Future of emerging trends – consumers’ own AI agents

[Music] Welcome to "Mad Digital Next". And today we're picking up the series on the future of Virgin Trends. Specifically today, looking at the world when consumers will have their own and AOA agents. My name is Robert Carr and I guess today is Professor Kastan Moralsi. Kastan is a decision scientist at the University of Melbourne and director of the City of the Brain, Mind and Models. An interdisciplinary research center focused on human and machine decision-making. Kastan's academic history traces through a Zurich University and Columbia in New York. And he continues that international connectivity by overseeing the joint PhD program of the Universities of Melbourne and Boa. Kastan and I are recording today at Mads Melbourne office on the lens of the Rundrapegal. Kastan, welcome. Thank you for joining us. Thanks very much for having me, Frant. I'll cover the privilege of learning about your work by my involvement on the Zurich Board, the Institute for the Picture of Business at the University of the University of Melbourne, that our prize CEO wants to give it a couple of kicks to start. But maybe to start with, can you tell our listeners about the City of the Brain, Mind and Models? I surely can, thanks, Frant. So our center focuses on as you mentioned human and machine decision-making. But what's special about it is that we work across different disciplines maybe is reflected in our name. So neuroscience, psychology, economics and computer science. Because we believe that decision-making really spans all these different areas. It needs to be integrated in order to successfully further our understanding of how people make decisions and how can we make decisions better. Now, the way that we think about, if you like our program for research, is that we have two areas, one which we call basic research, which is really about how do we make decisions? What are the brain processes, for example, that are involved in decision-making, what different aspects of decision-making? There are two major themes, perhaps, that we are interested in, that are linked to two major challenges we encounter in decision-making. What is the fact that we almost never have enough information to predict the outcome of potential courses of actions? So we call that probabilistic uncertainty and they are interesting questions to address, for example, how do people perceive that kind of uncertainty, how do they learn about it, how do they, or how different aspects of that kind of uncertainty, for example, extreme events reflected in their decision-making? But then there is another important aspect or challenge of decision-making, which is the fact that we almost never have enough cognitive resources, if you like, to process all the information that we do have when making decisions. So a trivial example is, I'll get back to that maybe later in the podcast. Suppose you want to choose a new credit card, there are more than 250 options available in Australia, each of these options have multiple attributes, some of them are quite difficult to understand. And so that's just what we would probably consider a relatively simple decision. And so we face as humans, we face constraints, now cognitive resources, and we are very much interested in how various constraints be it memory, be it, if you like, processing capacity, and other constraints, how they affect decision-making. And we actually believe that a lot of the biases that we call biases that pedelics exhibit in their decision-making are consequences if you like of these constraints. That also sort of explains why we believe that an interdisciplinary research approach is important. Decisions always involve the brain and a lot of the constraints that I just mentioned in capacities are related to the anatomy and the arching of the brain. And so the integrating, if you like, different levels of analysis, with neuroscience, psychology, and economics is really, really important. So that's basic research. And then we are quite interested in sort of translating that into, like, real-world applications, which we call translational research. And here the areas we are particularly interested in is consumer decision-making, decision-making in health, for example, how different psychiatric or neurological disorders affect decision-making. And then a new area we are sort of just trying to build up is high performance decision-making. What are the good decision-makers and what makes them really good and what can we learn from them when it comes to everyone else? Well, it really feels like right at the moment, where it is just where some of these issues become more and more to the fore. And it's a credit to here in the State of the University to have done the poor side. So we've set up a centre in this space, you know, I think, some time ago, probably ahead of the rest of the sketching on. I think that a little bit of what you can just describe there around is we have the challenge we see at the bank, but also a lot of our clients in areas like the big superannuation funds for instance. The challenge of customers who are perhaps entering to stages of life where you will prone to areas of problems to decline at the same time as you're being targeted by scanners. And the vulnerability of some of the best people in terms of being able to detect a scam and to be able to how you process and how you react and the hate is possible. But also, I think you're probably wanting a little bit of the space that wants to be able to have our own autonomous AI agents. What are the areas that might be best done by a human and what are the things that tasks the cognitive processes that might be better available by the AI agent, but to take that example and be going with the critic outcast, there is such a that it lays in your expansive information available with a human what we get across in those might be the kind of tasks that an agent an autonomous agent might be able to for you. Is it thinking better than the right one? I think so. I think one point you just made is how the life force affects decision making and one aspect of aging, which is actually maybe not coincidentally one area that we are going to be able to do. And the other thing that we are going to do is that we are going to be able to do that. And the other thing that we are going to do is we are going to be able to do that. You know making decisions about hundreds of thousands of dollars. How do we handle those kinds of situations? These are pertinent questions where we need to sort of make inroads in terms of what actually happens with people, what happens to their decision making capacity, how do we approach that from an ethical perspective or societal perspective. But then of course there is the question, how can we leverage technology? But in order to answer that question, I think we do have to understand how people use technology. And as you said, which aspects of their decision making, but they actually need help with. We don't want to be able to create one app that is going to do everything. It needs to have specific functionalities that help people in certain situations. And so I think we just need to understand that out. But the leverage points are. Yeah, so if I think about AI agents in the form of when our customers will have their own AI agents and our customers will want those AI agents to be able to do in tall. Whether that's for instance in initial payments or like purchases or living money. You know, I kind of gravitate a bit to the view that we have to overture on that digital next both last year and again a couple months ago. But it was visiting from the UK and David talks about both the boring and the complex transactions. And the boring ones of things like when I part my car of a train station, my phone is in my pocket, my phone, those tracks, the it knows exactly where I am. It knows my patented by egg up. I shouldn't have to go over the wall machine and wait my credit card or do anything like that. So paid my car parking, why I agent should just do that for. But then at the other hand, and it probably relates a little bit to what you're saying in terms of some of the super iteration decisions. The complex transactions where it might be optimal for me to move a particular amount of money in my recite it. We're talking about savings account. I particular day under the legislative and CTOs that exist. And I probably were not able to get my head around. What those looks like. What those will bite on our side of mind. So that's the complicated transactions if you want. The David's view is that that boring and the complicated are probably the ones where I most willingly hand over the. The economy or the decision like to buy I get to just take him out of this. What will we. Does that. Is that sort of relate to how you think you know. So I think what I like about this approach is that rather than looking at finance at one big thing that's going to be solved by one app. We pass it out into a little decisions and a little problem. that we could tackle one by one. So without trying to do everything at the same time, as you're starting with the easy targets by having to pay for parking, it's probably a good idea. I think the fact is, many people find finance or managing their finances daunting. Not just budgeting, but getting insurance of being across their insurance needs, managing their retirement funding. And even if you know a lot about finance, if you're a finance professional, these tasks can still be daunting. Right? So, for example, figuring out an optimal savings strategy for your retirement funding or investment strategy, is anything but trivial, even if you have a PhD in investments. It says, "No gold standard that tells you or formula, that tells you this is how you do it." There's a lot of complexity in a lot of areas of finance. But then we also have this issue as mentioned before. There's a lot of information you need to do to digest. It's time-consuming. A lot of tasks are really quite boring, like utility payments and so on. And so I think, even if you had all the knowledge you needed, and all the information you needed, you probably still want to outsource it because you would rather want you to spend your time on something else. But then most people don't have all the information or no one else, right? And all the knowledge you'd probably need. And so there's an even greater need for a lot of people to have some support. And so the question to me is, what's the best strategy to tackle this problem? Well, great. It becomes a process, you know, what you would actually trust, and I would do it for you. And that trust that might take some time, it might be, but you know, at an initial stage, I had over the responsibility for chasing up this particular third-year-old that still has my address from three years ago, and then I don't want to add to it with those personal, late-naked tasks and why I should do that for me. And maybe, but it's on, I think, get more confident seeing and trusting it to actually make quite actual transactions in decisions for me. That's right. A few years ago, we sat down with our team and not to brag, but we all have PhDs. We are all sort of quite confident, handling our finances, but we are still annoyed by a lot of it. As I said, you know, we want to sort of go and lock on to their computer to pay some utility, but it's just another chore that you'd rather also. So we sat down and asked ourselves, this was just at the time when smartwatches became available. Is this fortune? Then it's a fortune. So I wanted to bring this up. Because I think you might become a little bit of a brave. I think you should brag because you did this in '28, and I went to the CES conference and last being his this January, and I wanted what they were showing for the things that you covered in that video six years earlier. So I think you should brag that you were the union balancing, did a spectacular, thanks for this. And we wouldn't put it in the show notes, I'll link to the video that you've got on BBO of this fortune at the police. Can you continue? So the idea was, okay, we now have smartwatches. Wouldn't it be cool if we could have each have our own personal financial assistant or perhaps our own family office in our watch will basically handle our personal finances for us. Anything from monitoring our budget, paying utility bills, making sure that our insurance policies are up to date, making sure that we are on track with regards to retirement savings, managing our investment portfolio, and so on, right? All the things that you would outsource. As the reason we did this exercise, because we wanted to get our heads around about what sort of research would be or development would we have to do right now. Also, this was a few years ago in order to make that app reality, knowing that at that time, and perhaps even today, we don't have all the building blocks, right? To us, it was more like, this is our vision, what would we have to do? What would we have to do as decision scientists and computer scientists and so on to get there? And at the sort of macro level, there were two major areas that we think needed work. One was the user interface. So you want the user interface that is very easy to use for someone who knows very little about finances. And so it talks to you like an assistant, asks you all the right questions, and then basically either makes autonomous decisions or makes recommendations to you. So that's the one big challenge. And then the other big challenge, which is perhaps the bigger challenge, is the logic that it would use. How would it work? How in these different areas that I'd mentioned, we budgeting, or build payment, and so on. What sort of logic would it use to either make its own decisions or to make recommendations for you? And what we found at the time is that at the time we didn't have the user interface bit, we are now a long way, particularly with large language models, like chat GVT, a long way of having a technology that can basically have very sensible conversations with people in almost real time. We don't have it with voice at the moment in real time, but I'm sure we'll have that in the much-adjusting future. But so let's say that that's a built. And I'll voice one of my favorite egg notes at what time, but in the US, was that the Amazon Alexa device in my kitchen was very smart in detecting that as an Australian X-stat, would it use the US that I would occasionally want to order at VGM or even the Arnold Sigeumwha play that shows, and it would sometimes prompt me, it's say you have an old VGMwh for a while, then you might need to add that to your share and go. And I would say no, because I actually I hold it some. And it wouldn't actually be able to recognize my Australian accent. That was so incredibly astute in order, I did define what an Australian X-stat might want to buy. But the moment I sit down, banks would go, "Yep, that's all fine." Added that to your cart for you. (laughing) - So the interfaces are working very well. - So we're all those there, but I'm quite. But anyway, I would say, let's sort of say we're a long way to solving the interface problem. But then, if you look at the logic problem, how would it actually make decisions or even recommendations? That's I think where a lot of the work needs to be done. And let's start with something really simple that you mentioned earlier, making simple payments, for example, in relation to parking. So the scenario would be, this wasn't part of our vision at the time, we had something similar, sort of autonomous build payments. So the build comes into your system, it automatically pays and just lets you know, "By the way, I've built the paid electricity bills." Could be the same with parking, right? Your app, geolocation, you're now in this sort of parking area, might ask you how long do you think you're gonna be here? Maybe it already knows. - Yeah. - And then we'll just take care of the transaction. So this is something that we could probably implement as a prototype very quickly, because if you think the online parking apps, they already have a lot of the infrastructure in place, we just need to add a little bit of tech over that. So that could probably be implemented relatively quickly. But then there are other things for you know, be it investment decisions. Let alone help with retirement savings. Where I don't think we couldn't even spell out conceptually how what code we would write in order to make decisions or even good recommendations for people. Why? Because we don't have an industry standard. How to advise someone how to go about their retirement saving. Right? If you ask 10 advisors and maybe actually, this might be a bold claim, you would probably get 10 very different answers. And so then the question is, well, what would that act do? And there's a whole raft of hollow on issues to do with acting in people's best interests, acting responsibly. And so on that we can't address until we've got solved the initial issue in terms of what would the logic be? I know of that for the Dishurid GD emission, but you get a huge financial cost that's quite a big, big regulated space. And one that has deep regulatory complications in each jurisdiction, that's right. As opposed to some of the one of the mental use cases that would be universal across the world. But one thing I want to figure out that I think you, in light and go there is, I had tended to think the real paradigm shift the Gen A and I come on them and it's been so dominant over the last two years. But the real paradigm shift was that was placing AI in the hands that the average citizen, the consumer, that every previous application of AI that we talked about and always need AI being deployed by the court. Or it quiet down the agents. But now some of these is AI and hands down. Now, I like the fact that you picked up that our lens have really been driving this notion of making the user of the post-murics, but that perhaps that's the beach. That the signal must support the driver or the cabalizing one that really has had that effect. So I think the agenda, or that it is actually AI the hands of the people now. - That's right except the Gen AIs that are available are also run by culprits. So they, to overuse the word, they do empower people to some extent and that they make a technology available that can be used. be incredibly useful and I use it, not for example, encoding or lots of other tasks where they are excellent. And I can definitely see how it could be used in many other important areas of life, but the issue is that it's not owned and run by the consumer. It's still owned and run by a corporate entity. And so if you wanted to use it with finance, then there would have to be some financial institution connected to it. And so I think that it comes back to your question or to the point who would build and operate those kinds of apps. So there are some obvious candidates. In finance, it could be banks, but there's an actually obvious conflict of interest, which products would that app recommend. And how would it use people's information and so on? Could be governments, of course, but then there's probably a privacy issue when I want to share all of my financial transactions with the government. Could be a big tech platform. That was probably my instinct. I would expect the likes of the firm like result or Apple to be deploying an app that even what they need. That easy powered by IOI makes this advice and potentially execution transaction work. There was also a start IOI app slot. It is a new initiative in this place, which I think is pretty catered and we use it with a walk-chart. So you probably see some of those new returns. But I was thinking probably the big techs would be at least likely providers. Well, I know from what I know, I think some of them at least have been trying to move into this space, perhaps not surprisingly. It's probably not used target, but we haven't seen as much maybe inroads as I certainly would have expected a few years ago. My expectation was that Apple for example would move into the personal finance space very twitching that haven't really. And then if anything, they've been retreating a little bit with, for example, their credit card product. And the question is why was it really consumer uptake or other issues? One issue with finance is assignment of responsibilities. So as you know, finance is often high stakes and therefore high risk for the operator. So imagine you had your upward and app that was authorized by the customer to make certain kinds of payments on the customer's behalf. And there's a coding error at the app mistakenly transfers $10,000 to the wrong back account. Who's responsible? And I'll scale that up. It gets hacked. Or there's another coding area and it transfers mistakenly $10 billion, who's responsible. And so, you know, this is sort of me speaking rather naïve because I've never operated a product or a platform like that. But I just, I think, you know, there's one thing is about how do we, how do we build the core functionality? But then there's this other question, what, what do we do if something goes wrong? And A, it was responsible. And then if you do have to take responsibility, how do I manage that risk? And it might just be that at this point, there is a, there might just be a lot of hesitancy apart from the regulatory issues. Which by the way, I think is also an issue in the, in the health. So I want to go with this thing, a lot of while we're around some of the challenges, it's quite a bit of a flash. So it's all pivoted among it. So it's a bit of that confusion, which is another subject on a year of the very year. But I'm not these challenges and implications. I think a bit of it for the perspective as a bank, or we will at some stage have customers who will have their own A or agents, and they will want their A or agent to be able to do this. And we're going to be to firstly have the authentication capabilities to work out, for instance, that this is actually a finance box. And this box we can see, let's try to trade that on custom. But then secondly, you probably give your, your box, your agents some delegation, what things are, but it's allowed to do it for you. And that kind of, perhaps starts to sound a little bit more apparent, certainly, never apparent, reflecting that there is a potential duty or expectation of these writers acting on the art and pure best interests. We're going to need to be able to authenticate that it's actually behaving in line with the delegations that you're pibidant. And there perhaps you also get into, you know, if you've all this making further situations, it's maybe going to be less than now that by advertising, for instance, and all why reliability, ABI studied as well as price, obviously. And they need it also that comes as systems to be a little question if an AI agent is able to, for instance, work money faster than we ever see today. Or even if they could be AI agents working for different people, follow a very similar algorithm and you wake up with an highly correlated movement within the economy. What is the sort of challenges or implications that you would have in mind? There are some of the challenges. So I think I, I, I, another big, big challenge, obviously, is the whole issue of how to get the app or the AI to act in the customer's best interest. Right. So again, with things like paying, paying for parking or bill payment, some of the, some of those more basic functionalities, maybe that's not so much of an issue, but again, you know, think moving on to things like investment decisions, even just savings decisions. There is a question around how, how do, how will the app know what the goals and preferences and financial circumstances of a customer are? How are you going to elicit that? And you and I, we might be able to specify that probably to a reasonably extent to an AI, but a lot of people may not be even be able to state their goals, let alone detailed their financial circumstances. Right. So how do we elicit that information that we need in order to make good recommendations? And then how do we build a technology that then computes? If you like, good recommendations based on, based on that information. And as I said earlier, that that's also a no brainer. And so the, the, there are some keywords here. Obviously, these are big issues in AI in general, the alignment problem. How do you align an AI with its human user? How do you make a eyes? What's your Russell calls human compatible? And as I said, there are conceptual questions around, for example, how do you elicit peaking goals and preferences? Are there certain types of preferences that people have that digital system? Carpe, you would elicit it all. And then the question, when can an AI make a good decision on behalf of a person without the person in the loop? Or how do you decide when to bring the person into the loop to make the decision themselves? Right. And so there are these conceptual challenges that, of course, you know, people are working on it again. There are certain areas where that these challenges are not so pertinent in others. They are where they are more pertinent. And so I think a related issue is that finance or financial, particularly financial advice is what we call it credence good. And so that's a good or a service where a customer can't tell either before choosing the service or other after having received the service, what the quality of that year. So the standard standard case that's often given by academics are healthcare decisions. For example, the quality of a doctor or the quality of a treatment. And so as a patient, I can't tell the quality of a doctor. I might feel good after I got some treatment, but it's still, you know, in the long run, it might turn out to have been not so great. And so I certainly can't tell the quality of a, and we learn it to kind of actual. You don't know what how we're going. Yeah. And the expertise and look, that's right. And so in healthcare, what we do is, you know, again, standards and regulation. Everyone's doctors, doctors are trained to certain standards. And there's a lot of quality control. I want to finance a lot of financial services are treating good as well. While we do have a lot of regulation, I'm not sure we have the similar type of regulation about the quality of individual pieces of advice or particular services. And again, I believe that we do need to move into that direction. If we want tech, including AI to succeed, because we do need these kinds of reference points. How to decide, for example, the effort regulator, when was it a service good or the different properly, when was it not do that properly? Right. So I think with a roll out of technology, it's, as you know, developing a prototype is one thing. But then if you want to market it, you do need to think about the edge cases. If you like, what if something goes wrong? What if I hit you? And I think a lot of these like edge cases, having been solved. So I think that's another challenge. But I think the way to go about those is as you suggested earlier, you start with the easy, easy ones and learned. And then you build the next country complex thing and you learn. And so on and you know, step after step, you build this ecosystem and the ecosystem that I would like to have built is one that goes way beyond just managing finances. So think about the internet of things, no, and your fridge and your pantry, ordering as maybe Alexa did a perfect But ordering groceries for you, I mean how often do you open the fridge? This happens and don't judge me in our household quite a bit. You know, we want to have eggs for breakfast and all we're out of eggs. I would love to have a fridge that realises that we are about to run out of eggs and then puts an order in and automatically pays, then it's just get delivered. And so I think we need to think a little bit beyond finance about how, you know, this technology could be used to automate just a lot of other areas in our lives. And finance is that always the flip side, right? Every transaction in the real world has a counter or another transaction that's a financial transaction. So building this financial automated financial ecosystem would then unlock a lot of potential, I think, for automata. Well, let's sort of go up a little bit. People will like transactions or have their AI agents making transactions. Consumers purchasing decisions. And as a bank, we'll make a support that. We'll need to be able to work out how we authenticate and validate how we, whether it can share as bought is dealing with the merchants bought. The bank will need to be able to authenticate both the ends and actually settle the transaction. I think probably where a bit of where you'd clarified for me and the discussion is, there's probably a distinction between investing in a purchasing and the investing buy-states are is complicating whether it's acutement making the decision or an AI. Whereas my AI agent can probably optimize for me in a lot of my purchasing. You know, my real world scenario this week was that I needed to want some more barbecue cleaning wipes, whether I'm better or ignoring those involved, and whether I'm getting the best price from barbecue school or Amazon. Yeah. How quickly will get delivered and whether I will pay for expedited, deliberate support like an AI agent can probably just work all of the AI for me, without me spending violence toppling between you know, browsing with those that I find or work it out with. That's right. Let's pivot to quantum, which is one of the other big, murdy things that we're covering on that digital mix. And we haven't built on that furthering our next episode, but I know it's a supposed you've been researching and I wanted to tap your mind on it also. Interesting, firstly, in how you're saying the emergence of this technology, and perhaps what's most vertical in your wider report. So I definitely committed either as a decision scientist or perhaps a finance person. So my perspectives are very much, you know, from that vantage point, if you like. And you know, at the most general level, quantum is a completely new computing technology. It's not like neural networks or machine learning, which sort of runs on existing technology. And it's a completely different technology that physically works very differently. As you know, exploiting properties of quantum systems like superposition and entanglement. And the hope is, and I suppose to some extent, the reality already is that it'll speed up. At least certain kinds of computations substantially. And that therefore, or that that's something that makes it very attractive, particularly in areas like finance where speed computing speed is 10 of the essence. But for example, one application in the area that we just talked about using AI in consumer finance would be human computer interaction. So we said, you know, we want an AI that talks to people in real time, extra commendations in real time. That's very computation intensive. And if we could get a technology that halves or reduces compute time if therefore delivery of the product or recommendations, substantially, that would be quite attractive. But then of course, there are other areas that come to mind like trading where a microsecond can be a big strategic advantage. And so on. So the hope is that many areas in in in finance can benefit from the speed up. Now it's of course still relatively early days. And lot of the work is on hardware development and different people I suppose have different use on when production, you know, system that's going to help in with real-world applications, it's going to be available whether that's five years or 10 years. We don't know. But let's suppose it's going to happen. I mean, the systems we have, they are already you know, they have grown quite a bit over the last few years and they're sort of a trajectory to which they call the quantum road map to having production type systems in the not too distant future. But a lot of other work at the moment is really conceptual work trying to understand how to use quantum computing in different areas of application, for example, in finance. Because the computing technology and the logic is if users is very different from classical computing, so during computing, we have to rethink, you know, how we solve problems like derivatives pricing, but only optimization, various risk management problems using quantum computing. And there are teams that have been working on that around the world at the big financial firms, for example, with the Japanese Warman and Goldman Sachs, some of their hedge funds, but in their often collaboration with universities, trying to work that out. And that's where we are at. Where we are at. There's probably, I'm sorry, because there's sort of, there's the constructive it was that effects you've sought. That's the, there's the effects you've sought around. That's right to be friction, but the financial also makes a lot more tricky communications. And I think that's probably where we're saying very heavily heavy levels far and deep from US and Chinese governments, particularly following it. And there'll be constructive side. There's the opportunities in finance for better risk modeling and optimization, if you mentioned, as well as things like eDrive discovery and an agriculture and the way that the nitrogen molecule can attach in fertilizer and so forth. So there's this mix of the effects, signals of the friction standards as well as some of these constructive new opportunities. That's right. And they are related. So if one firm, for example, was successful on the opportunistic side, for example, developing a substantially faster trading algorithm or a tool that can effectively reverse code other firm's trading algorithms, that's also a potential threat. Because if you have automated your trading, and suddenly someone can reverse code your automated or your trading algorithms and then exploit it, that economically just as big a risk as someone being able to decrypt some of the encryption algorithms that you might be using. So you're right. A question, how can we use it to make our existing operations better, to be faster and so on, but also the threat perspective. How do we do it in order to prevent us being exploited by someone else, whereas access to this technology and also how to use that technology? So let me keep talking, include on that point, how you prepare and I'd say you build capability. One thing I often hear is that there's a room-short ball even at sound while at night. That's not the volume of quantum computing scientists necessarily coming through academia or through education yet, and also that you can't just converse a retrain, a bicycle computer scientist, in building a quantum computer scientist necessarily. So there is a fear, I think, that there is a talent shortfall that we've been digitally manipulated. Interesting, whether you think that's a fair assessment, and as if there's anything you'd recommend that businesses both ourselves in finance, but also our court requirements, things that businesses should be abstain from you about on that, and that would appear in the results. That's right. So I think one reason we might have a shortfall in talent or trained, say, graduates in this space is that there's probably also a shortfall of academics who could teach quantum computing, and one challenge with regards to teaching is not just is that you want to teach not just the theory, but you also want to be able to actually get people to code or to program a quantum computer. That requires access to a quantum computing system. Very few universities have that, or they don't have the staff, you know, how to actually operate it. Now if I am allowed a little plug, please say University of Melbourne has been offering. I have those things built in the ground. Logs on flight and partner initial strategy checks. So please continue. It has been offering quantum computing subjects, both at undergrad and post-grad level. And one reason we have been doing is I've been able to do it is because we do have at the university, or have had a team that's very strong in quantum computing and works very closely with IBM that gives the team, but also this our students access to actual quantum computers so that they can also get the practically experience of developing circuits and doing the actual computation. So there's probably a shortage in the training education space that we need to expand out. So we need to grow capacity in that space. And I'm sure that will happen. Then, and then where do we go from there? How do we grow the capacity, for example, with people who already work in industry? And I think one question is, would as a map, for example, as a bank, would you rather want to do that in a house, or would you rather buy it in? And I know that in the tech space, or maybe in other areas as well, there's often, and maybe in Australia, more than other countries, there be you that, why would we do it ourselves? That's just, let someone else do it and buy it. My point with quantum is that, A, you're not gonna get the best out of the technology. Again, you're gonna, you need people who understand your business inside out, and he can then translate it into a quantum computing problem in order to solve it. And I, ideally, you want someone who really understands your business needs and understands quantum computing. And I'm not sure, you know, that's something that you can easily purchase, this is what we do, I believe. - That's right. - That's right. And the other, the other big issue, of course, is dependency. Right, so if you buy everything else, everything in, you're gonna pay a premium, probably quite hefty in a space like that, where there's a challenge for it, and you make yourself dependent in an area computing, that is absolutely mission critical to everything you do, and will only become more mission critical as we go forward, right? So there's no banking anymore without computing. And so would you really want to outsource your core expertise in that core capability to a third party, probably not? And so I think it is important for Australia and Australian industry to grow a capability in that space, working with universities and others. And I think we are already behind, so we do need to speed up, but it's a priority of the federal government, it's priority for our university, certainly, and so hopefully we'll be at the forehunt. - Probably for us here at MAP, so we're very much committed to that upstill in journey with we've started pulling in a number of that, through our tiny programming, so that's one we'll need to develop that bench track. - Kielsen, thank you. I'm gonna try to catch your appeasings that I think most resonated for me at least from your comments. Firstly, I like the way you started in explaining the brain processes, as you've been exploring them in the centre of them, those two major things you mentioned, the fact that we never have enough information and dealing to probabilistic uncertainty, as well as often that we just don't have the submission, permitting for a stall since. And it really opens the door, and the sort of space we've been talking about here of the extent to which AI is going to be a bad health or walk mentaic. Great discussion, I think you're capitalised for us there around where the sort of scenarios, what are the where we can best get to, what would be most useful. We also did in that touch on the theme of trust and how we will over time, perhaps we'll take time to develop trust with our AI agents and what we are accountable with and do it. What the fact you made a point about needing to have a user interface that is easy and that makes it accessible, add a particular role that LL needs to play, that capitalised environment over the last two years, with Gen AI, which led us into talking about who will build and operate the apps and a few different scenarios we talked about there, but looking at back to the assignment of responsibility that is such a high stakes issue, the particular refinance, that really up it goes to the distinction between perhaps some of the investment scenarios and the purchasing uses. So it creates that shot through, well I think it's going to be what the really big, emerging spaces, probably in the cause of the next two to five years, plus the thank you for sharing those insights. No thanks so much for having me. Now coming up on that digital next, we're going to continue with the big future trends and specifically with more aquatic computing. I'm going to be joined next by my friend Steve Sware as the former Chief Innovation Officer of HSBC, now leading Quantum Education in Part of the Shigwit MIT. And also in supporting climate transition, we're going to check in on a very exciting energy, efficient air filtration solution that's being developed by my nabs and liacustins. So please join us again for those. Thanks for being with us on that digital next. (upbeat music)