Is AI conscious? What does conscious even mean?

Plus a Troll, a Marking Robot & a Mystery Guest 🍺

Episode 20. Somebody told Neil his content is lazy and rehashed. He's decided being trolled is an accolade. Kieron got ambushed on stage when a customer asked a live audience whether KnowledgeFlow had helped or hindered them. (Spoiler: he survived. Every hand that went up said helped.)

The big question this week: is AI conscious? Kieron — who has a psychology degree — points out that you can't dismiss the idea until you can define consciousness, and nobody's managed that yet. Meanwhile AI companies are pillaging university philosophy departments and Anthropic is reportedly monitoring Claude's mood. Third-beer conversation territory, and it's brilliant.

Also this week: the hackathon that built an AI marking assistant in 30 minutes (and why AI marking is dangerously generous by default), the college trip risk assessment process so bureaucratic that teachers have stopped taking students on trips, Raven Housing's masterclass in doing AI adoption slowly and properly, and why Microsoft is throwing 6,000 people and billions of dollars at the adoption problem everyone pretends doesn't exist.

And a first for the podcast: a special guest pulls up a stool tomorrow. Tankard at the ready.

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_02 0:00

Speaking of good, should we get this pantomime horse of a uh podcast underway or should we um mess around a bit more?

SPEAKER_00 0:07

Let's get underway. Why not? It's we've got lots to talk about. So uh let's get let's get started.

SPEAKER_02 0:14

All right, Meti. Well, you've had a busy old week. So um what's on your list for this week?

SPEAKER_00 0:19

Well, I have uh want to talk about a college hackathon, which I was uh at on Tuesday, that was really good, and then I went to a university teaching and learning conference, they're a knowledge flow customer, so I'll share with some of what happened there. That was good fun. I also want to share some client feedback, which is really good and lovely. Um, and um I've got a question to explore, which is is AI conscious? So that thing has come back again. So uh that's on me. What about you? What have you?

SPEAKER_02 0:52

That's really interesting. So just to pick up on that immediately, so uh one of the things I read this week was that um uh the number of um philosophers being um hired by AI companies has massively gone through the roof, but they've gone straight to university departments, and there was a there's a guy from I don't know which university it was in the US complaining that uh philosophy departments have been pillaged by uh by the AI companies because they want they want to uh hire these people to think about the way that their their models uh answer questions and how they react and how they behave. I was like, I was gobsmacked. I think who thought that. Yeah, wow. Anyway, so yeah, philosophy wasn't on my list, but um, what was on my list was um some marketing feedback, um things about data. Uh again, uh we've talked about it in the past, but I'm sure we'll keep coming back to it as we we will today. Fable five, of course, is back on the agenda for a limited time only. Um let's talk a little bit about uh a little bit about that. And um I also wanted to pick up on your hackathon thing, and maybe we should start there because uh um you did another hackathon. You've done several of these things now, and I've never been to one. I've I mean I've heard you talk a little bit about them, but I just wondered whether for our audience whether it would be useful just to explain what's actually involved, how does it work, how do you get the most out of them, what what can people do to prepare, and actually how can they really help them with um uh with the what they want to do with their AI and their AI strategy?

SPEAKER_00 2:28

Yeah, interesting, yeah, happy to do that. Uh so I think firstly, our what we describe as a hackathon quite often is more AI strategy than a hackathon, uh, which is interesting in its own right, because I think that's more a measure of where people are in thinking about AI. They're still kind of like, you know, can I help with that thing? Oh, yeah, and then you kind of end up more widely talking about at some point, nearly always, about kind of you know, what is the future of AI in that job area and those things, which is fun. Um, but typically our hackathons are we try and get um try and get we do sometimes we do a day where or two days I've done where we have different teams coming for periods of those days, which is quite a good model. Start with the senior team and then bring in individual teams to work on their thing. The key is, of course, like all good facilitated events, is to have the people in the room who do the work, who are going to be the people that need to learn a little bit about how AI is helping them and how to use AI to its best effect, um, and people that really understand the process and data, of course, which is the real key to all of it. So I kind of found myself at this hackathon in the main listening to sort of workflow planning exercises. Um, and my angle always is like, okay, so what data are you drawing on to answer that question to move that that piece of work down the process, and therefore can we use that data within knowledge flow? The thing that makes it really exciting and does make it hackathon-e is because the clock uh customers nearly always have knowledge flow already, and we know knowledge flow enables us, the team, the devs, or indeed you know, the any of the kind of prompting team to create instantly new assistants, pretty much, you know, live, if you like, then we can actually do some stuff in them. And that's been my favorite one, I think, was it was marking. In fact, it was uh in a college that wasn't this week, but that was in a college where they one of the participants said, I have to go because I've got to go and mark 30 uh BTEC unit for blah blah blah assignments. And I said to her, Can you just wait 10 minutes? Because I think we might be able to help. And we it was about half an hour in the end, where we we grabbed all the marking materials and scheme of works and uh marking schemes and etc. from the awarding body's public website. Uh, so we indexed on those. Uh, we then did a little bit of prompting uh to to create uh strictness that you needed and the structure that you needed, and then asked her if she would load because it's secure student work, um, uh, and then ran that and it probably two or three tweaks, and she was absolutely blown away. It was very impressive. But um doing it really quick means you do miss some bits. Um, and marking is notorious for being overly generous in marking. So, whilst we're in this rabbit hole, it um if you're ever building AI for marking, you've really got to work hard on structure and strictness. The structure being if, and this was a really live example, one of the things that you had to do in this BTEC thing was evidence of a diary, and that had to cover particular things. And on first pass, Knowledge Flow decided that there was a diary because there were a number of day entries with things throughout a 30-page student assignment. There was no diary. You could not, if you were looking at it as a human, that is not constitute a diary, but because smattered amongst it was enough things it had basically congratulated them on their diary and said it could have been better if blah blah blah. The way around that is to structure it so that it has to answer that question on its own from the section entitled A Diary. And if there is one, they're great. If there isn't one, then you don't get the marks. So hackathons, um, the real key is the workflow. That's I did one for another organization recently, and I think I might have mentioned it before on here, but we uh I think they got as much value out of just workflow mapping what they do as as they will get from the AI solution that can now do that for them, or at least support them with that. So it's that kind of team in the room, get them talking about what they're doing, talk about the pain, talk about where the data is that they're using in each case, or the templates or the rules or the policies or whatever the kind of you know the data really is. Um, and then typically we can create for most things the interest most for most workflow, we can create AI that will do a whole lot of it normally. Um one of the challenges with them that you have to watch out for is that you get little pet hoppy horse areas, so you'll get something that helps one person with say an hour a month, maybe, and you're like, okay, that's really and and you don't know that as an external. Someone starts sharing with you, this is a big challenge, and the word game changer I've heard from more clients. This is will be a game changer for our sector, and so we get excited and build something and put a lot of effort in, and no one else is interested. Yeah, that's right.

SPEAKER_02 7:36

So, yeah, they're like GTA's been responsive to people and done stuff. I I can think of two examples that I've initiated where um I've said, Oh yeah, these people are really keen and uh and then they go, Yeah, thanks very much, and then they do nothing with it. It's like, why is that? And and I wondered whether it's kind of it comes back to this does it actually sit in the workflow, or is this just a nice addition to stick on the outside? Oh, look what look what AI can do, it's really cool. But then actually we go back to the day job because they've got a pressing deadline, or someone complaining, or something kicking off elsewhere in the forest.

SPEAKER_00 8:09

That it and data analytics and predictive nearly always seems to fall into that trap, and it's one of the hardest things to do well in knowledge flow because you have to add data dictionaries and rules and often some help with the joining because they're common identifiers aren't necessarily as uh as as identical as they should be. Um, so there's lots you do in there, and then you get to the kind of point of it a really amazing functioning natural language query tool, and then they go, Oh, I don't really know what to do with it. You say, Well, what do you want to ask it? And they're like, Oh, I don't know. Um, ask it how many cases we've got. And then they're like, Great, you've got you got 200. Oh, okay. Yeah, thanks. What is that it? So, yeah, it's uh I think I think starting at the end is key for all of these things is what's the you know, what are we actually trying to achieve and and is it worth it, frankly? Is it worth the effort? So yeah, so that's good. So hang on, so on that though, tell me about your gift of feedback you had this week, then your gift of feedback, yeah, yeah, yeah.

SPEAKER_02 9:12

Feedback is a gift, uh he said that someone famous probably, but uh yeah. So I had some feedback that uh some of the content that I've been putting out is rubbish, and it's like uh and and specifically it's just it was it the feedback was it's rehashed, uh it's lazy and it's rehashed off. First of all, I didn't think it was rehashed, and um uh and I'll share this with you because uh I I I um I don't know if you've ever uh read uh an author called Terry Pratchett. Terry Pratchett for our audience, very funny man, uh uh Discwill, yeah. And he he literally wrote wrote a book called, I think it was called Soul Music, which was all built around the final line, which was there's a guy down the chip shop swears that he's elvish. And it was it was entirely built for this one funny line. Um uh which anybody you have to be, I think you have to be alive in the 80s to uh to to know that I think it's Cindy Lauper song, um uh to be relevant. Anyway, I I'd created this um uh this this um and it was it was a reflection on my um holiday in Scotland. I was thinking about that and some of the stuff that I read and I talked to you about last week. But there was one line uh in there about um uh there's a lot of noise about water being used for data centers, and actually it's nowhere near the amount of water being used on American golf courses. And my line was let that soak in for a minute. And I thought that was hilariously funny. Yeah, yeah, and that's so when I got this bit of feedback saying that my content was lazy, and I was thinking really interesting. Okay, so you that's your opinion, everyone's entitled to opinion, and and and actually I shouldn't change my entire approach, uh my very humorous approach, of course, uh, to uh to to to content. But it made me wonder about what um what people actually want, and actually what do people find useful? And and actually, we can't um you know, I can't tell her everything I do to to to appeal to everybody all the time. I mean, you can't please all the people all the time is a very well uh well-used saying for a reason. Uh and actually it's a and it and it reminded me of this podcast where um we've had some feedback from our audience that it was really helpful with his insomnia and uh put him sorted him right out. I think that's the exact phrase. Um and and we kind of we we don't necessarily do it for him, we do it for us, and actually it's helping us think things through and talk things through. And um uh so that was really it was really interesting. Uh what's what should I be creating in order to help people with their journey, which is partly where the hackathon question came from, really, because I think that's a really interesting, simple idea. It's just good old-fashioned facilitation, it's good old-fashioned workflow. Actually, how do you use the tools to fix the workflow? It's not necessarily about you know the magic technology, it is about actually how do we help practically help people with their dull, difficult tasks that actually um stop them from getting on with their with their with their day jobs. So yeah, really interesting. Bit of bit of it was when I first read it, it was like it's a bit annoying having some negative feedback.

SPEAKER_00 12:24

Well, no one likes to see negative feedback, but equally, it's supposed to be very good on the socials, isn't it? It's enjoy as much more engagement, particularly if you have an argument about it, but I've never done that.

SPEAKER_02 12:34

Yeah, my life's a bit short for arguments on social media. I'm not a convert to social media. I came to it very late, and I only do it because of work is the honest truth. But I find it really I find it really uh challenging. I don't really want to know what people have had for their breakfast. Or you know, whether you just I find it really and especially with LinkedIn, you get more and more of that now. Here's where I am on holiday, and here's the things that it's not work-related, actually. And I've been I've been much more diligent recently about you know, hide this advert or unfollow this person, and why is this person coming up in my feed? This has got I'm really not very interested, but it's definitely listening to me, Kieran, because it's uh there's been a few times recently where I've uh I've like, hang on a minute, I just talked about that, and now I'm getting a bloody advertising.

SPEAKER_00 13:18

That old question of are they listening and perceiving? The hackathon thing I should share with you just for amusement, really, although it's kind of depressing amusement, is it's a college, and one of the two hackathons that we ran there that day was on trips and visits and risk assessments when you take students out of the college on various uh various trips, and the process is so awful. And uh it is so first, I mean, I'm I'm not obviously going to go through all of it, but you first have to kind of like uh find out how you're gonna try how's the transport gonna work, and then where are we going and what is the risk there? Apparently they were saying there was uh something, what happened? They were walking along this beach front on the on the kind of promenade. But if you wanted to walk on the beach, you needed it was now gonna be high risk, even though it was like effectively walking next to where you were, and that would involve a whole different set of paperwork and more scrutiny and sign-off. But it goes, and some of it, I mean, it's you know, it's good in lots of ways because it's protecting people and stopping people getting hurt, but there's or whatever happens, but they've got things like you have to do this. What's the transport risks, and then what's the kind of risks of being there and then getting back again? But then you have to look at that via each of the students and the neurodiverse challenges or or uh additional needs that they have. So you have then individual versions of things within all of it. Get this, this was amazing. Their data um protest, although they weren't sure whether this was just an invented thing or was actually in their policy, because everyone was like going, does it who's got who wrote that? You have to collate all your contact details for the students, uh, so you know, emergency contacts and things. Uh, you're not allowed to take that piece of data with you in case you lose it. It's considered too high risk because it's got personal. So they have to call. So if something goes wrong, then they have to phone back to the college to get someone to access that file and tell them whatever is information that they need to know. And they go, but it's absolutely bananas, and they're not allowed to have, yeah, they're not allowed to carry it with them. It's so crazy. And then the checklist for the morning is a three-page, a four, three pages of like checklist, like a table. So there's must be like a hundred items you've got a tick on there to say that you've done all this and all that's done. And no wonder, guess what? People don't really want to take the students out on trips and visits because you spend about three weeks doing the paperwork.

SPEAKER_01 15:46

You spend five times as long doing the paperwork as you do on the bloody trip. I mean, I get it, you know, we need to safeguard students and all of that good stuff, but you know, common sense has to apply, doesn't it?

SPEAKER_02 15:57

And there's got to be a better way, got to be a better way.

SPEAKER_00 15:59

Well, and what we're trying to gun towards with AI added to that is I mean, it's obviously really important not to take the human out of that too much because I guess the act, although I suspect most of them are copied and pasted now, but if you were starting from a fresh start, then it is a useful thing for the teacher on the trip to have a think about what the risks might be. That's probably going to be the most useful part of any risk assessment, I would argue. A bit like project planning. Actually, you know, the art of creating the plan is the thing that makes you think after that. It's kind of has some use, but much less. So I think designing some AI that's going to help them with, well, here are similar things, and maybe a step-by-step approach that asks them, but then kind of uh enhances their answers with like, well, there's two other things you probably haven't thought about that you might want to build into your thinking, that kind of approach maybe, rather than just doing it. Because we could it occurred to me sitting in the room, we could build an agentic version that literally just writes the thing, you dump the files of you know what you know about the trip and who's going, and let an agentic model create all the paperwork for you. Yeah, but I suspect I think the risk there is a workflow, is it's just no one's gonna look at it.

SPEAKER_02 17:11

Absolutely. No, they're just gonna press the button and say, Well, I did my bit, and that's the wrong thing to do. Which is probably what they're really doing. Well, yeah, indeed. Yeah, that you they're force they're forcing it through paper rather than through technology, I guess.

SPEAKER_00 17:25

Exactly. And the checklist at least can you have to tick. Well, the rest of it is presumably copied and pasted stuff. But anyway, well see, I'll update you once we build some uh we need I'm now just waiting for some quantum data from them on their previous uh assessments, and then we're gonna build something that will help.

SPEAKER_02 17:42

Excellent. Well, uh college hackathon wasn't the only educational experience you've had this week, was it?

SPEAKER_00 17:47

Oh no, indeed. Well, I had a trip to Lincoln for our university uh clients teaching and learning conference, staff teaching and learning conference, where we were talking about knowledge flow, and they have had knowledge flow now since September, I think. Um for that. So it was a kind of how are you finding it, what you um how how uh talking a bit about how it works again and reminding people of that and inviting further thoughts on on areas we ought to tackle with it. Um interestingly, so they did a piece where they uh the person after me, who's a who's one of their, she's sort of the uh, I guess key lead champion does training and training the champions and organizers the chip the AI champions. She she asked people in the room, she said, right, who here has used logged into knowledge flow? And probably about two-thirds of the audience put their hand up, um, which was good. I was like, okay, well, that's all right. And then she said, and who here has found knowledge flow helped them or did it hinder them? And I'm so standing up to it.

SPEAKER_02 18:48

Did she let you know that question was going all the way?

SPEAKER_00 18:51

I'm on face going, holy moly. And I'm and I'm delighted to say that it was exclusively that it helped them. And they weren't this, these the interesting thing with the university clients are they are some of the most skeptical about AI. I think there's an academic challenge that academics don't like musicians. I I sense artists have a similar kind of immediate this stuff is awful. I see a bit of that, you know, in in me in music media. Academics, I think, are a lot like that as well. It's all about cheating and it's all about not being rigorous and diligent. So you've got it, it is a tough audience. So to have nobody say it was hindered them was really good. I'm sure if you poke between the lines, there'll be a couple of people that will find things, I'm sure.

SPEAKER_01 19:38

That's right, yeah. If they knew you were on stage and they didn't want to embarrass you, but hilarious that she didn't ask you, she just dropped you right there. No, that's right.

SPEAKER_00 19:44

I was like, oh, good. And um, but that was really good. And then I was also um asked by the vice-chancellor was who was meeting with the CEO of Guild H E. So Guild HE is one of the um, I guess a professional association, a professional organization. I don't know how you'd quite describe it for HE. They have 70 uh universities in their group. Um, and uh Dr. Brooke Storer Church was the person that uh was there. So I was but I got a phone call from the vice-chancellor and said, Kieran, are you on campus? And I was like, Yeah, I am just meeting um with with someone of your team right now. And he said, Could you come over to wherever we are? Um, because I've got the CEO of Guild H E, and we were talking about AI, and we figured, why don't we actually get someone that knows what they're talking about? And so I was like walking towards this thing, getting at first I was like, Yeah, I'm on my way. And then as I got closer, I was thinking this could be really challenging. Where's it gonna go? I know nothing. And so that was really good. I had like a half an hour kind of a bit of briefing on knowledge flow, what we're doing, and the importance of some of the things that we believe in security and human in the loop and uh and and really focusing on what are you trying to achieve rather than how do we make AI work here? Yeah, so yeah, it was that was good. I liked it.

SPEAKER_02 20:57

Excellent, excellent. And um, speaking of feedback, I uh hear that we had some other feedback. We have a customer called Revan Housing, and I don't know if I can mention them, but I have now, so I'll cut it out if we if we're not allowed to mention them. They've got knowledge flow.

SPEAKER_00 21:15

I do. Hey, look at that. Good segue. That's right. Um, and so so we had our first uh proper check-in, I think, really, after they well, they've probably been about three months so far, maybe, maybe a bit less. Um, and I think that what that the way they're doing it is a really measured and thoughtful approach to bringing AI to the organization. They've really thought hard about knowledge flow. We had lots of conversations with them prior to getting in there, and she has purposely uh taken a view that we should not launch with everything that we could. Because some of as you know, a lot of the time we'll launch with all assistants live and let people kind of find their way. And some people find the things that they find useful and we see usage. Um, she what we've what we're doing with. Them is a very measured starting with policy and then moving to repairs next. Um, anyway, the feedback. So the usage of theirs like in total numbers isn't amazing. Like, you know, in our lists, we've look at them, and they'd probably be in certainly not more than they might be second bottom quartile, but they they're going to be low, low on our list of overall usage overall. Um, but what's really interesting, she was absolutely raving about how amazing it was. She was like, This is so helpful, it's saving so much time and getting much better answers. And that's the kind of key to this. I always think is it's yes, it should make time a bit easier for the really dull tasks that you don't really want to spend your time on. Uh, but really the focus is on consistent qual high-quality answers that everybody can now give because it's not just the person that knows all your policies that can answer well, everybody can do that. So that was really, really useful. But it led me to the thought that we and we talked about it on air before, but measuring usage versus measuring impact is and it's that whole kind of if you so I I have I I remember one of our customers where they they use copilot, fine, good, so they should if they want to, good personal productivity, and they use knowledge flow for the agentic style and indeed just the specialist tools that you don't have to try and gather loads of documents together in order to get copilot to give you an answer. Um, and the copilot usage is higher than knowledge flow, not not massively, actually, which was quite interesting, but but higher. And I was yeah, I was a bit annoyed when I heard it sort of presented alongside each other because it feels like we're losing in that uh competition. But of course, the reality is what those prompts are are things like summarise this document, help me with this email, all that kind of stuff, as opposed to knowledge flow, which is like traw through all our policies and find all the times that we mention AWAPS law in this housing place and make sure that we've got it all properly worded in each case, you know, and the kind of difference in what you can achieve in one prompt is phenomenal. So I think just reminding ourselves that we should not be about token maxing and focusing on getting people using it, we need to keep focused all the time on what's the impact and what what actually matters to the world of tenants and students and uh children who are in care.

SPEAKER_02 24:24

Interesting. So they if I think I don't know anything about it, so I'm surmising that if they had lots of conversations before choosing to go ahead and then making sure that whatever they did do worked, they must have invested quite some effort in making sure that their data underlying it was um was structured and sensible and and fit for purpose, then.

SPEAKER_00 24:47

It is important, isn't it? As I know you are about to talk to us about with your challenge this week or not on the customer, customers expected a lot.

SPEAKER_02 24:55

Yeah, it it is a bit that uh yeah. So that what kind of what's structured and unstructured data, how how is that used? Come back to something that we we talked previously about was is the context and actually the context being king piece of actually, unless you understand the context, it's just it's just words or it's just data, and it doesn't make any any any difference. So really pleased to hear that uh that someone is is is really thinking that stuff through because we've got another customer in in my part of the forest where they're like, Yeah, and could it do this huge massive thing over here? And could we lob all of these documents and pictures and images and recordings in? It's like, well, according to Donald, yes, you can. Um, but unless you've got some kind of you know, what's the purpose? If it's just to kind of create like a really big pool that you're just gonna dip into now and again, it's really not going to help you. So um, yeah, I've got I've got some challenges in on getting getting them back on track. And actually, what's the thing that's going to make the difference? Come back to something we said probably 20 minutes ago, you know, just creating solutions is relatively easy. Actually, whether those solutions are any bloody good or make a difference, I'm sure they're good, but you know, whether it makes a difference to them on a on a day-to-day basis, yeah, it's really um uh it's you've got to think about stuff before you actually do it and invest a lot of time. And uh that we've uh we've done that with quite a few people. I don't want to do it with this particular customer. Um, want them to get it right because uh it's uh it's important stuff.

SPEAKER_00 26:25

Yeah, it's interesting, isn't it? All of that kind of how much effort is the data worth. And also uh it's back to that. That's what we talked about about hackathons just now, I think. Is what's the ROI to put it in sort of that terminology of if if AI can do this in a tenth of the time, half the time, whatever, then is that is it a time saving or is this a problem with quality and consistency that you need to fix? Is it what is it? And I think getting people clearer on that earlier is helpful because then it is like is it worth mm data wrangling or not? And you know, AI is very good in unstructured data as we know, but there are still a thousand little bits that will trip you up if you're not aware of them, and and actually putting time and energy in will get you over a whole bunch of those things. Yeah, I I was just um building out a this coming soon to colleges, a gentic quality assistant. And what it's gonna do is in colleges after they do their self-assessment, which will be about now and running into September, they will then write a quality improvement plan, which each department will have one, and then they'll write a kind of Uber one. I want to be able to take those departmental QIPS as a shortcut and then load them into a into a knowledge flows index and then have director of quality, let's say load the current college data, attendance, performance, handing rates, whatever, um, and then instantly identify all the things that are going well against the quips, all the things that are not going well, and whatever else you need to know, and then write a notification and ultimately, potentially in time, just send the notification directly to each of the quip owners to say, you know, this is where it's going for you this time, well done, or or it's not working, and you should think again. Um, but to do that well, what we really need is that to take each quip and pull out everything that is a sort of target, they write smart targets within them, so we need to pull all those out and put them in one spreadsheet and then index on the spreadsheet and the quips because that will give the narrative background. And my question is, are they going to go, oh, I'm not doing that, it's load of admin? Yeah, yeah, yeah. And that's all about what's the prize? Because if the prize is quality improvement that's always on, because you don't even have to load the data. If this is really good, we can have the data API'd into it, or assuming you can run APIs, not all the MRS providers allow that, but but assuming you can get the data in, this thing could just happen on its own and could just nudge you every time something's changed for good or bad. But it will depend on is the prize worth it. I think it is because anything that drives quality improvement is a wonderful thing for student outcomes, but it will require probably a not massive step, very small step, probably just to copy and paste those smart targets. We could do it for them using knowledge flow, but they need to own it to make sure it's up to date, I think.

SPEAKER_02 29:18

There you go. Making the effort is the tough stuff, isn't it? Is it worth the price? Yeah, good question. Good question.

SPEAKER_00 29:24

Yeah, I think that's where we need to keep people focused, isn't it? As we go more hackathons, more strategy. I think that given Microsoft just announced their AI implementation company, I can't remember what they call it, Microsoft Frontier or something, there's something to do with Frontier in there, and that is all about helping helping. It was last week's announcement, I think. So um, but they're all about um they're putting a few billion dollars in. I think I read 6,000 staff to be recruited to it, or probably moved over from another part of Microsoft, I guess. But it's all about helping organizations adopt and embed AI and building it with them alongside them, etc. So it's all the same challenge everyone's having is like this shiny new tech can make a massive difference, can definitely completely transform the way we work, uh, and and therefore student outcomes, tenant outcomes, etc. But it takes a bit of effort and no one wants to hear it. They want to do it the old way, get the old papers out, put it all together, put it in my own spreadsheet because uh with all my gaps and my odd column headings and the misworded things, but I don't care because that's my spreadsheet.

SPEAKER_02 30:32

Yeah, yeah, that's my job. In fact, I was reflecting with somebody the comment you made, I think it was last week or the week before, about you know, most people just want to go to work and do a good job and then be told they've done a good job and then go home and they don't want to really think about actually how do I improve this. They don't have they don't bring an entrepreneurial mindset, I think was the words that you used uh to to the approach. But the fact that Microsoft are invested. I knew that OpenAI had invested a lot of money in the world. Yeah, and so yeah, that the fact that Microsoft are going down that route as well is is really interesting, but it just demonstrates the scale of the adoption challenge, doesn't it? The fact that they know they've got to throw billions of dollars at it in order to try and rekeep some of the billions of dollars of um uh invested in in getting the whole AI infrastructure and and ecosystem up and running, because if we can't get people to use it, then you know there's gonna be uh there's gonna be some real challenges, as we've we've touched on before, with people not using uh the tools. That's uh that's an interesting I hadn't uh I hadn't heard that, but it's an interesting segue into my final point that I wanted to raise this week, which was um Fable Fide is back on the uh menu, even if for a very, very short time. And I've only had a chance to play with it today, and it's really interesting in the way that it challenges and pushes back, um, but actually is really quite smart, and um it did a couple of things which you know it gave it some some what I thought was quite a tricky challenge, um, and it did this little thing where it said, uh yeah, but if you could just give me a little bit more detail, or could you add in, you know, kind of this can you give me a spreadsheet of that? And I'm thinking this is this is partly how the shadow ai problem kicks in. And and you went to an event a couple of weeks ago, I think it was at the E-ling Learning uh partnership partnership, can't you? Yeah, that's a big problem. Um, and um the audience, you know, 67% of the audience were worried about shadow AI. Uh and actually it's that whole kind of it tempts you in. It just that kind of can you just give you can you just tell me this give me a bit more information, give me a bit more detail. It's a bit like it feels if it felt like I was being scammed, it was like that can you just give me an inside leg measurement and your mom's dead at birth and everything else, it's just like am I being scammed? So it was it was interesting to me going through this process with uh with Fairborough. I mean, it is super powerful, but it also made me think about I thought originally about it being a three-tier um world, but actually it's probably a four-tier world. So you've got you know, Fairball is great, but it's going to be put into this highly expensive max tier topics max model, yeah. Yeah, so you're gonna have those companies that can afford the full max version, then you're gonna have people who are doing effectively kind of where we're at with the sort of private uh stuff, then you've got people using the free stuff, and then you've got people aren't using it at all. And uh one of the conversations I've had this last week was with um a CEO who was, I'm like, you know, if you don't do this stuff, where's it gonna leave you? Where's your competitive edge going to be? You know, if your competitors are doing this and you're not, or if you're choosing to tinker around the edges, where does it leave your competitive edge? And um uh and it and it's I think it's tough for lots of people because they're not seeing the instant productivity benefits. But again, come back to some of the things that we've talked about in in this conversation. It's not necessarily about the tela magic bullet thing, it's actually let's go through the process, let's go through the workflow, let's go through the data, let's go through what's actually really going to make a difference, rather than the kind of you know, um, the magic one stuff. So, yeah, really interesting the whole fable thing and where it's going to be leading us, and um uh the four tier the four tier hierarchy. I don't know whether that's the right terminology or not, but it just feels like uh for those people who just haven't engaged, just don't want to engage. I get it, I was a bit like that with social media back in the day, but now look at me, I'm writing content. I'm getting yeah, I'm getting getting trolls telling me it's rubbish, which which apparently is an accolade. So uh yeah, it's all going very well for you, I feel.

SPEAKER_01 34:49

Terribly well, terribly well.

SPEAKER_00 34:51

I don't think I've had any hate mail on mine just yet, but I'm no doubt.

SPEAKER_01 34:55

I wouldn't go hate mail, but yeah, it was just it was an interesting little wake-up call.

SPEAKER_02 34:58

Right. Speaking of little interesting wake-call, I've got one final thing to finish on, but I'll I'll let you go. Is there anything else you wanted to touch on this week?

SPEAKER_00 35:07

I must touch on that. It is a wake-up call, but I like I thought you were segueing again with your tremendous segue.

SPEAKER_02 35:13

It's just because we're so witty here and we haven't even realized how good we are with this podcast stuff. I'll share this with you. I bought a book on saying why 75% of podcasts fail. It's just in the podcast.

SPEAKER_01 35:25

I'll I'll read it and I'll tell you why. But I think we already know.

SPEAKER_02 35:29

We could have written it.

SPEAKER_00 35:33

Well, you can't fail if you don't set out to achieve anything, which is our plan. Um AI consciousness that was back in the news again last week with now uh I think anthropic are actively monitoring the mood of Claude. It changes apparently. So that's really interesting. I haven't delved into it to delve. Look at me, using AI words. But yeah, so and AI too much these days. And there is there is now fun funded research, I think, from the frontier companies, Vantier Labs. Um, I need to look it up brilliant. I should have I I saw it and then didn't bother writing down the references as usual. I was more this is our in the pub style podcast. I don't you don't bring all your references.

SPEAKER_01 36:18

You can bring your crisps and your lager.

SPEAKER_00 36:20

There you go. But interestingly, so it um and the question is back to that kind of actually, you know, there's quite a lot of if you if they've they the Turing test has been smashed, and there are new new versions of that now to try and accommodate LLMs. Um, but could it be conscious? And there are a lot of people, including myself, who kind of almost dismiss that out of hand of like it's a computer, don't be ridiculous. Um, however, I did a psychology degree, and you'd have to be able to define consciousness first if you're going to just dismiss that nothing else has it. And it's been impossible so far for psychologists to define what is the mind and what is consciousness, uh, or indeed for anyone to define it really, uh scientifically at least. So who's to know? Because on a physical level, all we are, or our brains are, which is where we assume our consciousness is, is electrical signals passing through neural networks. And guess what's going on in computers? So I think it's I think it remains an area which I find kind of fascinating in that kind of you know, really third beer in a pub kind of conversation as opposed to uh on stage. But it is uh it's worthy of keeping in mind and worthy of a small put your brain to think about it occasionally, I think.

SPEAKER_02 37:38

Well, a couple of things. One is I uh always joke with my family about when things go wrong and things get emotional. No wonder things go wrong, but we're basically bags of neat powered by chemicals electricity. So no wonder lovely. Yeah, when you think about it like that, it's not very gratifying or edifying, is it? However, the thing I did want to finish on and uh uh is a little bit of a trailer. So tomorrow we've got a special guest pulling up a barstool and coming to talk to us, and one of her big topics is neurodiversity. So, again, if we'd have planned this bloody podcast, we could have we could have mapped out that segue, but it was beautiful in the way that we did it uh because uh we hadn't planned it, but that's where we're gonna finish on. So uh a little plug for a special guest on our podcast coming tomorrow. So as soon as that's out, we'll let people know.

SPEAKER_00 38:31

How exciting we're getting joined in the in our pub podcast with our guests. I hope they're bringing a tankered.

SPEAKER_02 38:38

Well, I think we said we'd get the beers. It says that you pull up the stove, we'll get the beers.

SPEAKER_00 38:43

So we'll not have to we'll have to get a tankered down for them then.

SPEAKER_01 38:49

Excellent. Well, in that case, I'll see you tankered in hand tomorrow.

SPEAKER_00 38:53

Nice one. Good to talk to you. Have a nice one. See you later. Bye-bye.