a book called Brandwank, which is um really interesting and uh talks about um how corporations and organizations, especially on LinkedIn, if you look at someone's LinkedIn profile, uh it's just absolute gobbledygoop nonsense. And his his question is what does that even mean? What does that mean? What are they saying? I don't know what they do, or what does that mean? And he's really, he's really, really good. So I I I uh uh I was uh uh one thing about Mate is is actually he can translate that that technical nonsense into um into simple language that anyone can understand. But yeah, I'd highly recommend Peter's book because uh it's most entertaining and uh actually really good um really good messaging uh on um uh language in general, but also how do you kind of not fall into the trap of which lots of people clearly are of using LLMs to to write their marketing, their LinkedIn profile, their website copy. And actually, his his real major point is if you do that, you're not gonna stand out from the crowd, and therefore whatever you're spending on advertising or whatever you're doing is just gonna be wasted because you're just one of one of a gazillion others. So yeah, really interesting um perspective.

SPEAKER_05 14:19

Yeah, it's good. And I I think uh your point on Nate and trustworthiness. I mean, I've said a few times to to customers, I think the future of brands is trust, given uh, and I'm and really what I'm talking about is where I was talking to the Kings Fund, um, who are a healthcare, what do they a think tank, I think they call themselves. Um, and uh in the UK. And um, you know, what in a world where you so they do all kinds of things and they do research for customers and lots of lots of things, and their website has lots of kind of answers and the latest reports and things. In a world where AI can get you straight to a lot of that stuff, yeah, there is clearly going to be an erosion of their market. And true of so many organizations, particularly if you're in the information sharing collation business. Um, but the key, I think, is the differentiator's trust. And it's a bit like you know, if you use perplexity for search and you have a look at the sources, I'm much more likely to look at the stuff that's you know a trustworthy source, the FT or something, than I am from unknown caravan weekly website or whatever. Um so I think that that that's what brands should be thinking about is how do we maintain the trust over our stuff and make sure that what we do is good and accurate and we continue to put it out. But the thing that Nate has done, which is I think a a future for the world of newsletters and news in general, Nate's got his MCP server where he does he has all of his kind of content, you can plug your AI into it and effectively ask your version with all the stuff your AI knows about you of Nate's story. And so instead of reading something like we all do, and you know, bits of an article are relevant and bits aren't, the idea is that everything you read will be personalized to your situation, to the challenges you have, but obviously coming from Nate's perspective on whatever it is, that I think is really interesting. That idea that in the Kings Fund world, I mentioned it to their marketing team. Uh in the Kings Fund, they you know put out a newsletter, but use this version of the newsletter, and it will now only or specifically bend everything to the why that matters if you are a clinician in a hospital or uh a manager of somewhere else. I think that's really interesting.

SPEAKER_01 16:37

Yeah, the personalization piece, really good.

SPEAKER_05 16:40

Yeah, really, I think it's that's clearly a future, it's already happening. And I think that got you know, in the other sort of my other interest in the world, I've spent um uh not not insignificant amount of time in my career uh doing transformation programs with the big advertising groups, so with WPP, with Dentsu, Group M, um group M a part of WPP. Um, and the in so I'm kind of you know I'm always sort of slightly interested in their world and and the things they grapple with. And the effectively the SEO search engine optimization, so getting your your page seen on Google uh and on the first page. Here's the interesting thing with AEO, I think they call it various, there's various different phrases. I think people are still trying to win the the name for that new new version of how do you get your website seen by an LLM? Well, the first thing that's interesting is LLMs will um largely ignore any sponsored content. So if you're paying to be on top of Google, well, whatever, LLMs will look on page three, four of the search results, whereas it was always about getting on page one. I mean, uh this is all just stuff to think about. I've got no real expertise in in what you do about all of this, and then Nate talked about this a long time ago, is the way chunking works and the sort of token sampling, if you like, that an LLM does when it's doing a search for you is different. And um, it seems so you can either try and write your copy in chunk sizes that LLMs will like. It's one way. The latest thing I've been reading is making your copy more conversational is apparently working to get you onto the LLMs when someone's searching for what what snack should I what chocolate-based snack should I eat on my break today to make me feel joyous and happy? How does Kit Kat appear at that point? Is they need to kind of be matching that that kind of search as opposed to someone searching for chocolate-based snacks, which they might do in Google, you're gonna be saying, I'm having a break, and I'm wondering what kind of chocolate I should have.

SPEAKER_01 18:48

That's really interesting. So two two things. One is um come back to the Peter Wendt point about writing for humans, um, and I I get it, you know. Actually, if you if you if you just got if you've got AI word salad slop on your website, you know, no one's gonna understand it, no one's gonna read it. So you're making it more human. But also, how do you do that and make it LLM readable? Because actually the last two customers who've come to us unexpectedly found us, told us that they found us through Chat GPT. Now I was amazed at that. So you know, we we we'd not structured our website or our content for for that particular route. But actually, one of the things on my mind after reading Peter's book and and thinking about the NERT stuff is actually we need to do both. We you can't I think you I think you need to do both, you can't do one or the other. I think you need to do some kind of both. Now, uh some people might think that's uh uh heretical, but that's yeah, I think that's what I'm trying to try to a while ago.

SPEAKER_05 19:55

I've seen one website that has an LLM friendly version of their entire site. It's basically just one page with a markdown file of their website.

SPEAKER_01 20:03

Yeah, you you you mentioned that to me three or four times, and I think you've given up because you think I'm not gonna do it. That's not actually clearly well the reason I haven't done it is just waiting to get the website finished, aren't you? Yeah, the whole the website will never be finished. It's the fourth, you know that. Yeah, yeah, yeah. But the point the point it's uh all the feedback I've had from Claude and others uh is is that actually it doesn't make any difference. Oh yeah. It's actually the quality of the writing, not the fact that there's a markdown file, because it already strips out the kind of the block, the the Gutenberg block. See there you go, you know what a Gutenberg block is. Uh uh exactly. And actually, you don't need to know because actually it's strip it doesn't matter because it doesn't work.

SPEAKER_05 20:43

What would Peter Wentz say of that comment you just made to me? Would he would he would he approve of your uh throwing some difficult words at me to try and throw me off the scent?

SPEAKER_01 20:52

He would absolutely be cross with me for that. He just shake his head again and go, Oh God, Neil. Uh so yeah, but the point is that the the AI strips all that stuff out anyway. Then actually you don't need to just do it as a markdown file. Interesting, I wondered whether that was kind of virtue signalling on behalf of whoever's writing a bloody website. It's like, look at me, I've put my LLM markdown file down at the bottom so that everyone gets it. Because actually, people like you fall for it. Uh oh you're cheating bunky.

SPEAKER_05 21:19

Maybe it's the thing that'll make a difference. People like you are biased against biased against it for no reason at all.

SPEAKER_00 21:25

You've got I'm not buying. I'd like to see the evidence, Kieran.

SPEAKER_05 21:29

I need to try the evidence. There isn't if no one can really see it, can they? It's too difficult. Reddit is um, so I I see a lot of that in um perplexity search, and uh I hear that AI search want it because those fast moving sites, Reddit is I think is open as a you can search it, whereas Twitter have uh you know restricted the ability for robots to get in there. Yeah, um, it is uh sort of you get to get more current stuff. So I think they rely on it because you're getting things you know so fast moving if you search. It you'll get some things from right now. Uh, but then you also get the weirdness of all of that coming into your because Reddit is a bit like Twitter, and it's kind of like just random comments and trolling and all the rest of it that comes with it. So um, yeah, interesting. The whole world is a very peculiar future, well, not peculiar, fascinating future. And that sort of world it does interest me a little bit of how do you get audience attention still. And what will be the and it seems like, and this is great, there's a uh organization I've done some work with who I very much admire the Ozone project, um, who are uh uh I I I'm not even gonna try and use their brand marketing, but they basically work only for on behalf of advertisers work only with high quality journalists. Um so you if you work with them, your ads are not going to appear on Pornhub and God knows where else, they'll only appear on trusted websites. Um and I think that whole world of trusted media, the LLMs are using that too. So to go, so I think they may be well onto something. Now the question is how do you get somebody from a perplexity search to the website to see an advert? And that, or how do you get the advert into the perplexity site? That's the that's the next challenge for everybody, I think.

SPEAKER_01 23:18

Well, I'm going to concentrate on trying to produce um better content, which is very uh human-centric, it's very clear, uh, it doesn't use any of the it doesn't use AI W star star uh language, and um uh but also that uh there's lots of it so that um actually the SEO and the and Chat GPT continues to um put us out there as as uh one of the people doing trusted um uh private uh regs. So uh yeah, very um I'm I was super excited by that actually. The fact that we're we're actually making some progress with that and um uh as I say customers coming to us out of the blue with that is is really exciting.

SPEAKER_05 24:01

Yeah, it's good. So I was um this week, yesterday in fact, I was um on a or giving uh host what's the word I wasn't hosting it, I was on a webinar anyway, for with JISC for our JISC pilot. So JISC are in the UK. They're I think they're the biggest, the world's biggest provider of private um is it cloud private connection? Connectivity. Connectivity. So they run the use the eight.ac.uk um network, uh Janet as it was back in the day, probably still is, who knows? Um, but they also so they also are well respected or are well respected in uh education in the UK, as you know. And they uh we're gonna they've asked to they can pilot knowledge flow. They had a test of it uh back in January, in fact. There was a group of them, slightly slightly alarming time, frankly, because we gave them access to uh to our to knowledge flow, and uh that was their head of AI, their AI team, their you know, people that really know what they're talking about. Um, and they gave it a good hard kicking, and then we're like, yeah, we think it's really good, and we'd like to pilot it. So that'll be 10. We're gonna work with 10 institutions in H E and F E and we're going to uh have give give them each knowledge flow and the onboarding, and then at the end of a three-month pilot, there'll be an evaluation which JISC will lead. So I'm really excited. I mean it's mildly terrifying. But the webinar the webinar yesterday was great. We had lots of people come, um, and we showed off both the uh tools for colleges and the tools for universities. Nice. Um, and we had some a nice piece of feedback uh this morning from um somebody that is really well known in college land in uh in AI, saying that she one of her team had been and came back raving about it and said, We really want to be involved in this because it looks like a real step up from the stuff we're doing. And they are respecting most people in colleges would point at this college and go, they are the people to go and ask about it. So to have them see our stuff and go, that's what we need is uh quite quite accolade. So yeah, that was exciting. Nice, yeah, nice question. Cheery, cheery part. Well, let's see what happens.

SPEAKER_01 26:09

But was there much difference between um because uh for our audience as our audience knows uh H G and FE is not really my world, um uh because I'm dealing with other parts, you're dealing with the education piece, really. Uh was there much difference in the tools and things that you showed from university and and college, or are they because they see themselves as very different, don't you? Yeah, well that's what I was gonna say. Just whisper it, just whisper it, no one will hear it. Oh, wait, they see themselves as very different.

SPEAKER_05 26:36

They do, and here's the thing, because they unis hate it if you say, Yeah, we're doing this for colleges, and this is effectively the same tool, just changed. You can't you have to be careful. So they want to feel like they're very much uh much more important than a college, very, very different, far more diligent and uh academic. Um, no, they're basically all the same, just different language. Um, so like uh in college, they will do a thing called a SAR self-assessment review. We've got a great tool, as you know, that um will crunch through all their data and we give them that first draft of this is all the data that shows you're doing well under this heading. This is the stuff that shows you're not doing well, and here's gaps. I can't find any data. Um, and if you want, I'll write a sort of summary paragraph of what how you might write that in a SAR. Um, the point being, all that time you spend staring at stuff, searching documents, reading things, and hoping that you're going to take it all into your brain, or you let AI do that bit, which it's amazing at, at way better than any of us will probably can ever be. And then you engage your brain on the kind of what does this mean? What can I do with that? Oh, yeah, I know that's a problem, but we can never fix that because of this. Anyway, that's the SAR. AMR is a thing called an annual monitoring review. You know where this is heading. It's a load of data that you upload and you have to review and read, and then you've got to reflect on and decide what's going well and what isn't, and explain why based on the data. It's identical. It's just a different set of questions. Well, actually, they're not very different set of questions either, frankly. Um, but it's yeah, and they and so and I could labor that on and on.

SPEAKER_01 28:11

Come back to the point that we started this conversation some 20 minutes ago, uh, speed of change. So uh two weeks ago, timetabling didn't exist as a product. Did you show the time tabling tool?

SPEAKER_05 28:22

Yeah, time tabling. I did show the timetabling tool very briefly. Um, yes, and it it feels like timetabling is almost there, but then it what's hilarious about as soon as you it's as soon as you kind of get to that like that, there it is, then another thing, so whole can of worms opens up, and there's so many things, and we haven't even I mean, in many ways, we've we've got to we've got to a place where a university or a college can load some data about their room staff courses and it will produce a timetable for them that is compliant with everything they've sent, i.e., teachers teaching the subjects they teach, rooms that not double booked, rooms that are allocated based on the requirements of the room, labs, PCs, specialist equipment, um, uh staff working to the hours, etc. etc. But I think then is when all the kind of hard work starts because that's when Mrs. Miggins goes, Oh, you know, I hate teaching that class in that room.

SPEAKER_01 29:17

Yeah, and all of those things that we need to accommodate. Kieran can't teach in that room because the temperature never gets below 30 and his man boob starts sweating, so he can't teach in there.

SPEAKER_05 29:26

Bringing it right back to the beginning. I see what you're doing there. I should tell you one more thing before we try and wrap up, which is interested me in my own behaviour. We talked about um last week, we talked a bit about whether AI may be um uh conscious or not, um, and the the fact that it's probably beyond now just simply ruling that out and laughing it out of the room as at least worthy of a consideration, even if it is after three beers. I am being much more polite to AI since reading that stuff. I plead in nearly everything, and thank you. So that which adds a bit of tokens, but the worst thing you can do, as we've said before, is if you thank it for doing something, that's bad because everything in the context window goes back to the LLM just to simply go, you're welcome. But here's the thing I also read, which um was somewhere, it was a clawed response. I think it was a behind-the-scenes clawed response as opposed to to the user, or maybe in the reasoning. Um, it said, I'm doing this, blah, blah, blah, and the user has said please, so I should search harder. There was some phrase, I'm not sure that was exactly the phrase, but it was the it recognized there was a please, and because of that, was going to do something additional, which is uh really intriguing. So um, and and maybe it's coming full circle and to just simply that's part of language, and you know, everything it's read and knows about and sort of inferred in its own neural network is if someone's polite, you should probably work a bit harder to help them.

SPEAKER_01 30:53

Yeah, it's interesting because I I I think Julia probably wrote something about it about two years ago on on the blog. Um, uh about actually if you put pleas in and it kind of it responds better.

SPEAKER_05 31:06

It used to, and then it stopped it. Jeopardy worked better for a while, or just threatening it with or you telling it my job's on the line. If I don't get this right, it would improve the responses.

SPEAKER_01 31:16

Yeah, and I um and I I rem I remember um I think might have been your mate Sam Altman who uh you've been conversing with sends me post. Send you post, send you post uh saying something along the lines of uh it's worth burning tokens, you don't need to do this. Um I think they changed it. Um so it's interesting that it's come back round. I read an article from some professor of AI at a university, I can't remember which one it was in the UK. Uh I'll look it up, but who basically said, Yeah, it's nonsense, uh, this consciousness debate is just absolute rubbish. I was like, Well, we should go to the pub with you and have a couple of conversations with the case.

SPEAKER_05 31:52

Well, well, maybe I should ask him, yeah, I would like because you know, I hear that, I hear that, and I've met people that have been like ridiculous, and they're kind of like almost totally dismissive. But I mean, I think any, I mean, I'm sure he's a professor who's probably thought about it a bit. Hopefully, he's got more than that's rubbish. Um, but you know, the reality is any point that someone makes should not be treated in that way, in my opinion, because it's like why? And but secondly, I really I think that there need there's a there's at least another two or three sentences required after why that's nonsense. Yeah, yeah, you need to be. You better be damn sure you know what consciousness is. That's right. You need what's the definition? Approve it. Must it have a heartbeat, or if it could be powered instead of blood and oxygen, it is electricity. Well, who's to say?

SPEAKER_01 32:40

Right, well, I know you've got a busy old day ahead of you. Um so good luck with all of that stuff. I'm going to be uh held hostage on the train network in uh uh in super hot temperatures. So I'm going to hydrate uh thoroughly uh with relevant liquids and uh and I will catch you very soon. You have a good day.

SPEAKER_05 33:01

Good stuff, and you enjoy yourself wherever you go. Cheers, baby. Bye bye.