The Evolved Leadership Podcast

#21 Artificial Intelligence And Leadership In Today's World, with Ross Farrelly, Director of Data Science and Artificial Intelligence for IBM Asia Pacific

• David McDermott

Share episode

My guest is Ross Farrelly. Ross is the Director of Data Science and Artificial Intelligence for IBM Asia Pacific. He works with companies throughout the region to develop and execute on their strategies to adopt and realise the benefits of predictive analytics and AI. He has a Master of Applied Statistics, a Masters of Applied Ethics, a first class honours degree in Pure Mathematics and a PhD in Information Systems.


Some of the highlights of our conversation include exploring the ethical questions that AI provokes, the explosion of ChatGPT, how AI is helping the planet, and what organisational leaders need to be understanding about AI to lead more effectively in today's world.


Enjoy the conversation.   


You can reach out to Ross directly at: https://www.linkedin.com/in/rossfarrelly  


To learn more about what it takes to be an evolved leader, and to check out our other podcast episodes, go to:  https://www.evolvedstrategy.com.au 


Please note,  the views and opinions expressed in this podcast are the personal opinions of the speakers and do not necessarily reflect the views or positions of any entities they represent.

00:01.64
davidmcdermott
Ross welcome to the show.

00:01.71
Ross
Thanks David thanks very thanks for inviting me on.

00:07.92
davidmcdermott
Um, tell us about your leadership journey so far.

00:11.68
Ross
So I've um had a rather. How should we say or a slightly unusual career journey. Um I started off when I first left university I went and worked in england as a primary school teacher. And after that I actually was the deputy head of a primary school for many years and that was a very interesting learning experience in terms of working together with a large group of people, both students parents and teachers to grow a small school that had only about 20 or so students up to about 200. So ah, certainly learn a lot about working with diverse diverse stakeholders and also different personalities and I think that's probably been one of the most useful things I've learned about leadership over the years is no 2 people are the same. You can't treat you need to treat all people as individuals. Um. Subsequent to that I moved into it and I worked at a number of multinational it companies and I'm now working at Ibm. But and my most recent couple of roles have really been what I would call informal leadership in the sense that I'm not. Um I don't have any direct reports I'm not a manager but what I do do is put together informal groups of people to solve important problems and I think that's a difference sort of leadership because you're not these people don't formally report to you.

01:41.21
Ross
So You need to use your soft skills to encourage them and to convince them that working together with you and in in in these informal teams is is good for their career. It's good for the business and it's good for them and I think that's that's actually very common. These days is where. Informal groups of people are brought together to solve a problem and it's really it's It's really opt in it's voluntary and so you really need to sell individuals on the concept and get them bring them along on the journey.

02:17.38
davidmcdermott
Well I'm going to ask more about that in terms of your what you're doing now at Ibm but tell us a bit more about the ah you said you work for a number of multinational companies in the it space and I think 6 sigma was that one of the.

02:30.00
Ross
Um.

02:32.50
davidmcdermott
Sort of Frameworks that you worked with can you tell us a bit more about that sort of that part of your career.

02:35.20
Ross
Yes, yeah I was involved. Um sure so six sigma is a process improvement methodology. It's been around for quite a while. It's extremely It's it's based on statistics essentially and it's a very it. It was. Has origins in manufacturing and the whole idea was if you wanted to improve a process The first thing you had to do is stabilize it and then you needed to be able to have metrics around the number of defects being produced and so on and so forth. And then you could design and implement interventions which which would improve that so it's known as statistical process control. It's used right around the world particularly in manufacturing but nowadays it's also being implemented in service industries. Also often combined with the lean methodology. So lean. Six Sigmas are a commonly used term you may hear.

03:32.92
davidmcdermott
Right? And is that something that is relevant to your work now or have you left that that kind of you know 6 sigma lean methodology and work behind.

03:45.50
Ross
Um, so I specifically at the moment don't work on that methodology at the moment but I would say my my career has very much been one of evolution. So linsi sigma is where I started in terms of statistical consulting and so on. That then evolved for me into data science and as part of that evolution I continued to study university so while I was involved with lean 6 sigma. First of all I did the the lean sex segma blackbout courses. But also did a master's degree in ah in applied statistics and another one in applied ethics which is actually very relevant to the work I'm now doing with it Ai, but um, so yeah, the evolution was really through lean sex sigma evolving into data science which was sort of it was the hot. But buzzword at the time and I just happened to be in the right place at the right time with my skillset when I moved to another us-based multinational. Um, and then from there my my career is when I moved to Ibm, it's really evolved through cloud computing. And then into Ai.

04:59.34
davidmcdermott
Well I know listeners are looking forward to hearing your views on on ai it's it's the hottest topic today you know chat Gpt is is everywhere on Linkedin and and in the news and people are wondering. You know what is it is it useful. You know is it is it a concern. You know how? how can we use it to our best benefit you know should we be avoiding it, etc, etc. And ah and I'm particularly interested in the fact that you've also got an ethics background and and stance in regards to Ai and ethics. So before we dive right into those topics.

05:33.99
Ross
Are.

05:39.17
davidmcdermott
Can you just explain a little more about your current role. You mentioned you know you're running these informal groups. Um, you know how how large generally are the groups where do you run them and ah what? what is the intention generally of of these conversations that you're facilitating.

05:57.82
Ross
So yeah, my role at Ibm it's um, it's a broad role and when I say I bring together groups of organizations I'm not talking about running you know user groups or anything like that. It's more around putting together teams within Ibm to to solve customer problems. So. You know we may yeah might be in Indonesia. For example, we may be working with a customer who's got a particular issue I'll pull in the right experts from around Ibm and we'll work together to put together a solution and implement that so that that's how that works in in my role at the moment. But um, yeah, in terms of i.

06:35.22
davidmcdermott
Right? Well what one that question I I just wanted to unpack a little more so in the groups you're working with is the customer in the room and and other stakeholders as well as Ibm team members. Was it just an internal focus.

06:57.72
Ross
Notes it is customer facing so normally well sometimes with these large deals they they could be up to 4 different major stakeholders so you obviously have the customer. You may also then have a outside consulting firm or a business partner who you're working together with you then also possibly have a distributor who looks after the distribution and licensing of the software and then you have the actual vendor like Ibm who. The original creators of the the software and even with Ibm there's going to be multiple stake outholders so you may have your your consulting arm. You may have your product management team who are responsible for the sort of evolution of the product. But. Then you may even end up working with actual developers who wrote their software in the first place and you may even bring in Ibm Research who are doing some cutting edge work and maybe 1 or 2 of their Api Apis may be applicable. So it's ah ah these are complex seals. There's many moving parts. Um, and that's partly why I enjoy it.

08:01.65
davidmcdermott
Yeah, really fascinating and is it all over the world or more Australia focuseds.

08:11.15
Ross
For my for me I focus on asiaific. So yeah, it's um, basically the Philippines Vietnam Indonesia Singapore um Australia New Zealand Korea

08:13.29
davidmcdermott
I go.

08:25.86
davidmcdermott
So in terms of the actual content and and Ai let's ask a broad and explorer broad question to start with and what what can you share with listeners I guess in terms of um, a quick synthesis of. Ah, where where Ai is at in today's world um and um, an initial perspective on you know what? your I guess top um insights are around ethics and how it applies to Ai today.

09:01.33
Ross
All right? How long have we got because it's ah it's a big big topic. But that's good. So let's start off with what what we actually mean by Ai because I think you know like many of these buzzwords that get sort of.

09:04.91
davidmcdermott
Try try and keep it to under 50 minutes for this answer.

09:20.41
Ross
Get thrown around and sometimes they mean all all things to all people which means they they actually lose their meaning. So my definition of artificial intelligence is very simple. Let's start with human intelligence. Let's think about what we mean when we meet an intelligent human being. Um. Obviously there's many facets to human intelligence. Some people have you know an artistic intelligence. Some people have linguistic intelligence. Some people have numeric intelligence. Some people are just maybe they're deep subject matter experts on one particular field. And so they have a lot of factual knowledge about a certain field but they may not have that broad intelligence. So human intelligence is multifaceted. Um, now if we take any one of those facets and we train a computer to demonstrate that sort of intelligence. Then that's ah, an aspect of Ai. Okay, it's quite a simple concept and um, you know there's some things that computers do that we don't call Ai now but when they first were when they were first created they were mind-blowingly artificial intelligence and 1 of them for example is memory. So one so 1 aspect of human intelligence is someone who has an incredible memory now. We take it for granted that computers don't forget anything? Yeah, that computer memory is basically infallible unless you you know unless your hard drive gets destroyed. So um.

10:43.36
Ross
You know that's an aspect of human intelligence which we take for granted that that computers can do um the same with arithmetic you know before mechanical calculating devices that was 1 aspect of human intelligence as being extremely good at arithmetic we take it for granted that computers can do that now. But the um, the real evolution of Ai basically is that computers are now being trained to do more and more sophisticated aspects of of things that we associate with human intelligence and a very good way to think about this is what we call narrow. Broad and general Ai these are 3 broad categories of Ai you so narrow. Ai for example is where we take any type of task and it's a task that's narrowly defined. It's well defined. We know exactly what success looks like and doesn't look like. And then we use 3 themes to solve that problem. We use data. We use compute power and we use algorithms and we solve that problem to a high degree of accuracy. So an example of a narrow Ai problem would be can we translate from one language to another. That's very well defined. We know exactly what success looks like and basically that has been solved to a very high degree for many languages and there's many of these narrow arrow problems which are slowly being ticked off 1 by 1 by by computers.

12:12.55
Ross
But where things get really interesting is when we move into what we call broad ai and this is where we take a much more complex much more human problem and try and train a computer to do that and I'll give you an example of that. There's a very interesting research project come out of Ibm. It's called project debater and the the challenge was. Could Ibm research train a computer system to hold a live debate with one of the best debaters in the world from Oxford University on any given topic under standard debating rules. Okay, now that is a much harder problem. It's not like it's not as simple as just translating one language to another. And 1 of the challenges. For example was how do you know when you've succeeded well because you know debates are normally adjudicated by a person and so to win the debate. You need to convince that person that your arguments are more compelling than the opposition that is. Much frazzier. It's much less well defined than just translating from one language to another for example and if you if you search for project debate on Youtube you can watch the debate. It is mindblowingly amazing. How how accurate and how compelling the debate. Is now how how was it done. How is this broad ai I topic done. It was done by breaking down the challenge of holding a debate into multiple narrow Ai problems. So these are things like can we detect claims in ah in the document can we negate claims.

13:42.69
Ross
Can we analyze the logic of an argument can we identify flaws in the in the logic can we fact check what's being said and and bring up fractuual errorrans all of these are subtasks which need to be solved in order to hold a debate. And that's exactly how that that research problem how it how it was solved and general ai I won't talk too much about because it's sort of it's it's far in the future. Basically, but essentially what general Ai would be is when computers train themselves to solve your problems. Um. And not only that that computers would have cross domain transferability of knowledge at the moment say if you're a human being and you work for a certain industry for 10 years if you go to a different industry. You don't go back to 0 all that knowledge that you've gained in the first ten years stays with you. But you can transfer it to a new domain. You can make those connections make those analogies and you're up and running computers generally speaking are extremely bad at that if you train them to do 1 thing and then put them in a new environment and they normally don't do very well so that's that's the vision for general Ai, but most researchers will say that's. Sort of around twenty fifty or or beyond. So. It's really the broad ai I think is is where things are really taking off so the ethical considerations around Ai is.

15:10.64
davidmcdermott
And and just before you go into that ross sorry to jump in um, but with the general Ai which you say is you know, perhaps around twenty fifty and beyond is this the kind of Ai you know the the movies that come out. Ah um, that ah i. Predicting you know where Ai might get to and showing scenarios of you know robots overtaking the world and humans becoming the slaves or exterminated by robots is that what would would that fit in the category of what you're calling general Ai or is that something else.

15:39.93
Ross
Ah, that's really that's one sort of scenario that some people think um, you know could play out if if general ai becomes readily available and is not used Well. It's some people call it the singularity where the basically the the robots take Over. Um. So yeah, But as I said that's more in the realm of Science fiction at the moment but some people are really concerned about it and they take it seriously. Yeah.

16:05.34
davidmcdermott
Are you concerned about this singularity occurring. No.

16:12.60
Ross
No, not. Ah um I mean it's not I Really seriously can't see that happening. Um, you know I think when we're designing. Um, you know when we're designing systems like this. People are aware of the potential. So. It's I believe that the you know safeguards will be programmed into it. Um, and at the end of the day these system to design and run by humans and I think they're going to remain in control. That's.

16:43.51
davidmcdermott
I Guess there's always the power button that you can just press Yes, so more you know more to and our immediate experience. What.

16:46.90
Ross
Ah.

16:49.69
Ross
Um, yeah, yeah, take me.

17:02.23
davidmcdermott
Ah, what? what's the ethical and perspective that you hold or ethical perspectives around what you what you understand about Ai today which is a lot more than most people. It's it's your area of expertise and and what we do need to be concerned about ethically.

17:20.48
Ross
Um, well there's so many well I think first of all, it's it's good to define what we mean by um what we mean by the ethics of Ai so whenever we have a subject like Ai. Normally you can look at this the Ai through 3 different lenses. And the one that most people look at it from is the technological lens and the technological lens tells us what a I can do and that's that's the thing that people are really excited about you know we could do this with ai look what Ai can do now. So that is purely scientific technological then we have the legal lens where we look at it and we say what does the law allow us to do with Ai right? What can we do with ai and what can't we do and what should we not do in order to avoid. Being fined or sent to dro so that's the legal it lends and then the ethical lens has a completely different question. It says what should we be doing with Ai what is is it the right thing and there's plenty of things that ah technologically possible are entirely legal. But.

18:30.80
Ross
Ethically questionable, um, and the thing about ethics is ah there are many different ethical positions that people hold so there's a whole range of different. So basically what what ethics does is it gives you a framework to answer the question is this action. A good action and there's but there's lots of different ways to answer that and 1 of the most common ones today is what's called the uilitarian view of things which basically says if an action creates the greatest good for the greatest number of people then it's a good action. Okay. That is extremely modern. It's quite a modern framework. Um, but then there's other there's ah, there's other ways of answering whether or not it's in good action which are things like you know religious ethics for example or dale day o onto day are.

19:24.92
Ross
Ontological ethics which basically says we have certain duties to each other as humans so depending on which ethical position you come from? you're going to answer this question around what should or we shouldn't be doing with Ai in a different in in different ways and that's part of the reason that. Um, you know the ethical questions around ai are quite challenging. Um, so in terms of um, some of the big ethical questions. Well one of the big questions is this Um, as we. As ai becomes more advanced through technology. Essentially what we're doing is we're outsourcing more and more of the things that humans used to do to computers and the big question is what effect does that have on humans right? and just in very very simple terms. You know when. When calculated become ubiquitous people become less good at arithmetic just because that that part of their mind didn't get exercised. Um, we see this with spell checking now people are becoming not less less capable of of spelling because Spellcheck will do it for them. And the same with grammar and now with chat gbt which you mentioned in your introduction if people start using that to start writing and composing sentences and paragraphs is that does that mean that that part of their mind is not being exercised and therefore they'll become not so good at it now. These.

20:55.97
Ross
These questions are I think they're quite important because they are going to have long-term effects on what it means to be human and and what it means to be a person. Um, and there's there's other aspects to it as well because you see it's quite entirely possible. For example that. You know these tools that can do things for us. There's 2 different ways of looking at them one is if you already know how to do the task if you already know how if you're if you're already been trained up in arithmetic spelling Grammar Composition reasoning you can already do those things you can use those tools to help you do them more efficiently but you have oversight of them. You can see if they're doing a job or not right? You can see where they've made an error if you've never learned how to do those things and you're using the do using the tools to do them for you because you can't do them.

21:37.63
davidmcdermott
E.

21:50.85
Ross
That's another. That's another question and it could well be that it becomes a divide between people who have taken the time to educate themselves or have been educated well and use the tools as opposed to people who are just using the tools because they can't do it themselves and I think that's quite important.

22:06.80
davidmcdermott
Yeah, really interesting point and how might that play out I mean no one has a crystal ball. But if if this divide does become a reality and there are this category of people who haven't been educated and um, the. You know they're using the tools but can't spot errors and and don't really understand how it works and then there are these people who are able to have oversight of these tools and spot errors you know, but what might happen if the world continues that way.

22:33.67
Ross
Well, you know you could get.. Basically yeah, you can get a you may get a divide between you know between those two groups of people and basically increasing and inequity in societies which would be. Unfortunate.

22:53.64
davidmcdermott
I Mean there already is a fair bit of inequity. But that's that's often around um wealth and those who for various reasons managed to acquire a certain size of asset base and ah and those who didn't and there's that obvious obvious economic divide.

23:08.75
Ross
Is.

23:10.54
davidmcdermott
I mean is it more than than what you would call the economic divide because obviously if if we're talking salaries and payments or you know entrepreneurship Those who have oversight and and understand what's going on are more likely to be paid and paid better and and create more wealth and those who.

23:26.27
Ross
Um, the.

23:30.17
davidmcdermott
Who don't will be ah at lower ends of the spectrum is it really? ah fundamentally an economic divide that you see increasing or is it other things as well.

23:37.16
Ross
Well I think it it is. It's it's yeah the economics is a big part of it. But it's also just access to information and knowledge about what's happening in the world and those sorts of things. So yeah I think it could lead to in in equalities in in many aspects.

23:57.43
davidmcdermott
Well you and I spoke previously and one of the things you mentioned that really stuck with me in terms of ethics is if and and let's talk chat Gpt and and other related platforms. Ah you know I know. Microsoft has has invested 10000000000 in openaii and that's publicly public information openai I are the creators of chat Gpt and and it is the the platform that's getting most um publicity at the moment. Um I've I've been playing around with it myself and know. The jury is out for me at the moment I think it has some uses personally and and I wouldn't use it for other things. But what's you know? what? That's the the point you mentioned that I remember was if young people particularly. Ah. Ah, going to be using chat gp to write essays. You know do homework for them. You know or whatever in the future and in jobs you know documents they're they're working on and using platforms like chat Gpt to um to create what otherwise they would have had to apply their you know their brain and and use thought processes to.

25:05.99
Ross
Move.

25:06.77
davidmcdermott
To create you. You made you made this comment. You know they can just let chat gbt do that and then just scroll tiktok and as as an alternative use of their time which is a very ah very um, ah present reality I mean I and I know a whole bunch of young people young adults and.

25:13.96
Ross
Fifth.

25:26.27
davidmcdermott
And teens and they are very much when this spare time often on their phones on Tiktok and and ah and other platforms and hitting getting getting hit up with the. The dopamine and oxytocin with the the scrolling which is scientifically well proven and and in addiction addictive patterns of scrolling newsfeeds but can you can you speak a bit more about that that aspect ethically of know as ai becomes more and more. Usable and understood by um, at least a portion of society and and the planet. What what some of the concerns are around that that part of it.

26:00.83
Ross
Are.

26:09.32
Ross
Um, so I mean I think there's a number of aspects The first thing is I think we shouldn't overlook how much good Ai is doing um because there are there are thousands of examples where ai is increasing increasing the efficiency of what people do and again we take. Um, some are so much of it for granted we daily' even know it's happening so the the amount of good and the amount of economic good that it's doing is unbelievable. You know and these incremental increases in efficiency and therefore of economic activity. And wealth creation that are taking place through that are being driven by Ai are incredible. It's the global impact of of Ai across the world is measured in trillions of dollars and that's getting better all the time so we shouldn't overlook that um but the other is. Of your question was you know the vision a bit like in the industrial revolution. The the vision of you know machines or computers. In this case was that they would take away the drudgery they would take away the drudgery the um yeah know the donkey work of what humans used to do and that would leave humans. To evolve and spend time doing the the high level creative. You know, really challenging really rewarding work now. Um, as you mentioned that that was the vision. The reality can be that.

27:38.59
Ross
Some some aspect you know some a certain proportion of of people when when their time is freed up that don't necessarily evolve and elevate. They may spend that time on on less worthwhile activities at the end of the day. Ah, you know it's a free country people can spend their time as they Wish. Um, but um I guess that's that's the you know I think people are still coming to terms with it. You know that it could be.. It could be a bit like the gut in the hype cycle is that where. You know when when new developments take place and we have additional way we have different ways of working in additional. You know, discretionary time at our disposal. It takes us a while to to work out how best to use them.

28:32.66
davidmcdermott
So you you talked about um you know the very positive impacts of Ai and saving trillions of dollars worldwide can you break that down a little bit more and ah you know I know a number of clients I'm working with um.

28:37.54
Ross
Um, um, yeah.

28:48.24
davidmcdermott
Ah, looking at digital strategies for certain offerings and Ai and machine learning are um, ah absolutely the focus of you know how to optimize and and gather data most effectively and use that data most effectively for better outcomes in in many different ways and um, you know there are many proven examples of. Organizations that have been doing this for for quite a long time actually whether whether they called it ai and machine learning or not but can can you break down a little more for us how how is ai saving the world trillions of dollars um and and what are some examples of that specifically in terms of stories.

29:11.15
Ross
Um, money? Yeah, um.

29:17.73
Ross
Are the. Um, that's right? So um, and it's not just it's not even just the the efficiencies. It's also you know what we like to call Ai for good where things that rarely benefit humanity that would be impossible without use of Ai or. Not impossible. They would be cost prohibitive so one. Ah 1 yeah one of the very encouraging stories I can share is things like the or early diagnosis of breast cancer using ai technology to scan the images and detect. Ah, early signs of cancer which are undetectable but to the human eye that to me is a perfect example of a really good use of Ai. It's something that um you know you can train you train the image classification algorithm on thousands of images. And then that that learning that learning that's been inherent and is now inherent in the algorithm is available to benefit women all around the world and that's a classic example of how technology and Ai can can seriously benefit another very similar one is the early detection glaucoma.

30:39.28
Ross
So simply by using images that have been scanned from the back of the eye you can train algorithms to classify those as to whether or not that person's at high risk so that so I think medicine is something that um yeah, everyone would agree. Where Ai can really play a huge part and it's even other ways that it's being used in medicine which is for example, using patient data to make predictions about whether this where the certain people are at high risk of of hospitalization in the near future and then. The practitioners can then recommend interventions which prevent the hospitalization ever taking place now. That's a great outcome for the patient but it's also a great outcome for the health system because hospital action is one of the most expensive things that ways of treating somebody. Think things like that they are really exciting. It's a win-win all right? It's a win for the patient because they avoid going to hospital. It's a win for the public because they don't have to pay for the hospitalization. Everyone benefits from that and there are there are many many examples of that. And there's other examples that again, we take for granted say things like spam filtering that is pure Ai classating classifying emails as they come in as to whether they're spam or not. We just take that for granted that would be impossible without.

32:05.78
Ross
Machine learning models doing that and the same with fraud detection and cyber security detecting cyber security breaches those sorts of things again a lot of that happens almost below the radar. We just take it for granted that it happens. But it it doesn't happen by accident. Those are some quite exciting you know advances in Ai and there's another one that I'm very interested at the moment and this is it's a very different use case and it's a very original piece of thinking that's gone into it.

32:26.70
davidmcdermott
E.

32:44.47
Ross
But basically the the scenario is that when governments do large-scale infrastructure projects. They need to gather community feedback to find out what the community thinks about whether it's a new bridge or a motorway or. You know if there's even if they want to get community feedback on say the transportation needs of a certain city and typically they do that by sending out surveys now what was noticed was that say in ah in a multi-ethnic society. Survey responses were almost invariably only from people whose where english was their first language. Um, but surely it can't be that just because you're english is not your first language doesn't mean you don't have a view on the the transport needs of the of the of the city. So a company based in New Zealand actually called frankly Ai they developed using using an ai virtual agent. They developed a tool to allow underrepresented communities to give feedback about community projects in their own language in their own time. So it wasn't a structured survey the virtual agent would simply ask them in their language of choice. Do you have any concerns about this project. You know what do you think of it in a very open-ended way and that way they found it was a much more effective way of gathering feedback and could even swap between languages.

34:19.91
Ross
Or for example, this is quite amazing. Some languages like Torres Stra creole often. They have a mixture of the torres straightit words and english words in the same sentence but the virtual agent was able to understand that and respond in kind. It's I think quite. Quite mind blame.

34:40.40
davidmcdermott
Um, it it really is fascinating and yeah, the the bit of research I've been doing and some of the some of the other stories I've heard in terms of organizations.

34:47.30
Ross
Ah.

34:50.88
davidmcdermott
Um, that have been applying this for for quite a while really and are at the cutting edge ah organizations like Netflix Amazon and peloton you know Netflix and Amazon are very well. No peloton you know they sell. Ah um, ah a kind of um stationary bike. It's an exercise community and and a big part of there.

35:01.78
Ross
Um, were.

35:09.78
davidmcdermott
Revenue base is the online courses and the groups you know the live groups. It's kind of like a Zoom format where yeah, you can have apparently up to 2000 people in a in a big sort of Zoom room kind of format that people pay a subscription to and they they've been using um machine learning to. To understand their users' needs as you know as finely as possible to give them. You know what? what? they're really looking for and continuously refine that Netflix has been doing exactly the same since its conception. You know you can you can? Ah, it's another question as to what the ethical aspects of that. Are you know that comes down to as you said, what. What it what is good and is it good for people to be watching Netflix more and more you know that's not not a topic that we have time to dive into today Amazon um, optimizing service delivery you know, understanding what products people want and how can they give people more and more of what they want. So. And tell me if I'm wrong, but my understanding is what we term ai and machine learning has actually been going on for for quite a long time It's not even though chat gp has exploded over the over all sorts of platforms recently now what this technology if if we if you can call it that.

36:20.39
Ross
Um, nothing.

36:26.92
davidmcdermott
And is not a new one. But what's what's your perspective on that has it been around a long time. Is it new. Is there something new. That's you know, suddenly here or or has it just suddenly hit hit the public in a bigger way than before.

36:39.16
Ross
Now it. It is a it's certainly an evolution It's so yeah, it's it's not ah, it's not a revolution. It's an evolution and Ai has been evolving since the first you know babbage calculating machine was created so it's absolutely in evolution and even. Things like say hat gbt. In fact, that's just 1 example, it's it's but it's runs off what's known as well. There's a foundational model known as gpt gpt is one of many foundational models. Um, and. They've been around for quite a long time. They've been used in many different scenarios. In fact, they're used in a lot of bibian products already. Um, but what is really interesting and not many people probably understand this very well is one of the big challenges with. Building a machine learning model for example is to have a a labeled data set. So for example, if you want to predict say you know, let's say you wanted to build a very simple model to predict the the weight of that person who of the next person to walk into your. Doctor studio if you want to make a prediction of what their weight would be the first thing you need to do is build a data set and 1 of the features in that data set must be the weight of actual people real people so you can train the machine to say that okay height is you know, correlated with with weight and we can.

38:03.80
Ross
We can build a model on that and make a good prediction so that training data set the label data set is it's absolutely fundamental. But they're very hard to create the very They're very expensive and very time consuming to build a really good training data set. And what the thing with foundational models particularly work with respect to language is the the insight behind them was if you have a sentence for example, any good. Any standard sentence which you know makes sense. You can use the first half of that as the the exam question. Basically.

38:40.70
davidmcdermott
A.

38:40.22
Ross
And the second half you know is the response variable and it's it's it's a labeled data set so you know you can say well if we know the first few words of a sentence can you predict what the rest of the sentence will be now a a sentence which is grammatically correct becomes a labeled. Data point in that dataset and what that means is you can now train models on thousands and thousands of documents and that's where that was the really exciting breakthrough with respect to foundational models and that actually applies to any sort of data that has a sequence.

39:11.91
davidmcdermott
Um.

39:15.47
Ross
Whether it's music whether it's image that has a sequence of pixels or even computer code. For example. So yeah, they've been around for quite a while. Um, the reason that it's all come to. People's detention is that chat gut was exposed. You know to the public. But.

39:19.73
davidmcdermott
Are.

39:33.32
Ross
They've been running in in other products for a long time. Um, so what was the other part of the question.

39:41.21
davidmcdermott
Well just just to explore that one just a little more deeply so with a platform like chat gbt which is the publicly known one. Um, so is is it true that they've ah.

39:43.79
Ross
Are.

39:54.40
davidmcdermott
Essentially has a very very large database of all sorts of datasets that it can draw on as you say you know certain grammatical sentences that are that are correct that if you if you put into chat Gput you know and create a website landing page to sell. Um.

40:01.26
Ross
Are the.

40:12.59
davidmcdermott
You know to sell baked beans to ah you know to ah a Target market in Singapore you know just to pull to pull a product out of the air and and a market and and it will actually have a first draft of that it will just you know type it up for you and and then you can say you know refine it to um.

40:22.35
Ross
Um, rather um.

40:31.74
davidmcdermott
Exclude that bit and talk more about you know why baked beans particularly um or you can even just say make make this bait bean product more attractive to this particular market and give me a second draft of this landing page and it will do it and it will just keep refining it as you suggest refinements is is it. Ah, creating new things or is it literally just drawing on pre-existing data sets very vast amounts of them and linking them together in ways that it believes is the best fit for what you're looking for is it actually creating new things.

41:00.79
Ross
Um, the.

41:06.30
Ross
Well, it's It's doing both so it's using the data that's being fed into it and recombining it in new ways to create a new artifact. Yeah, it's ah it's called Generative Ai so you know you can generate. Other programs which are similar will generate new generate new images which have never existed before but of course it's doing that by being trained on on previous images and previous Algorithms. Yep.

41:25.40
davidmcdermott
E.

41:33.29
davidmcdermott
Um, and it does that fit into either narrow broad or general or is that a separate matter.

41:40.16
Ross
Um, it's probably yeah I would classify it as being on the cusp between narrow and Embroid Certainly it's It's certainly doing a harder. It's it's it's it's addressing a harder question than just a narrow narrowly defined one. So yeah, it's where I would place it at the moment.

41:59.11
davidmcdermott
Okay, so time is rapidly running out and I I don't want to miss. Um, the discussion around Ai and leadership. You know this is the evolved leadership podcast and there there are a whole bunch of listeners and. Here on the show. You know who are who are running businesses in in all sorts of industries or um, ah, senior execs in you know across corporate government and social sector organizations of all sizes and and in all sectors and I know that you know my clients I'm speaking with are.

42:25.79
Ross
Um, the.

42:32.38
davidmcdermott
Ah, deeply interested in this topic of Ai but obviously not experts in it and and are asking. You know what? what do we need to be to be understanding as this is expanding um particularly recently much more publicly around the world and people are are taking a much deeper interest in it.

42:43.77
Ross
Um, you know.

42:50.53
davidmcdermott
Many organizations are using these terms Ai and machine learning which I also actually want to ask you you know? are they different things. Um, you know what is Ai what is machine learning if you can also speak to that. But maybe maybe just address that and to begin with but the um the main question to to finish with is.

43:00.62
Ross
Um, yeah, are yeah.

43:10.31
davidmcdermott
Ah, what would you suggest to leaders you know of um of organisations of all sizes. Whatever their leadership position. What what do they need to be understanding about this space to be able to be more effective as leaders in in today's world

43:29.60
Ross
Um, sure. So um to address the I'll address the difference between machine learning and Ai first of all so the clue here is in the grammar so machine learning is a verb and artificial intelligence is a noun so machine learning is the activity of training a computer. To be intelligent artificial intelligence is the state that the end state of the computer which then demonstrates some intelligence so that's simple breakdown of the difference training. The machine is the activity that leads to the end state of having an intelligent machine.

43:52.71
davidmcdermott
M.

44:03.47
davidmcdermott
Yep, perfect.

44:05.20
Ross
Um, so that's that one so in terms of leaders and using ai it's ah it's a big big question. Um, you know a lot of people I speak with a lot of leaders. They they ask? Well what? how can we use Ai to improve our business and um. Sort of 2 aspects to that one is how can you use it to improve your existing business and generally speaking the way I break it down is if youna analyze what your business really is it will generally speaking and come down to a series of decisions that you make. And then a series of processes that you then implement those decisions so you know if you're a bank you obviously making them decisions like who do I lend to and who don't I then to or those sort of questions. So the. And and many organizations like if your sales organization. It's Barr you know some of the big decisions you're making on a daily basis are who do my sales reps speak to today. What do they talk about what do they pitch? What do they lead with and so on so these are. You know, a lot of these questions are sort of subconscious but they're actually that's how your business runs and um, then generally what you can do as a leader is you can quantify. First of all how much it costs you to make those decisions at the moment and secondly what is the cost to your business of.

45:33.40
Ross
Not necessarily getting those wrong but making suboptimal decisions and then you can say well is ai is the current state of technology within Ai is it capable of helping me make those decisions better. And what would be the economic impact of improving that decision-making using ai and that's what Ai is really really good at that. It's good at collecting data and then making decisions basically around classification or predictions. So that's a high level way of really. Or 1 way of finding ai opportunities within your business where you can implement Ai to improve decision making and then the second part is the processes of it. So once you've made decisions you need to implement those through a series of processes and that's where Ai can often help with Eff Efficiencyiencies so are there manual tasks basic. Which can be automated through through the use of Ai so that's one way of looking at it. Um, and then the other aspect is is there is there a new business that I can create from the ground up using. The new recent advances in in artificial intelligence and you mentioned a few of these platforms that have come out recently. So you know uber Netflix airbnb and particularly airbnb and and uber they're really quite unique because you know.

46:59.60
Ross
They don't own Uber doesn't own cars Airbnb doesn't own real estate but they've made a market. They've identified a need on one side and a supply on the other side and they're using Technology. They're creating a market and that's where the really exciting and sort of groundbreaking changes are taking place. So this market making platforms I think they're they're quite exciting. But. Word of warning. They're very challenging to start with because you need to solve the chicken need Problem. You won't have buyers without Sellers. You won't have sellers without buyers so you often need to burn a lot of cash in order to overcome that chicken need probably um.

47:44.87
davidmcdermott
Well, we're at time unfortunately Ross but maybe there might be a part two down the track as we um, you know as we continue to watch this space but that was really fascinating and I know listeners will will um. Ah, be 1 deeply interested in in what you've shared and 2 and hopefully get a few insights as to how they can apply. You know what? what is current in the in the world of Ai and machine learning to their own leadership practice. So thank you very very much for your time. Um, for those who do want to. You know, reach out to you in some way or learn more at least about this space. How can they do that and and what would you recommend? they look at.

48:26.00
Ross
Ah, so in terms of reaching out is simple. Simplest way is via Linkedin for me. So yeah, I'd love to I'd love to connect with and converse with as many of your listeners as possible.

48:33.90
davidmcdermott
We'll put that in the show notes.

48:39.63
davidmcdermott
And and what would you recommend for people who want to learn more about this space whether it's books or articles or and particular artifacts.

48:47.83
Ross
Um, yeah, so really depends on how deep they want to go and how much time they want to spend so there's obviously you can yeah there's reading about it and there's any number of good books recently one recent. Book I read is by deep. It's called deep thinking by Gary Kasparov who was notoriously beaten by deep blue the ibm chess computer and 9097 but um, he. That that book I strongly recommend he really reflects on where ai is going and the sort of promise and perils of Ai but there's any number of other sort of general interest books about about artificial intelligence. Um, another. Source of really fantastic information that I would recommend as a podcast. It comes out of Stanford University it's called entrepreneurial thought leaders. Um, and it's basically alumni from Stanford who are invited back and to give they give a talk on how they did it. And a lot of them are Ai companies and some of the sites are really magnificent. So I certainly recommend that um, and then there's online courses. You know there's any number of what they call moocs massive online courses where you can literally be sitting in the actual lecture theatre.

49:51.36
davidmcdermott
Are.

50:07.60
Ross
Of some of the best universities in the world Most of these Moocs are often free if you want to actually get the qualification you have to pay. But if you just want the information and the knowledge a lot of those and I can't I can't recommend them highly enough. They're best in the world and obviously.

50:24.19
davidmcdermott
Um.

50:26.42
Ross
That's only if you want to go deeper if you really want to get into the you know if you want to educate yourself on it and understand what's happening under the covers. That's a very very cost effective way of doing it and a very um, educational way of doing.

50:41.23
davidmcdermott
Cool. Well thank you for those suggestions and thank you very much for your time. It's been fascinating.

50:47.16
Ross
My pleasure. Thanks for the invite.