We're witnessing history unfold right in front of our eyes with AI

Mark Smith: 0:19
Welcome to the MVP show. My intention is that you listen to the stories of these MVP guests and are inspired to become an MVP and bring value to the world through your skills. If you have not checked it out already, I do a YouTube series called how to Become an MVP. The link is in the show notes. With that, let's get on with the show. Today's guest is from Sheffield, england in the United Kingdom. He is a technical director at Flow Simulation. He was first awarded MVP in 2018. He has been working with computers for over 30 years and with AI since the 1990s. He regularly speaks at international conferences and user groups about Microsoft 365, ai and Azure. You can find links to his bio, social media, et cetera, in the show notes for this episode. Welcome to the show, Bill. Thank you, mark. Hi. Good to have you on. I always like having a little getting to know you segment at the very start, which is food, family and fun. What do they mean to you? What do you do when you're not doing your technical day job?

Bill Ayers: 1:30
What do I do when I'm not doing the technical? I don't think there is any time when I'm not doing the technical day job. It feels like uh well, we've got, uh, my wife and I we've got three children. They're all grown up and, uh, mostly left home, mostly um, and we're kind of now getting to spend a little bit of time doing a few things that we've always meant to do. So we're going to go um cycling in the netherlands, uh, next week, so I'm just hoping the weather is is nice for that.

Bill Ayers: 2:03
Uh, what were the other ones? Food, family and fun, fun I don't have time for any. Any fun, that's, that's not not, not possible. Uh, do a bit of, uh, yeah, uh, everything seems to revolve around technology. I don don't do any computer games. That's the one thing that I just avoid completely. It's just not my thing at all. But yeah, I spent a lot of time doing technology, as a lot of us as MVPs, do. You know that has become our hobby as well as work, but it was slightly different slant on it, I suppose. But food, well, I discovered that I really like Italian food and I really like French food and I really love Indian food and I really adore Asian food. Then I realized it's just food that I like. So just every the different countries from around the world. I love all the different cuisines and I'll just eat pretty much anything and sushi.

Mark Smith: 3:12
Yeah, very good, very good. I'm particularly interested in your AI journey. You know AI is obviously top of mind for a lot of people nowadays and you've been doing it since the 1990s. Tell me about how you know, what caught your attention in those early days and then bring me up to speed about you know. Since generative AI has come out and the impact that's having on the world, what are your thoughts on where things are at now, but also your journey to getting here?

Bill Ayers: 3:42
It's kind of funny. So as a child I remember seeing a talk given by Professor Sir I've forgotten his name now. Anyway, a famous professor gave a talk and said AI just isn't going to happen because it's exponential growth in complexity, and so it kind of put me off and I thought this was going to be the next big thing and I wanted to get into it and he persuaded me that this was not the future and it was never going to happen. It was sort of just technically impossible, uh, and he convinced everybody else as well, and so there was this complete decline in funding for ai and uh, and that kind of died out until the 1990s when it kind of came back again, but it was as expert systems, which is a slightly different technology. So it still counts as AI, but it's not neural networks, which was the first wave of AI, which started in the well late 50s before our time. But it came back in the 1990s in the form of these expert systems, which was a way of sort of capturing knowledge of people, and so if you wanted to have a, an expert system that would diagnose illness, you'd get a try and persuade a doctor to make a sort of decision tree of questions and it was kind of semi-programmatic. But they eventually came up with sort of software packages which would capture all this information rather than hard-coding it. And that also failed as a technology, because I think not everything lends itself to that approach, perhaps a bit of a reluctance of people who spent a lifetime trying to learn how to do something they don't want to just dump it into some expert system and have their jobs taken. So there's a kind of a bit of that as well, so that all kind of petered out.

Bill Ayers: 5:53
But I was involved in the expert systems. We were trying to do things for the time I was working in engineering software and we're trying to solve various engineering problems. That way we started getting quite good image analysis and speech, text-to-speech and speech-to-text and a certain amount of natural language processing got better and better to the point where about the middle of the last decade we got to the point where you could actually do some tagging of an image, say to a reasonable accuracy, and it got to the point where it was similar to human error levels. And that's the game changer, because now you can start in these admittedly very narrow fields to be able to do things like tag a load of images, which would be. You know, if you've got a million images in a SharePoint site or something like that, a document library, it would be very expensive to go through them and tag them, but you can get a machine to do it, and so that was a game changer. But then with the evolution of this transformer architecture particularly originally for text translation, so it was the encoder-decoder pattern for text translation that became so good and so efficient and the way it kind of finds its own level and you don't have to put quite as much effort into kind of designing the structure of the AI model. It kind of figures it out itself to some extent.

Bill Ayers: 7:47
I'm hugely oversimplifying the transformer, how transformers work, but it was such a breakthrough that the large language models which were trained on huge amounts of data the first one was I think the first one we were looking at was I think it was GPT-2 I remember using, and then GPT-3 came out and that really got the attention of people in the. That's GPT Generative Pre-trained Transformer really got the attention of people in the AI community, really got the attention of people in the AI community. And then GPT 3.5 was the version that became ChatGPT and suddenly it just exploded because people realized kids could do their homework, essays and they could answer questions and it was better than a search engine and things like that. And that's then improved even further with GPT-4. It's the GPT-4.0 now, which is the multimodal model and that's continued to develop as well.

Bill Ayers: 8:55
That can do things like you can talk to it and it can look at the text. So the old way was you'd you'd use a speech to text and then put the text into a large language model. But these multi-modal models can take the actual, the voice input and get additional signals from it, such as your tone of voice and things. And so now it's more like that film Her, wasn't it with, was it Joaquin Phoenix? It's just like that. You know, you can have a conversation with an agent. That's really like talking to a real person, so it's almost scarily realistic. So I don't know, Maybe in a few years' time these models will be indistinguishable from whether you're talking to a real person or not.

Mark Smith: 9:53
The gains seem a lot smaller. These days we seem to get a lot more models, and I think Sam Altman's just come out and said that if we don't get fair use on copyrighted material, we're kind of going groundbreaking turning point in the AI era. Is there the potential that we're going to see another breakthrough that will totally shift the paradigms again and, you know, it might not even be people that find it. Maybe the current levels of AI will be able to produce the next differentiator that's going to leapfrog us once again. What are your thoughts about this?

Bill Ayers: 10:45
This could go two ways. So we might have a big clampdown and people are saying, okay, copyright, so you could argue I think you can argue both cases. So it is kind of using other people's material, uh, but then isn't that what human beings do? Is we? We? We don't invent music from scratch. We listen to what's gone before and we are influenced by frank zapper or somebody, and then that that's our style, that we develop. So so that's that's just normal, for for the art world is to uh build on what's gone before and develop in little increments, uh, from it. So that might be a factor.

Bill Ayers: 11:31
Also, there's some people say, well, aren't we going to run out of training material?

Bill Ayers: 11:35
And eventually the ai is just, uh, the training data is actually just more AI output and the thing kind of just exhausts itself and reaches some sort of heat death of information or something.

Bill Ayers: 11:53
But another possibility is, which I think you alluded to, is, as we're able to use AI tools to improve the tools themselves, then kind of knowledge begets knowledge and it's able to then take off in an exponential way.

Bill Ayers: 12:15
And that's also a possibility and that has also got certain ethical and existential risk implications as well that if we could have a artificial super intelligence takeoff where they get it gets so much better than us, and then there's some awful accident and somehow somebody gives an AI model the wrong instruction and it goes and does something which has disastrous consequences, and there are a few uh a few ways that this could happen that have been discussed in uh, particularly um. Nick Bostrom has written a number of books about these various scenarios and how they could uh play out, so that's a depressing read, but we need to think about it when we're thinking about these models, and we have to be a little bit careful about what we do, both from an ethical point of view and as far as AI risks are concerned. We need to be aware of them and we need to think about it now, not in 20 years' time, when it's too late.

Mark Smith: 13:22
Yeah, will capitalism, though, prevent the actual? And we need to think about it now, not in 20 years' time when it's too late.

Bill Ayers: 13:26
Yeah, will capitalism though, prevent the actual level of thinking needed. I'm not sure capital has the power to stop this. If this happens, I think it's beyond any political system.

Mark Smith: 13:42
No, but that's what I'm saying is that the danger is that capitalism, the more money we can make off, what AI will generate is that the capitalists will keep throwing money and risk to. You know.

Bill Ayers: 13:53
I see what you're saying.

Mark Smith: 13:55
You know who cares what will happen. Let's just make us which is, I suppose, trump's probably mindset at the moment is that let's just go full steam ahead, let's innovate and let's worry about risk later yeah, I see what you're saying so well.

Bill Ayers: 14:11
I'm not sure it's capitalism that's the problem. I think it's human nature might be the the risk factor here, that people are acting in their own interests. So, as a capitalist, it's not in my interest to bring about the end of the world because I can't make money if the world comes to an end, so it's. But it's human nature to think. Well, I'm not worried about that risk because I can see this goal that will benefit me in the short term. And yeah, that's a danger. I don't think we're at this point yet, but I think we need to think about it.

Mark Smith: 14:48
Absolutely. We're talking about MVP Summit before and I attend a lot of the trustworthy AI sessions and just really understanding how we're thinking about this. You know, obviously the EU is quite often ahead in how they think about, you know, risks factor and particularly the potential on on the citizen base, and so I I find it interesting, even though I am, you know, barreling full steam ahead myself for my personal life and my business life around um ai, since you know, generative AI came to market and your work across Azure. You know M365 and in AI, what have kind of been the delightful moments you've come across in the last three to four years? That may be surprised or intrigue or good use cases. What jumps to mind?

Bill Ayers: 15:46
um I, I think, uh, seeing being able to save time. You know, it's like the industrial revolution you used to. A farmer would have to plow a field. If they're lucky, they had a horse pull the plow. Once you can get, you get a tractor or, you know, steam powered tractor or a steam-powered tractor or something like that. It means you can do 10 times as much work and achieve 10 times as much with the same effort. And I think it's the same, used correctly. We can do the same with A-high is. We can just amplify our brainpower and do more and avoid having to do the boring things. So it's really nice when you have some tedious job that you would have had to do and you can just click, click, click and when it comes together, it is really great that you can really save a lot of time. Now you might have a bit of a journey to get there, with prompt engineering and designing the right systems around it and maybe having some multi-agent system, which is the current thinking around some of the AI technologies. We kind of have different AI models to do different jobs, a bit like having a team GitHub.

Bill Ayers: 17:07
Copilot has been an absolute time saver for me as well. So, yes, I can make a. When I'm doing a presentation in PowerPoint, I can make an image. Danger is, you know they all look a bit like they've been generated by AI, but apart from that, you know that can save time doing that.

Bill Ayers: 17:28
But in GitHub Copilot, when coding, it's a real time saver because I can ask the model what the syntax is for whatever language I'm using. I might be using JavaScript this week and C Sharp the next week. I can't remember the syntax always from one week, or Python. I can't remember the syntax always from one week. Or Python. I can't remember the syntax, but I don't need to anymore because I can use GitHub Copilot and it can suggest things and quite often it'll just give me a block of code. I can just indicate what I want to do and it will give me a block of code and I would say about 50% of the time I'm done and I can move on the other 50 of the time. Yeah, okay, that's not what I wanted and I gotta work on it, but that's still already a huge time saver and it's getting better and better.

Mark Smith: 18:20
So, uh, really enjoying that as we go to wrap up, tell me about the benefits of being an mvp. Um, how, how has it been? How has it affected your career having, since you know, 2018 the designation MVP?

Bill Ayers: 18:39
Yeah, I think it might have been 2017 actually yeah, I can't remember. It's been really nice to have a greater degree of access to people within Microsoft who are working on the next versions of products, see what's coming, just to get a little bit of a heads up and really ask questions. So it just there's just a little bit more attention that you can get, little bit more attention that you can get and events like, as we mentioned, the mvp summit, where we can talk to, uh, the people who are actually doing the work at microsoft, not some second and third layer of of uh support people, but, but the people who are actually making decisions and and doing the work. But the other thing, the other for me, a big thing is the MVP community itself people like you, mark, that I then get to meet and talk with and have fascinating conversations with people in the able to socialize with and talk to and and uh people who are, like me, a little bit nerdy and uh want to talk about this kind of thing. I love it I love it.

Mark Smith: 20:00
You're so right. The community, I think, is for me, is what makes the mvp program um, and they're all over the world. You know, I've traveled a bit and I've been able to meet many of my MVP friends all around the world, so it is something special.

Bill Ayers: 20:17
Yeah, that's great, and different technologies as well. They're not necessarily working on the technologies I'm working in, so I get a different perspective.

Mark Smith: 20:25
Excellent. Thanks, Bill, for coming on the show.

Bill Ayers: 20:28
It's been a real pleasure.

Mark Smith: 20:29
Thanks, bill for coming on the show. It's been a real pleasure. Thanks, mark. Hey, thanks for listening. I'm your host business application MVP Mark Smith, otherwise known as the NZ365 guy. If you like the show and want to be a supporter, check out buymeacoffeecom forward slash NZ365 guy. Thanks again and see you next time. Thank you.