Transcript
Well, good morning, everyone. I have with me once again, the great pleasure of speaking to Finba Joy. And Finba, you know, I’ve spoken about technology, about a lot of different things for a, yeah, on a number of occasions. And today our topic really is about Busting a few myths around AI and why everyone is getting so excited, because actually it just should be part of how we do business. Absolutely, yeah. I think there’s kind of too much drama on both sides of the debate really. Yeah, you know, the people who believe we’re just, we’re just about to enter the age of Skynet and, and these things are going to take over.
And I think the other extreme people who just, uh, completely dismiss that, uh, that, yeah, absolutely. So, you know, it’s something we have seen before with a number of, of different trends. And uh I I think there is, there is undoubtedly a reason to take a good look at this technology and what it can do. It does make a dramatic difference to an awful lot of the things you would typically pursue in any corporate enterprise IT environment. And tell me very quickly, where are you at exactly?
If I have to ask you in one sentence, are you, do you think the, the, you know, AI is gonna take over our lives and we’re just gonna be, I was listening to a podcast actually from someone at Silicon Valley University, you know, the X Prize the other day, and they were, this one of the people there was on the spectrum of in future our friends will be AI. relationships with anybody. I mean they positioned it and I, I was like, I’m not so sure how positive that is right through to, you know, the other person who was there, it’s all negative and so yeah, so where are you at?
Uh, yeah, I think undoubtedly it will have a massive impact on our lives and and we’re already seeing that. So, um, you know, I am, as a technologist, unsurprisingly somewhat in the positive camp, yeah. Yeah. Um, you know, I do see the the the overall benefit, and I do believe it will have a greater and greater impact on our, on our lives and work. Whether we’ll, whether we’ll within even our lifetimes see. This this whole general AI capability, I’m not sure, I’m not, I’m not convinced of that.
And the generative track is a different track, you know, I I you know, even it, um, and that, and that’s maybe some of the things that are badly, you know, uh, hyped that capability, we’re then gonna get general. The two things are completely unconnected it is completely, you know, it. I don’t think that that that that’s anywhere significant, but the capability of generative on its own is itself very significant, and I think the trajectory for generative is very, very interesting. But for people who don’t know, let’s let’s start right at the beginning.
Maybe it’s helpful, like you did ages for me. What’s the difference between digit digitalization. Maybe you can also take the time to explain a little bit what’s generative AI? What is, you know, maybe just explain a few concepts to us so that we because I think you’re right, everything up. Yeah, yeah, so um after some pioneering work by Google predominantly. Um, who released some models that became the basis for this kind of capability on top of neural nets. So most people have heard of the neural net and deep learning capability.
It’s been around for a good number of years, and it’s basically multiplied that capability. I don’t think people will know and The capability of the, the models used to process that data become more powerful. So that’s the, and they get called the large language models, that’s like a capability that that evolved on top of deep learning. The pioneering work was done by Google, and then very quickly organisations like Open AI started to pursue that work as well around, you know, about several years ago now. And of course the big breakthrough was a couple of years ago, about 18 months ago when uh Open Eye released Chat GPT and that kind of that’s kind of what created all of the hysteria if you like, because it seemed like suddenly we could do so much and it it looked like it had come out of nowhere, but it had been under development for a number of years and that you know that large.
Language capability is relatively, uh, I’d say relatively mature now, but it’s on a very, very impressive trajectory. I think that’s the important thing. What does language capability and why is it on a trajectory all about um this whole ability to uh give a request then based on the data that it’s ingested. For instance, in the, in the, in the case of Open AI is practically the entire web, so you know they’ve up all the web then you literally all it’s doing is predicting when you when you make this request, how the sequence of words typically forms in response to that.
So it’s learned, that’s literally all it’s doing is it’s not really understanding. It’s sequences and that, which is why there’s such, it is wrong to assume that this is the beginning of general, um, you know, artificial intelligence, generalised artificial intelligence. It’s not, um, it’s literally just pattern matching. Um, but it happens to do it in an extremely efficient way, and it happens to be based on an extraordinarily large data set, and that’s where the competition’s being fought out now amongst the big guys. They’re all trying to hoover up big, ever, ever larger data sets on larger computing processing facilities, and that’s been interesting that we’ve seen that over the last few weeks, even all of the big players.
Uh, the big tech companies are now investing astonishing amounts in data centres to do this processing, um, because that’s where the, you know, the battle will be fought if you like, uh, who’s got the biggest engine and who’s got the greatest data sets. Can we pause there for a second because I think you’ve just made a light go on for me because to be honest, once you start engaging with Chat GBT and there are other others now as well. Even though it’s just pattern matching, you start thinking, oh, this thing is really clever.
It’s understanding me, but actually, it’s just be matching. It is beguiling, and I don’t know if you’ve had a look at Claude. But they’ve, they’ve worked even harder on the humanistic side of response. And that really is very, very practically disturbing in terms of, you know, you could genuinely could have a counselling conversation with her and. But it’s just a pattern matching and, and, you know, which is why we’ve had to rethink what is the test for general artificial intelligence capability. It used to be, you know, the test around if you can discuss with it and you can’t spot it’s not a human, then it’s there.
Well, that’s true, but it’s, it’s, it’s still not general AI. All it’s doing is pattern matching. It’s not genuinely reasoning. Even though it might, it might appear as though it is, um, AI is, is that the process whereby it continuously patterns and generates its net new content. It’s not just that it gives you a response to a piece of analysis, which it can do. Uh, but given a set of prompts, it generates net new content, and that’s the interesting thing, and which is why we’re seeing, you know, I’m starting to see, you know, very, very impressive use cases around marketing and generating marketing content around product management.
Um, you know, that’s interesting. Yeah, absolutely. So you know, as that companion, if you like analyst synthesise your data if you’ve got a data set which is customer customer response surveys and a very efficient job of synthesising that that itself I think is a good example of, you know, several years. ago that would have been a very, very expensive exercise with data scientists, data engineers, a pretty significant kit. You can genuinely do that now through through a browser for crying out loud. Um, so, you know, um, I think there are, you know, it’s the rate of that commoditization of the capability that I think is the opportunity for enterprises to get on board.
So, let’s take that further then. So I think you said that’s the opportunity for enterprises to get on board. So let’s just strip all this hype away. Because I think what we need to do is just see it as a natural evolution and you know, we’ve all gone through digital transformation and so on and so forth, or spoken about it for many years. But this is just, it just should be viewed as a business tool. It’s just another tool we have now we’re talking about the business context, I’m not talking about other than that, but I’m just saying it’s just another tool.
It’s it’s very, very interesting um for uh somebody as ancient as I am, for instance, working in this industry. Yeah, the, the, the repeating patterns of what we saw through so many kind of digital waves of capability, starting to see all over again. So just like in the dot com boom, and then we had the crash and everybody said this is all nonsense, um, it was only, you know, a short 4 or 5 years after that that we had Facebook, YouTube, um. And we got, we had cloud capability and we haven’t looked back since.
But there was a period where it looked like this is all just nonsense and it’s not going to amount to anything, whereas now it’s completely abundant. Um, and we had those episodes a few years ago where, you know, British retail companies were asking the government to intervene because this Amazon thing was, was unfair competition. It literally just appeared dropped out of the sky when these companies could have been competing with them 20 years ago, but decided not to because they made their minds up that this world wasn’t for them.
Uh, so missed, missed that wave and you know, we’re genuinely starting to see some of that again, you know, a lot of, especially corporate enterprises, somewhat in denial. Yeah. But either it’s not significant enough or it’s too dangerous, therefore not really grasping the opportunity to, to plug this capability in, uh, you know. Um, that because of what you’ve said there, amongst other things, what one can do now is to leapfrog your digital transformation journey. So if you’re an organisation that’s not digitised or started with digitalization, so both topics.
How can AI help you to accelerate that instead of going through all the basics that you normally would have to go to, um, you know, the, the, you know, for, for, uh, I think one of the big challenges we saw with so many digital transformation efforts, and bear in mind that depending on whose stats you subscribe to, anything between 70 to 90% of transformations fail. So you’d transformation. Anyway, yeah, um, um, one of the issues, you know, or one of the most repeating issues is the organisation doesn’t actually know where its advantage lies.
OK, because every part of the organisation put it, puts its hands up to be more digital and create the case to be more digital, then the investment just get complete, gets completely diffused. You know, tens of millions. But actually there was no genuine advantage gained, you know, you basically took some process efficiency advantages, you know, you made a small difference to the bottom line, but you didn’t actually reinvent what your organisation does. And typically there’s no there’s no attempt. To find out what the winning player will be.
So how can we win? Well, you know, when we compete digitally, what will be the difference for our organisation, so that’s. So I’m gonna pause you there because you’re saying some really important things. So what you’re saying is a helpful, a more helpful lens to put on to questions and conversations. Around digitization or digitalization, because it might be either of those, is how can we win? And then we use that, right? And, you know, I think one of the opportunities with the, you know, the generative AI for me is about the extent to which it just commoditizes quite a significant element.
Of what used to be regular data analytics. So give, give us a practical example again for people who might not be in this world. So you know, one of, you know, that was, that was, that was always one of the first hurdles if you like, once an organisation gets underway, it’s OK, where’s our opportunity gonna be, you need to do some basic analytics. And that’s when then you get the whole kind of inconvenience or or difficulty which a lot of organisations have in mobilising all of that effectively.
Um, this toolkit, you know, can give us an opportunity to, to prototype that that effort, at least to prototype it. So, you know, you’ve now got a companion data analyst. So you have the ability to get together some basic raw data about either customer feedback or, you know, whatever it is you’re interested in, then you can get, you know, a, a, a good short term boost from what does this, what, what, what does this initial view give us. By using and partnering and you said your, your partner, you can literally do that through a browser interface, uh, you know, at least gives you that workshop capability if you like.
So you know, how do we evaluate this? Is it gonna take us a few weeks or months? Well, it could probably take a day or two. and, and if you think about it, that’s very, very powerful. But I think it not only could take a day or two or a few weeks or months, I think it’s you have to find the person who can do it. And that’s sometimes even harder. So, you know, that’s part of the issue, isn’t it? It is, but for me that’s the kind of power of of of how well they commoditize these toolkits is that.
You know, beyond, you know, some basic familiarisation, anybody can use it, you know, no, no, that’s true. I was in the past, before you have this toolkit or this partner, you have to find someone who could do it. Now you need to, you can do that over a browser. Yeah, exactly that. Now, at this point, there will be data engineers, data scientists screaming, uh, and saying, this, this man’s an idiot, um, completely overselling this, um. I think it is interesting, you know, we look at a basic data set, uh, and get some basic analytics done.
You know, it genuinely is a capability that a short while ago would have taken, you know, those bodies to be mobilised. So you still need those individuals to do the job thoroughly, but you can tolerate getting it underway, I think, especially for the boardroom conversations, it’s a great way of, you know, a board, an executive team testing itself, if you like, um, and that in itself being useful to get familiar with, to to to become familiar with the tool set, just like you are with Excel. So you know, every, like that analogy.
Would, you know, would insist, um, you know, would be very highly confident on working with their, their great old Excel tools. Well, this is another example of, you know, companion app that can make a big difference to your workflow in your decision making, um, and how you evaluate and test your decision making, um, to, to, to generate better decisions. So, you know, I think even just having that as part of the, you know, companion toolkit, I think makes a big difference. Yeah, etc. right? It’s something to work alongside with.
Yeah, and I think, you know, there was a piece of research we saw, I think it was it was a month or two ago. Um, And You know, basically the, the something like, you know, uh, you know, a large percentage of the Fortune 1000 companies that started a generative AI project had since ceased, uh, typically based on fears around what will happen to their data. Oh yes, which, which absolutely is an appropriate. Concern to have, um, but there’s a very, you know, straightforward ways of dealing with that, um, especially where organisations can use their own cloud environments to host the same technology.
So all of the big, uh, large language model vendors, um, have their, their engines available on, on the, on the large cloud providers, whether that be AWS or whether it be a 0. Um, so it’s possible to get the best of both worlds, you know, to use this new capability, but the data set is private to you, it’s not going to, to, to, for instance, to Open AI to train their model that other people can take advantage of. Um, so I think, you know, the reticence is, is, is, is reasonable, the concern to get it right, but I think it’s going too far to lock down all efforts and not, not.
Yeah, and it’s a solution. Yeah, exactly. So what do you think, what would your advice be to business, you know. Really, um, be, be appropriately curious, yeah. Um, so, you know, and take advantage of the commoditization. So for instance, it’s possible within a day of, of, you know, looking at what it could mean for your organisation and exploring the opportunities within a day to have a direction set, um, and, and that. I think makes a significant difference, and then again because of the availability of the toolkit and the capability of the toolkit, it’s typically possible within a week of that period to have a prototype available.
And I think it’s it’s absolutely astonishing, right, exactly, exactly, but that’s what we’ve seen time and again. And and and it really does relate to, you know, having the ability to work through what would be the biggest impact in your organisation. Um, and when you’ve looked at the scenarios, you’ve looked at the examples, you’ve explored the capabilities, then that’s perfectly possible to achieve within a day. And then similarly, given that options, you know, here are our two or three ideas, uh, because of the, you know, how, how well commoditized this has been, then within a week you’ve certainly got got a uh a prototype available, at least in the public versions of these things, if not in your own private version.
So, you know, you’re still got to be somewhat careful about what you build, but I think being able to get something like that underway within a week is pretty astonishing. Yeah, and I’m wondering, you’re making me think of something else there now, which is, I put up my hand and then I was like, oh, the thought is gone. But, 00, now I remember. Would you say this holds true for digital businesses, less digital businesses, because I, I can hear the kinds of businesses you and I both worked with say, oh no, we’re not a technology business, this is not for us.
And I absolutely think that’s not the case. Yeah, absolutely, and indeed, I would suggest it’s where those organisations can genuinely catch up. Yeah. So, um, you know, um, you know, by finding the processing efficiencies at least, if not using it more in such as product management and marketing, uh, to make a genuine difference at how well they fulfil those functions, um, you know, it really can accelerate their progress down the digital path, um, and takes at least a hurdle or team out of the way. Um, but you know, that, that involves, you know, that an executive of that kind of organisation being fully apprised of the capabilities being to what’s possible, um, and still will need some investment in technology, absolutely, so you know, it accelerates, um, and it raises the potential.
We still will need to do some investment in technology, yeah. But I think, I think what you’re saying it will, it could potentially be a more appropriate level of investment because potentially it can be more focused. So quick last question because I know time is running out. Um, can you just share with us some of the tools, everybody knows Chat GPT, but what are some of the other ones we can try out? Yeah, so Claude is um you know, very frequently referenced. Um, again, it’s somewhat more humanistic compared to, to, to, to chat GPT, um, but we’re seeing, excuse me, we’re seeing other interesting efforts coming from, uh, well, Meta have got that open source Llama 3 model, so we’re starting to see versions of that appear online.
Um, it’s not as easy to access as Chat GPT, but you know, oh, it’s certainly being made available in, in several places. Google’s Gemini, and yes, so, you know, I think right where at the very beginning, everybody assumed Google was just going to completely dominate this, um, you know, this race if you like, um, so you know, their capability is obviously significant, um, you know, given their search history. And then you’ve got um, you can’t rule out X, um, so yeah, so Grok um generated AI with attitude if you like, will be important to look at.
But the ones that people are doing, you know, I think the most significant work in now tends to be either, uh, chat GPT, Open AI, or Clawed from Anthropic. So Anthropic and Open AI are, are, are certainly the two that seem to be getting most frequently embedded. Um, but absolutely, you know, Gemini, I’m sure we’ll see a lot more from, um, lava. Um, I’m sure we’ll see a lot more from as well. So, Fin, with that, I have a load of other questions, but we always try and keep these short and sweet.
Any last words of wisdom from your side? I, I, I would say, you know, uh, be appropriately, uh, cautious, uh, but do reflect on the pattern that we saw at the, uh, in the early 2000s around the whole web evolution. And this time around, uh, for this wave, uh, genuinely get on board. I think too many corporates did miss out on that wave and then spent the last 5 or several years turning themselves inside out with digital transformation. I’d say this wave, um, catch it, um, take advantage of the commoditization and, and just look at the overall trend, um, which, which does seem clear to me in terms of direction.
Um, so why not take advantage. Um, you know, at, at the very least gain those process optimisations, if not reinvent your or your business. So what I’m taking away in conclusion is appropriately curious, appropriately cautious, and let’s stop a little bit of the hype, right? Yeah. Wonderful. So thank you as always, Finba. I look forward to speaking to you again, definitely more on this topic. Um, but apologies for all the background noises. I don’t yet have AI that can silence alarms and actually deal with emails that come through because I forgot to close my emails.
But thank you very much for your time. Thank you. Bye bye.
