Podcast: In Business with AI - The AI Roadmap

In this inaugural episode of In Business with AI, Melbourne Business School Dean Professor Jenny George sits down with Professor Jon Whittle from the Institute of Digital Innovation and AI to provide leaders with a practical roadmap for navigating AI.

In Business with AI

In Business with AI is a podcast presented by Melbourne Business School that explores how artificial intelligence is reshaping business, leadership and the future of work.

Hosted by Dean Jenny George, each episode features conversations with AI experts and CEOs who are implementing AI in unique ways in their organisations.

We unpack what’s working, what’s not working, and how leaders can navigate one of the biggest technological transformations of our time. 

Episode 1: The AI Roadmap

 

In this inaugural episode of In Business with AI, Melbourne Business School Dean Professor Jenny George sits down with Professor Jon Whittle from the Institute of Digital Innovation and AI to provide leaders with a practical roadmap for navigating AI.

From his role as technical lead at NASA to leading CSIRO’S Data 61 and creating Australia’s National AI Centre, Jon has been at the forefront of AI for 30 years.

In this episode, Jon shares why every leader has to be an AI leader, how to help your teams adapt to one of the most disruptive technologies of our time, the impact of AI on blue collar work no one is talking about, and why the rubber has hit the road for the costs of tokens.

Transcript

Jenny George

Hello and welcome to In Business with AI. I'm Jenny George. I'm the Dean of Melbourne Business School. And today I'm joined by Professor John Whittle from our Institute for Digital Innovation and AI. Welcome, Jon.

Jon Whittle

Thank you, great to be here.

Jenny George

In this podcast, we'll be hearing from leading CEOs about how they're implementing AI in real time. And we want to make sure that you have something you can take away and put into practice in your own organisation.

Jon is a leading AI expert, the founder of National AI Centre, the lead of CSIRO's Data61, and a former technical lead at NASA. So John, you've got 30 years experience in the field. It's wonderful having you with us today.

To kick off this first episode, though, we're going to pick your brains, Jon, because you were in AI before AI was even a thing. You've written two highly successful books, and we have one right here, AI for Business.

So to set the scene, I want to ask you a little bit about your worldview on AI. If you read the news at the moment, there's lots of doom and gloom, AI coming for people's jobs, the problems of super intelligence and what might happen if AI runs amok. But I believe you're an AI positivist. What does that mean?

Jon Whittle

Well so first of all like you say I've been in the AI field for a long time so I started my PhD at the University of Edinburgh about 30 years ago at that time it was one of the very few places in the world where you could actually do a PhD in AI so I've seen a lot come and go in that time and I think I've seen a lot of predictions about what the future is going to hold and we do tend to have very extreme viewpoints in AI either it's going to solve all the world's problems or it's going to lead to the end of humanity I am an AI positivist which means that I ultimately believe that AI is a force for good but I also understand that there are some pretty serious risks that we need to manage as well and I think my whole shtick if you like over at least the last kind of 5 to 10 years has been let's not be so extreme about things but let's find the balance in the middle because that's typically where the truth lies.

Jenny George

One of the things I think that has been a real feature more recently is about artificial intelligence and the fact that it's come almost from nowhere quite recently, but that's really generative AI, right? AI more generally has been around for a lot longer. Can you tell us your history of how you got interested in AI and how it's developed over the 30 years or so that you've been working in the field?

Jon Whittle

Yeah, so first of all it goes back way beyond me. I was quite late to the party. The term artificial intelligence was actually invented in 1956 at a place called Dartmouth College in New Hampshire in the US and it was very much an emerging field and a group of scientists got together and had a retreat and said you know we need to shape what the future of this field is and of course the first problem they had is well what are we going to call it and a professor called John McCarthy came up with the name artificial intelligence. I actually hate the name artificial intelligence because I think it leads to a tendency to anthropomorphize the technology and it's not. It's just large-scale statistical pattern matching.

I discovered AI in about 1994 so I just finished a mathematics undergrad at Oxford and really didn't know what I wanted to do with my life. I'm not sure I still know but I'm figuring it out and I did what you did back then when you wanted some career advice. You went to a careers library and I spent probably three or so hours in the careers library flicking through books trying to find inspiration. I couldn't find any. I was about to give up and leave and go to a college bar when as I'm going out of the door I saw this one magazine on the table I picked it up and I flicked it through and at the end of the magazine there was an ad for a masters in artificial intelligence at the University of Edinburgh.

I'd never heard of AI at that time but it sounded so fantastical that I thought you know this is what I want to do. Very sci-fi. So I applied and got in, did a masters then a PhD and then got really lucky at the end of that because my supervisor came back from a conference one day and said hey I've just met these people from NASA and I told them about your work in AI and they're interested and three months later I moved from the UK to live in California right in the heart of Silicon Valley.

This was the late 90s early 2000s and NASA at the time was doing what I would still consider to be ahead of its day in AI so I worked with people who were putting AI into space for example something called the Deep Space One that had an AI on it that was a Deep Space Probe where the AI took full autonomous control of the spacecraft for a 24 hour period so it was really an exciting time I would say.

Jenny George

And you've also got a, you know, from NASA to a background in the performing arts. And I believe you've been on stage with the Royal Shakespeare Company at one stage. The arts and the use, I spose of creative output is one of the controversies in AI at the moment. How do you see AI interacting with the arts?

Jon Whittle

Yeah so that's an interesting one and again back to that time in the careers library one of my key choices there was do I stay in tech or do I actually go into drama and I actually seriously thought about going to drama school until I realized that I probably would end up waiting tables if I did that and so I thought well actually I can do drama as a serious hobby so I spent many decades actually doing theatre, acting and dance more latterly in a semi-professional capacity. 

I think creativity is really interesting from an AI perspective because for many many years the narrative that we had was that creativity couldn't be achieved by a machine.

AI is good at crunching numbers, looking for patterns in large data sets but it couldn't possibly be creative right? And then suddenly generative AI came along and we had some of the early kind of image generators like Dali that you know you could just give it a prompt, it could create an image, now you can create video and suddenly I think that myth was busted. It's probably true to say that AI still isn't creative in the same way that humans are but I do think there's actually a case to make that AI has its own version of creativity and that it's no longer a uniquely human characteristic.

Jenny George

Well, let's talk about what's going on here in Australia more broadly. And also, of course, you've been having with businesses as part of the Institute for Digital Innovation and AI. What do you think has changed most about how Australian CEOs are thinking about AI, perhaps even in the last 12 months?

Jon Whittle

So there's been a big shift I think even in the last six months from the end of 2025 to beginning of 2026. So 2025 was all about pilots, typically large companies were rolling out enterprise Gen AI licenses, they were doing a few use cases here and there just kind of test things out.

But that's really shifted now and the number one question that I get asked by CEOs is how can I scale this up at an enterprise-wide level and demonstrate real ROI and that's what everybody's trying to do. I think it's probably fair to say that the majority of companies are struggling with that for various reasons that we might get into but yeah there's been a big shift even in the last six months.

Jenny George

Would you say Australia is moving fast enough?

Jon Whittle

Look that's a difficult one so people always ask me that you know how far behind are we is the question that I usually get asked and look I think the things that we're struggling with in Australia are actually quite universal. There's been reports come out even from the US that show the same trends that I'm seeing in terms of struggling to scale up. I think there are some countries like Singapore and other countries in Asia where maybe they're a little bit further ahead but I'm actually not convinced that we are so far behind the rest of the world.

You've got to remember that Gen. AI at least is a relatively new technology. It's only really three years, three and a half years old and it takes time to figure out how to use that. It's an amazing technology, it can do amazing things but actually putting that into a business where you get real value out of it takes time I think.

Jenny George

We know that Australians broadly are using AI at really quite high rates. In fact, Claude just came out recently saying Australians are the highest user of Claude. So it's not individuals necessarily, but perhaps the way organisations are taking all of the work that individuals are doing and putting them into organisational systems, that's the tricky bit, I think, isn't it?

Jon Whittle

It is. So we are number one in Claude usage. We get the gold medal for that and I'm happy to have played a small part in that. Claude is definitely my AI tool of choice.

Jenny George

Mine too.

Jon Whittle

Yeah, look, everyone's using AI although I would say that again the tools have really had a step change in capability in the last six months. If you take AI coding as an example, at the end of 2025 we had AI coders and some people were using them but they were a bit hit and miss. And tools like Claude code really matured I think in early 2026 to the point that every software engineer that I know is now using AI coding tools as a core part of what they do. But I would still say the majority of people working in the workplace are probably only using about 10 to 20% of the full capability of AI tools.

You know, they're using it to summarise documents or write emails and things like that, but they're not necessarily unlocking everything that you could get. From an organisational perspective, you know, all the studies show and all the work that I've done personally with organisation shows that, you know, the real trick to unlocking value in an organisation is not about the technology, it's about the organisational and the human aspects.

So it's things like do you have a clear leadership vision about what you're actually trying to achieve with AI? Do you have the right culture in place? You know, are your people ready for it or are they afraid of it? And do you have the right governance mechanisms in place? So it's all those traditional things if you like, like organisational change management and transformation and how you do that. That's really what the barrier is rather than the technology itself.

Jenny George

And I think we've seen big restructures over the past year or so, some of them citing AI explicitly, the scepticism about whether or not that's actually the cause of much of that. Do you think that leaders are just using AI as an excuse sometimes to reshape the company, which was something they needed to do for completely different reasons?

Jon Whittle

Look, I think it's hard to really know. So there is this term AI washing, which is exactly that, which is, you know, your share price has dipped, you need to do something to get your investors back on board. And so you do some layoffs and you don't want to say that for the real reasons for that, right? So you say, oh, this is because we're embracing AI and automation.

There's some interesting data that's come out recently about the impact of AI on jobs that gives us a little bit of a clue here. So in Australia, we actually hot off the press. We had a report out just in the last day or so from the Department of Employment that shows that so far, AI doesn't seem to be leading to mass layoffs across the economy.

But of course, still early days. If you look at the US, there's some evidence there that junior employees are being more effective than mid and senior levels. And I also read something interesting about career just this morning, actually, that shows that again, in career, there is some evidence that junior employees are starting to feel the pinch from AI. But there's not there's not a lot of evidence out there yet, at least, that we're going to see those mass layoffs that have been predicted by some tech company CEOs.

Jenny George

And there seem to be almost two forces that are being described with regard to AI and productivity. One of the forces seems to be that AI brings the kind of lower performers or the more junior people up to a kind of a good base level. The other being that AI augments the ability of more senior or experienced people.
Is the evidence there for one or both of those? Does it look like AI is doing one or doing both of those kind of things at the same time?

Jon Whittle

Yes, I actually think productivity is a bit of a red herring when it comes to AI. So if you look at the evidence on productivity more generally, it will say that for personal productivity, AI is likely to increase your productivity most of the time. It can be anything from five to 70 percent depending on the task. But at an organizational level, those productivity increases are at least again not yet here. And part of the reason for that is that you can't just simply aggregate personal productivity to an organizational level. Organizations are complex. There's a lot happening there. And if you don't plan for how you're going to use the save time well, then you won't get those benefits.

But again, I think it's a red herring because what I tell companies is that sure, you can use AI to become more efficient in parts of your business. 

But if that's all you do, you're just getting faster at the things that you already do, where the real opportunity is to actually do new things that you couldn't do before. So I actually always say go back to your purpose as an organization, why you exist, and think about what are the things holding you back from delivering on your purpose and can AI help you with that? And then you start to see AI as a delivery vehicle for purpose rather than just driving efficiency. And if you do that, you end up with a very, very different lens on your use cases than if you're just thinking about speeding things up.

Jenny George

So helpful. And even on efficiency, I think, as you pointed out, you can have individual tasks that get faster. But if your process end to end hasn't changed, hasn't been rewired, that individual saving may actually have zero effect on the overall sort of process inside the organization. That's a massive organizational task for companies. And when we see that kind of technological change in the past, we see that it takes many, many years more for those things to land than the initial things. So in that light, what do you think leaders should be doing to get their organizations ready for both the purpose-led changes that AI will bring, but even those rewiring needs? Are there some upskilling or various different things that leaders ought to be doing now in their organizations?

Jon Whittle

Yeah, absolutely. And that's a really good point that you make that actually this stuff takes time, which is very much contradictory to the message was sold about AI, right? You know, it's fast, it's speed, you know, you can do things immediately. But as an organizational level, you've got to do the hard work, actually. So from a leadership point of view, my view very clearly is that every leader has to be an AI leader now. This is not something you can delegate to the tech leaders in your organization, because the people factors, the budget factors, the operations factors are equally as important to the tech itself. So you're not going to succeed at an organizational level unless all your leadership, your senior leadership team and below are aligned.

So I think there's four things for me that if you want to become that kind of AI leader, you should be thinking about. I say it's the what the why the how and then the bigger picture. So the what is simply educate yourself on AI. You know, there's a lot of hype out there. There's a lot of noise. And you need even if you're not in tech, you need to know enough about AI to sort out the true story from the rubbish, frankly. But on the other hand, you don't need to go away and study your PhD in AI. Like I did, you don't need to become a guru. You just need to know enough to be asking the right questions. So that's the what.

The why is really that purpose piece. So really go back to your purpose and think about why you exist and frame AI that way. The how is the good old fashioned stuff that we talked about, you know, organizational change management. How do you bring your people along on the journey? The good news there is that most leaders already have those skills to a certain extent. They might need a refresher, but they've done that in leadership training. So really lean into those skills. And then the bigger picture is really about understanding things like the ethics of AI, things like cybersecurity risks that AI can bring. And again, don't just expect your CTO to sort those out. You can understand those at the ELT level or even the board level. But yeah, I think every leader has to be an AI leader nowadays.

Jenny George

Do you think AI is changing the nature of leadership itself?

Jon Whittle

That's a really good question. And it's a question that gets asked a lot. I think my answer is yes and no. So I think in many ways, no, because a lot of the stuff that you need to do to get AI to work is the same. Same stuff that you need to do for any big change in an organization. It's about, you know, getting your leadership team aligned, getting a clear vision, really kind of focusing on the execution of a strategy, all those kinds of things.

But there are probably some aspects where AI is bringing, if not radical changes, certainly new aspects. You know, if we do enter a world where we've got humans and AI agents working alongside each other, then that's a different type of leadership challenge because you need to, I'm not sure you can just manage agents in the same way that you manage humans. They're very different things. So, there are nuances.

Probably the biggest thing I say now to leaders is, which is a change that AI is bringing, is that there's a lot of people out there feeling very overwhelmed by AI right now. There's so much information coming out every day. It's impossible to keep on top of it. Staff are afraid in many cases, you know, because we hear all this stuff about job losses. So I think leaders now really need to be leaning into that care aspect of leadership. We've always done this, of course, but I think really double down on that and really think about how can I take care of my people and help them through what is, you know, one of the most disruptive technologies that we've done. And that's not something that we've ever seen, right? So yeah.

Jenny George

I've been thinking quite a lot about how decision-making, decision rights, who makes the decision, but also how you take responsibility for a decision. That feels to me like an aspect of leadership that it's not a complete change. Absolutely not. But there are differences when AI is significantly involved in an organisation. Have you been thinking about how those kinds of aspects of implementing AI are going to change leaders' decision-making within organisations?

Jon Whittle

Yeah, I think that's a good question and it's very much emerging. And again, decision making is one of those things that people say should be reserved for the human because it requires judgment and all those things. I think that's a bit naive though, because I think what we're already seeing, at least micro decisions being made by AI all the time, by definition.

Jenny George

And agent is making a decision. Otherwise they're not accepting it.

Jon Whittle

Or even if you're not using an agent, even if you're just having a conversation with a chatbot to produce a document, there are little micro decisions in there that, you know, you might ultimately have full control over what that document looks like. But we know that a lot of people defer to AI. So we're already in a situation, I think, where AI is making decisions all over the place. I think the challenge then comes of how do you make sure those are the right decisions, given that you don't really have a lot of visibility. So this comes back to good governance, which again is probably one of the top two questions that organizations ask me about now, which is what kind of AI governance should we set up?

Again, the evidence there shows that, because there's a natural inclination there to say, okay, we have to have human oversight of every single decision that an AI makes. But all the evidence shows that first of all, that's not possible. If you try to do that, you'll just slow yourself down. So the trick is putting in kind of escalation levels based on levels at risk. So those lower level risk decisions that already AI is making, you just let those go through. But the more risky ones, you do need human oversight. And there's some really good research out of Stanford that has studied companies and their guideline is that something like 80% of the decisions you should just let AI make and then escalate to human oversight in 20% of the cases. So that seems to be the sweet spot.

Jenny George

And is there any particular guidance about how to experiment responsibly? Because that is where a lot of companies are at the moment, they're still in that experimental phase. What does a governance framework look like for experimentation?

Jon Whittle

Yeah, so I think you use the word responsibly there. So there's a whole area called responsible AI, which is how to use AI responsibly or ethically or properly. I think one of the challenges there is that there's too much guidance out there. I started working in this area probably about six years ago. In fact, Australia was one of the first countries in the world to have a responsible AI or ethical AI framework. Now, everyone has one and every consultant and every large company and every business school has brought out some kind of guidance.

So I would say try to really kind of distill it down and think about how you're using AI. And for me, there's kind of four simple categories. So you can surrender to AI, you know, you just put a prompt and whatever it gives you back, you accept it, wouldn't recommend that. There's verify, which is where you fact check the stuff that comes back. That can be okay for certain situations. I think where you really want to be, though, is what I call partner, which is where you're using the AI as a co-designer, as a thinking buddy. And you're having a conversation with it to help shape your ideas. And I think you get the best of both worlds there because you're not you're not delegating your thinking. You're putting your thinking in, but you're stress testing it with the AI. And maybe it creates things that you haven't thought of either.

There's also fourth category, which is just reject, which is where, you know, you get an output. It looks so terrible that you just say, it's been quicker for me to go and do this myself. I did an interesting experiment the other day because I was interested for me personally, which of those categories do I tend to fit into? So I thought, how can I measure that? Well, actually, you can use AI to measure that.

So you can if you're using Claude or whatever tool, you can just ask it based on my conversations. Which of these four categories do you think that you that I am in most of the time? The good news is for me, at least, that I seem to be in the partner category about 60 to 70 percent of the time. But that's I think whenever it comes to advice or guidelines for people on how to use AI, we need to really simplify it down because there's again, there's just too much information and noise out there in general.

Jenny George

What are you seeing as the most interesting AI use cases here in Australia? You talk to CEOs, you look at organizations, you advise on responsible AI, you've been looking at the outcomes of experiments. Are there any of those experiments that you see working really well and those use cases are bearing fruit?

Jon Whittle

Oh, great question. Look, there's a lot. I would say probably most of the AI use cases right now are relatively simple and straightforward kind of process automation kind of things. For something a bit more interesting, I'd probably go back to some of the work that I've done in my prior role.

So we did some work, for example, on the Great Barrier Reef, where we were looking at how can AI help reduce the numbers of Crown-of-thorns starfish on the reef, which are poisoning the reef. And we developed an underwater glider that can go up and down the reef and it had a computer vision system on it and it could detect where these Crown-of-thorns starfish are so that you could come back and remove them. And showed that that was an order of magnitude more efficient than a human diver doing that, which is actually the traditional way of doing this.

I'm always at pains to point out at this point that we're not putting human divers out of work because these are typically scientists that have got better things to do with their time than just look for Crown-of-thorns starfish.

That'd be one example, but we did work as well on, you know, we had an AI system that could detect the path of bushfires in real time so you could put scarce resources exactly where they're needed at any given point in time. So examples like that are really inspirational to me. It's probably not where most organizations are at right now, but I would say as an aspiration of where you want to try and get to, it's those kind of things I would be thinking about, which again is really going back to your purpose because you're only going to get to those more inspirational examples if you really start with purpose.

Jenny George

When you see businesses, where do you see the business functions taking AI on board and using the most effectively? Is it the operations that that kind of frontline? Is it in HR? We're seeing many different functions talking about AI and how they're going to use it. Where are you seeing most effectiveness?

Jon Whittle

So it probably is more in that operations field. Some of the others are kind of slower to adopt. There's a related question here, though, which I think is relevant, which is who should be in charge of an AI transformation? Because who you put in charge will determine where those use cases live, right?

My advice on this is that there isn't a single correct answer. I would say 90% of the organizations that I work with put CTO or the CIO in charge. That's not necessarily a bad thing. It really depends on what your AI ambition is. So if you if you're just going to make incremental investments and you're not really looking to fundamentally change your business, then I think putting it with the CTO or the CIO is perfectly fine.

But if you want to be a bit more ambitious than that, and you want to maybe start to disrupt some aspects of your business model, then it's not really a tech team led thing. In my view, it's got to be at the very least jointly led by maybe even like a chief people officer and a CIO. Or of course, you could do what a lot of organizations are doing now and they're hiring a chief AI officer. If you really want that genuine transformation, if you do that, again, I would have them report to the CEO.

Don't make the mistake of hiring a chief AI officer and then having them report to a CTO because you're just again, putting it in charge of the tech team. And I can I'm a tech person. So I can say that I feel I'm not against tech people. But I think through that decision very, very carefully. And I don't think organizations are.

Jenny George

I think there is a cost wall, if you like, that's coming for a lot of organizations as they start realizing that the amount of experimentation they've let loose in the organization is not free. And we're starting to see token costs, the sheer share prices of some organizations and particularly the AI startups are projecting token costs as if they're going to be the same forever. It doesn't seem like that's going to be possible because we simply cannot afford the kinds of use that are being projected.

Companies can't substitute people if AI is in fact going to be more expensive than what they're replacing. How are you thinking about that whole cost element that at the moment is almost shelved, I think, by organizations as they're in this sort of experimental phase?

Jon Whittle

There's lots of things to say about that. So yeah, you're right, first of all, that the rubber is starting to hit the road for companies. You know, the cost of tokens generally has come down, but the usage of tokens is going up all of a sudden. So it's suddenly become a budget line for many CFOs and they're suddenly losing sleep about how do we manage this? How do we cap this?

From that perspective, I think my main piece of advice would be choose the right model for the right job, because even if you're using Claude, you've got a choice between Sonnet, you've got Opus, you've now got Fable, and they cost very differently. People will often default to the most powerful model like Fable but actually for 99% of what we need in Australian businesses, you don't need Fable to do it. You'll be perfectly fine with Sonnet. So set that up as a default in your organization, first of all. 

I think more broadly though, my point would be that the technology costs are not the real cost here. In fact, I always quote Eric Brynjolfsson at this point, because he's got a nice stat about digital transformation more generally, but I think it holds for AI as well. And that is that for every dollar you spend on the technology, you need to budget $10 for the intangibles, the people factors, the training, the change management and so forth. 

And that's a shocker when I say that to people, because they're not thinking about those costs. And things like that. But actually, if you're going to do a proper transformation, you need to be thinking about those intangible costs as well, and budgeting for those.

Jenny George 

And of course, the way that these models are set up, there are frontier models that are genuinely game changers for very hard problems. But it's very much a sledgehammer and a nut problem when you use them to write a difficult email, which simply does not need the frontier model. Tell us a little bit from an AI expert perspective, where are frontier models at and what would you use them for?

Jon Whittle

Oh, good question. So the whole game in town over the last few years in AI has been scale. So more data, more compute. And there's a very healthy debate going on out there, whether just focusing on scale will actually fundamentally lead to new advances in terms of new levels of intelligence.

There's some evidence to show that the intelligence levels of these frontier models is kind of plateauing. And so although it's still increasing, just throwing more data and more compute at it is not necessarily going to lead to a step change. And certainly people like Gary Marcus, who's a big commentator in this field, and I would agree with him, would say that if we want to move to that next level, we actually need new types of AI. And there's something called neurosymbolic AI, which combines this kind of data driven approach with what is often termed good old fashioned AI, which is sets of rules and that you can get that step change from that.

And in fact, if you look at Claude, the big advances it's had in the coding is exactly when it's taken a neurosymbolic approach. So it's got what they call harness around the large language model. So the large language model is doing the whole data crunching, but then this harness around it is a set of rules about how you create software code. And the combination of two is what's given the real advances.

In terms of where you really need those frontier models, I mean, one of the areas that I am very excited for the impact of AI on society is AI for science, which is can we use AI to improve scientific discovery? We've already had some great successes in this area. So DeepMind created AlphaFold a few years ago now, which pretty much revolutionized the whole field of 3D protein folding overnight. Suddenly experiments that might take decades suddenly could be done in a couple of hours.

And I think that's where you're going to need those frontier models and things like that, where these problems are so complex, there is so much data and it's really a kind of system level problem that you need that power. But yeah, let's be honest, most of us in a knowledge job are not solving that kind of problem every day. And so we don't necessarily need the best and latest and greatest frontier model.

Jenny George 

That actually takes a really interesting approach to thinking about what the future might be. So if we are being AI positives here, and thinking that AI has great potential for business, but also more generally for society, science is clearly one of those areas where AI has a lot of promise. Do you see other ways that AI may impact society in a really profound way?

Jon Whittle

Yeah, look, I think you could almost pick any area actually. So you could definitely pick health care. I mean, the personalized medicine, but even just kind of streamlining and optimizing the system. I think government services is another one. You know, if you try to start a small business in Australia, the amount of kind of bureaucracy and red tape, you have to fight through. And I've done some experiments myself where I've got AI to just automate parts of setting up a business like registering for GST and things like that. And it's a lot easier with AI, I can tell you. I think education, I think there's real potential there. You're probably better equipped to talk about that than I am.

So you could almost look anywhere and say the potential here is limitless.

The question will be, how do you actually make that happen? Because any of those things I just mentioned, there are entrenched ways of working. There are existing power structures in place that are going to need to be disrupted for this transformation to happen. And human beings are going to be naturally cautious or reluctant to go along with that. So I think it's going to take some really hard work to kind of push these kind of changes through. But the potential was absolutely there.

I mean, if I look back over my 30 years, you know, I mean, when I started out at NASA just under 30 years ago, my job was to get AI to write code to fly in space. And I gave up because it was all just too hard. You know, we could do it for very specific problems. But if you'd have asked me back then in three decades, would we have the AI coding systems or even the AI chatbots that we have now? I would have said, no, no way, not in my lifetime. It's too hard. So if you fast forward three decades from now, where we are now, I mean, it's impossible to even I think where things might be. But I think it's pretty clear that the potential at least is to really make society a much better place if we can get it right.

Jenny George 

So if you're picking industries that you think are vulnerable to disruption, you've mentioned some where you see possibilities. Are they the same industries that you think will be disrupted? I guess if you were sitting here 150 years ago and thinking about horses and carts and the possibilities of, you know, completely different transportation, you might or might not have been able to view the future at all. Where do you think we are?

Jon Whittle

Yeah, look, I mean, I gave up making predictions about AI a long time ago because the only AI prediction I will make right now is that every AI prediction is wrong. So putting myself on record to make yet another wrong AI prediction is risky. But again, maybe I'll answer the question in a different way. So one of the things I'm quite interested in is the impact of AI on different kinds of work. 

There's been a lot of talk about the impact of AI on white collar work and how disrupted that will be. Almost no one is talking about the impact of AI on blue collar work. And I think that potentially could be disrupted as well. Probably not in the next couple of years, because for that to happen, we're going to need advances in physical AI, which is AI on robots.

But, you know, there's so much investment going into physical AI right now. There's over 140 companies in China alone that are building humanoid robots right now as we speak. Huge investments in the US. You can already buy a humanoid robot to come into your home and empty your dishwasher for you. They're not that great.

Jenny George 

They're not good though, as I understand.

Jon Whittle

They're not. And it's a company that's selling a robot and they are actually upfront about saying that it's not very good right now. And the contract that you enter into is if you take one of these robots, you can call on a human teleoperator to operate it for you. So you can actually get the task done. Now, why would the company do that?

Well, because if they really want to get good humanoid robots, they need the training data. And there's only one way you're going to get the training data. And that's actually put it in people's homes and watch what it does. So as a consumer, you're entering into a social contract there that you believe in this technology that you're willing to give up that data for whatever future benefits might accrue. So I think the one to watch actually is blue collar work as these not just humanoid robots, but robots more generally develop. I think they will see big developments in the next few years. It'll be really interesting to watch how that sector or our economy changes.

Jenny George

And there's a well-known phenomenon with humanoid robots and in fact, chatbots as well, where up to a certain point, it becoming more human-like is quite acceptable. But there's what's known as the the creepy valley. No, what is it? The uncanny valley. Uncanny valley. Thank you. Something like this. The uncanny valley where when something becomes just human enough, but not fully human, we are really repulsed by it. That's an interesting problem for humanoid robots, especially around our own homes, isn't it? Where we need to really think carefully about how human we want to make something because there's this spot, the uncanny valley, where humans are really repulsed by the idea of something that's sort of human, but not quite. Do we see that in Gen AI as well?

Jon Whittle

Maybe. But let me let me let me go back to the humanoid robots because I think if you've the uncanny valley is one, but there's another one which is interesting, which is called Moravec’s paradox. And this is the idea that things that are very simple for humans can be very, very difficult to automate using a robot. So opening a door using a door handle, for example, is fiendishly difficult to get a robot to do, whereas we do it almost in our sleep.

In fact, in my previous role, I had a team of roboticists based in Queensland who were among the best in the world, not a humanoid robot, but drones and robots on tracks and kind of robot dog kind of things. And they developed these robots that won or came equal first in a global DARPA competition to explore underground environments autonomously using robots. But they always used to tell me that these robots could go underground in like a collapsed mine or something like that. And they could climb over rubble and they could do it really, really well. 

But then when they took them out to a piece, a patch of grass outside their lab and tried to get them to walk over wet grass, they would just fall over. Right. So that's why it's really, really tricky. Having said that, we are already seeing commercial applications of robots and aged care is the kind of place where these things are emerging. 

Proud to say that there's an Australian company that is really leading the way here, a company called Andromeda that was actually out of University of Melbourne. In fact, a student at the University of Melbourne had this idea during Covid, built a company that has over a hundred million dollar valuation. I think she's now moved to the US, but they've got a robot called Abby that's got Gen AI on it and residents can talk to that robot as much as they like.

And that's already been deployed in kind of tens of aged care facilities around the country. The story that some of their people told me, which I always repeat, is that there was one resident in one of their clients who really liked Star Wars. And wanted to just talk about Star Wars all the time, to the point that nobody in that aged care facility would go anywhere near this poor guy anymore because they knew they would get trapped in a conversation about Star Wars. But suddenly they could put this robot, Abby, in front of him and he could talk about Star Wars all day. And in fact, the robot knew more about Star Wars than he did.

Jenny George 

That's fantastic. For people who are hearing all of this and feeling a bit overwhelmed, where would you advise someone to start?

Jon Whittle

Yeah, so first of all, I think if you give yourself permission to not know everything, it is overwhelming. Even for somebody working 30 years in the field, I find it very, very difficult to keep up. And in fact, I'd like to give myself permission to say, you know, it's OK if you don't know everything about the latest, greatest tool that has come out.

I do have a couple of recommendations that I'm a big podcast fan. I listen in the gym and all over the place. I like to listen to two sides of a story. We said earlier that there can be very extreme views on AIs. I always try to listen to the two extremes and then make my own mind up.

So if you want a podcast that's more on the kind of evangelism side, there's a podcast called Moonshocks, which is really, really good. Gives you a very interesting look at what the future might look like. But same vein, if you want the kind of anti-AI view, I would recommend a guy called Ed Zittron, who is just really entertaining to listen to. He goes on big rants about how terrible AI is. And sometimes I agree with him, mostly not. But I think, yeah, try to take both sides of the story and then figure it out for yourself and be somewhere in the middle.

Jenny George 

And of course, stay tuned for our podcast as we explore more about the Australian landscape for AI as well.
Jon, thank you so much for joining me for this very first episode of In Business with AI. We've got a little black book here, and we'd love to have your advice. So we're wondering if you would please fill in one piece of advice that you would give to the podcast listeners.

We will then be digitizing this and putting that up onto the MBS website so that people have access to that as we have our various visitors on this podcast, we'll start accumulating that advice.

Thank you so much for joining us today. And please do subscribe to the podcast for the latest in business leadership and artificial intelligence. See you next time.
Thanks for joining me.

Jon Whittle

Thank you.