Main content area

Watch an in-depth panel discussion on AI potential in Australian manufacturing.

Moderator: Mark Pesce

Panel:

  • Dr Jon Whittle, Director, Data61, CSIRO
  • David Chuter, Chief Executive Officer and Managing Director, Innovative Manufacturing CRC (IMCRC)
  • Dr Sue Keay, Chief Executive Officer, Queensland AI Hub
  • Aude Vignelles, Chief Technology Officer, Australian Space Agency

Transcript

(MUSIC PLAYS)

Description: Against a white background there is teal-colour text on the left.

Text: TECHTONIC 2.0. Australia’s National AI Summit. 18 June 2021. This session will commence shortly.

In the bottom right, still colour images of industry. A man in a hard-hat. A woman in a field controlling a drone.

In the top left, the Australian Government logo.

Text: TECHTONIC 2. Panel: AI Applications in Manufacturing. Dr. Jon Whittle, Data61. Dr. Sue Keay, Queensland AI. David Chuter, Innovative Manufacturing CRC. Aude Vignelles, Australian Space Agency.

In the top right is the Australian Government logo.

In the top left, the text reads, ‘TECHTONIC 2.0. Australia’s National AI Summit.’

Mark Pesce speaks as an insert in the screen. His caption reads: Mark Pesce. Leading Futurist, Author, Entrepreneur and innovator.

On the bottom, live captioning appears.

Mark Pesce: Welcome back to Techtonic 2.0, the National AI Summit. We are here in the opening panel of the day. So, we have heard a lot of plans from the minister, but now let's start to dig down to what it all means in practice. What does the intersection of artificial intelligence and manufacturing actually look like? What are the applications? Why is this so important to Australia's future? And I have to say, we've put together quite possibly the best panel you could put together in Australia to answer these questions. And remember, you get to ask your own questions to the panel. You have that Q&A section on the right side of your screen when you're watching this. So, you can pose questions, you can upvote other questions.

Alright. Rather than introducing all of the panellists at one time, what I'm going to do is I'm going to introduce each of them and ask them each to share, from their own experience, an example of AI in manufacturing, cause all of them have such broadly different experience that we'll see that AI in manufacturing is not simply one thing. It is a whole new galaxy. It is waiting to be explored. So, let's begin with Dr Jon Whittle. Jon is the Director of Data61 Division of CSIRO. Jon, good afternoon to you.

Description: Alongside Mark Pesce, the webcam vision of the four panellists appears as small squares in two rows.

Jon Whittle: Hi, Mark.

Mark Pesce: Alright. Can you share with us something that you've seen at this intersection of artificial intelligence and manufacturing?

Description: Jon Whittle, a man wearing glasses, appear via webcam. His video is on the top row, to the right of Mark Pesce’s. Behind Jon is white background with the CSIRO and Data61 logos in bright light blue.

Jon Whittle: Sure, absolutely, Mark. So yeah, I guess, probably the first thing to say is that as you've kind of hinted that, AI for manufacturing is a very broad area. And that's largely because, you know, not only is AI a broad area, but also manufacturing as it's interpreted in the modern era is also a very broad area. And I think manufacturing nowadays, it's not just about what we might immediately think of when we hear the word manufacturing, which is products going through an assembly line, but there's a lot more to manufacturing than that. You know, you've got various processes in pre-production, production, and also post-production.

So, if you look at pre-production, you know, you've got the design elements of manufacturing, but how do you optimise processes? Then you come to production itself and you've got quality control, you've got warehouse efficiency, inventory management, equipment maintenance, workplace safety, and even cybersecurity factories, cause some factories can be considered as critical infrastructure. And then even in post-production, you've got the whole supply chains and how do you collect data from end-users to improve products? So, manufacturing is really, really broad. And I would say that AI can change every single aspect of that manufacturing process, whether it's that pre-production, whether it's production or whether it's post-production.

So, as you said, Mark, I think it's hard to give one example that kind of clearly illustrates what AI manufacturing is. But certainly, when I think of the examples that I've seen, the things that excite me are where we've got AI algorithms and human workers working together. And I can give just maybe just quickly a few examples from some of the work that we're doing at CSIRO that illustrates this. So, in our Victoria Clayton facility, we've got something called the Mixed Reality Lab. And this is using augmented reality to allow engineers to better detect defects in products.

So, what they're doing is they're using off the shelf camera systems and then taking those camera images of an object that's been manufactured and then using virtual 3D technology to superimpose on top of that a representation of what the product should look like in an ideal world. And that enables an engineer to more quickly and efficiently see defects that may be present in the object. So, it's a really good example of where you've got... It's not just about the AI detecting the defect, it's also about the human being working in concert with the AI. And on a similar vein to that, we've also worked with a major car component manufacturer to produce an analytics dashboard for defect detection. And that's been shown to reduce testing defects in car parts by up to 50%.

So, again, it wasn't about that machine learning doing the job. It was about machine learning producing that analytics dashboard that the human being could then interpret. But there's lots of other examples as well, we've done work within Data61 on applying computer vision within factory environments, for example. So, some of the work we've done is about how do you optimise the way that robots move around the factory floor. And typically, these robots might have onboard cameras, but that only allows them to see what's in front of them. It doesn't allow them to see what's around the corner.

So, if you combine that with computer vision from fixed cameras around the factory floor, you can then get an intelligent system that can optimise the way that the robots move around the factory and make things more efficient. And even in things like workplace safety, again, if you've got humans still on the factory floor that are still needed for some tasks that might be more intricate or perhaps more complex, there's workplace safety issues. And we've got some work where we've been using computer vision to automatically detect worker posture, for example, to help them avoid injuries or even monitoring them automatically for signs of fatigue in workers, which, as we know, is a major cause of accidents in the factory environment.

So, I think probably the summary of all of that is that there isn't one example that sums up what AI and manufacturing means. It's a whole range of things. And that's the challenge in many ways, is how do we integrate all of that together? And certainly, for companies trying to adopt these things, where do they start? And once they do start, how do they integrate all the different moving pieces?

Mark Pesce: OK. So, I mean, you're really reinforcing this idea that this is a vast landscape that we're really only just starting to explore. Thank you so much, Jon. Next, we're going to turn to Dr Sue Keay, Chief Executive Officer of the Queensland AI Hub. Good afternoon, Sue.

Description: Dr. Sue Keay is a woman has shoulder-length blonde hair. Her webcam is on the far right of the bottom row of panellists. She is in front of a black background On the top right of her screen is the Australia flag. On the top left is the text in block font, ‘ROBOTICS.’

Sue Keay: Hi, Mark.

Mark Pesce: So, what have you seen, what can you tell us about the future of manufacturing?

Sue Keay: I think manufacturing in Australia, it's really going to rely very heavily on artificial intelligence because of the nature of the type of manufacturing that we do here. In general, Australia focuses on niche manufacturing. So, it's always low volume, often highly customised production, which really lends itself to the application of artificial intelligence because of the sheer volume of data that's sometimes necessary to be able to control that process.

We're never going to be a country that's doing high volume manufacturing. We don't have electrical component manufacturing as an important mix, part of the mix. And so, I think this is going to become increasingly important. It's not just because of the complexity of the type of manufacturing that we do, but also because in many areas, and I've just come back from Central Queensland, we're facing real labour shortages, not just in our agricultural sector, but in our manufacturing sector. So, one of the local manufacturers up there employs 60 people, they have 20 open vacancies.

And so, they're actively looking at how they can apply artificial intelligence to streamline their processes and make them more efficient so that they can overcome that unfortunate lack of supply in terms of the labour. But I think, in terms of additive manufacturing, I think artificial intelligence just makes sense. There's a number of different complex factors at play, and whether that be for plastics, composites or metals. And you really just have to apply AI to ensure that accuracy and quality of the manufactured parts.

So, some of the complexities that you get inform material properties and integrity, particularly in metal parts produced by 3D printing, can all be related to a lot of the complexities of using lasers in this process. So, the parts will vary depending on the power of the laser, the speed, the spot size, how much overlap there is between passes of the laser, what the exact powder composition is, how much contamination there might be in the powder. And all of these things, if you set them up with high-speed cameras, which are following exactly what the laser is doing, if you apply artificial intelligence, then you can actually start to change the process of the 3D printing while you're making the part so that you can reduce the possibility that you're going to be producing low-quality parts.

So, in a process like additive manufacturing, AI is really just an essential part of that process. And if we don't use it, then you have a lack of reproducibility, which means we can't use these types of manufacturing techniques for developing critical parts, like for example, those used in aircraft. So, one example is a local manufacturer using artificial intelligence for its automatic fibre placement process. They found that it actually improves the prediction of their defects, which normally they would detect by physics modelling.

So, their typical defect detection rate would have been 35%. But now that they're applying artificial intelligence, that defect detection rate is 98%. They've also found that artificial intelligence is very useful in terms of informing them the best way that they can inject and cook the materials that they're using. And they've also seen a 60% improvement in cycle time and the performance of the product during the curing process, simply by monitoring the temperature profiles and combining this with modelling that they've been doing using synthetic data. The same manufacturer is also starting to experiment with how this can be used to their 3D metal printing process. As I mentioned, there's a whole heap of data that you have to consider when you're looking at the applications of lasers, particularly powders. And they're already seeing some improvements in that metal defect prediction, which is a far less stable process than fibre placement.

So, I mean, I think artificial intelligence just makes sense. And particularly for a country like Australia, where we do this niche manufacturing, the most important thing is that we start to get people on board with how we can apply artificial intelligence and make it more accessible to the workers in these industries.

Mark Pesce: Wow. And, Sue, you're really pointing to something that Jon was also pointing out, which is that AI helps humans work better, right. That it's not an us versus them, but it is, in fact, it improves our ability to get work done. And I'm interested in what you said about this body shortage, you know, is one of the accidental outcomes of having really long border closure going to be that we have to lean into AI to deal with the fact that we don't have people coming into the country in the same way.

Sue Keay: Yes. Well, I think it's fair to say that people have been able to keep their labour costs low by importing labour and that has not been possible during COVID. But I think, there is also, I don't think it's necessarily a trade-off in as much as what we've also seen when companies are becoming more efficient and profitable by using artificial intelligence, then often they will look to employ more people. So, I wouldn't say it's a complete, it's a one-to-one trade-off between putting the technology in place and necessarily replacing people, but in situations where you do have labour shortages, it can be a stop-gap measure.

Mark Pesce: Right. And it takes some of the pressure off. Alright. Now, we've got to turn to David Chuter. David is the Chief Executive Officer and Managing Director of the Innovative Manufacturing, CRC. Good afternoon, David.

Description: David Chuter is a bald man. His webcam square is the far left of the bottom row. On his right is the humanoid head of robot integrated in to a block aqua colour. To his left is the black, bold text, ‘we chamption manufacturing innovation.

David Chuter: Thanks, Mark. Hello, everybody. Mark's asked us to pick a success story that we know about, and to Jon's point, this is a really hard thing to do given how broad manufacturing is. I always look at the value proposition. It's easy to get hung up on the technology itself. But arguably, it isn't just about the technology, it's about the value you create and it's about the value with AI, initially, about how you can capture value inside organisation. Recent studies have revealed that the most popular AI use cases in manufacturing are improving.

Although primarily, this is about capturing value inside factories or businesses, and is not yet really about that value creation outside in terms of across supply chain sectors, or to create new business models. But that's where the exciting thing is going to come in the future. So, it's where are the best examples. Well, not surprisingly, these are generally within factories, and also not surprisingly, many are led by the automotive industry around the world as that industry has led many of the recent transformation waves in manufacturing.

Where companies are taking this up is probably a third of companies are taking up AI in the form of intelligent maintenance. OK. So, using AI, organisations can predict, prepare for failures in equipment, reducing or even avoiding downtime, and using a lot of camera systems to be able to determine, for example, whether robots are starting to move slightly out of their patterns of behaviour. Locally, we've seen companies like Movus, who remotely sense equipment vibration and they monitor and predict abnormal behaviour, and they can advise those companies, which is a significant cost saving, particularly if you can't get another motor from within Australia and you're constrained by supply.

So, COVID led to quite a bit of investment into getting much smarter about how you can manage maintenance activities. Another big area of investment is in product or process quality inspection or assurance. And Jon mentioned some of the work that CSIRO Data61 are doing in that space. It's really about making sure that you avoid defects in production and how you can pick those up. But it's not just whether you've made a defect, it's also starting to predict things like whether your tools are degrading, whether your process is going out of parameters. And so, you have companies like Audi and General Motors using cameras actually inside tools.

So, taking very high-resolution images of aluminium or steel body panels being pressed, looking for micro-cracks that humans simply wouldn't see. And what they're looking for is early indication of something that could cause a fairly big downtime in the industry thereafter. Another one's in demand planning, looking at historical sales data, weather patterns, third party data, social media, macroeconomic. We're not seeing that trend in manufacturing yet. But you only have to look at people like Domino's Pizza, for example, where they are able to predict in advance when pizza demand is going up or down. They know when we're going to order a pizza before we do, and even most likely the type of pizza, which is pretty amazing.

So, imagine that opportunity in manufacturing. The popularity in manufacturing is because there is an inordinate amount of data in manufacturing. It's very complicated. You've got lots of suppliers, lots of raw materials. And so, it's full of data that can be easily analysed. And the point of this is that, you can do that not just in the production environment. But where the exciting stuff happens to me is when you look at it from a design perspective. Artificial intelligence is now being used for things like generative design. So, this is letting a computer loose on how to design a new component and not being constrained by typical design rules.

Sue talked about 3D printing, additive manufacturing, couple that up with generative design and you have the opportunity to design freedom, multiple iterations and then to print it to test it. If you embed sensors in the product that you put out to consumers, you can then track it through a digital twin. And all of a sudden, you get a real connection between products that are made, what consumers are using, live digital twins. And all of that intelligence can be used to recreate and make better products, almost down to a single component level. So, sorry, there's not one example I've got. But it's an exciting space. And there's just about every step from design and engineering through production, logistics, even right through to how the consumers are using products and services, the end of lives, how things can be repurposed. And there are enormous benefits to be gained across the platform.

Mark Pesce: And we actually saw this last week. Google announced that they designed a new AI chip using AI, right? That they, in fact, were able to do the chip layout in a way that a human being couldn't do. And so, we are actually now seeing that becoming folded into the design techniques that we're using. Thank you very much, David. Finally, Aude Vignelles is the Chief Technology Officer of the very newly created Australian Space Agency. Now, Aude, you have undoubtedly seen some interesting intersections between AI manufacturing and this new space race that we seem to be having. It's not just the billionaires, right? Everyone is going and playing in space these days. So, tell us what you see.

Description: Aude Vignelles has her hair pulled back. Her webcam square is in the middle of the bottom row. Behind her is the logo of the Australian Space Agency. On the top right is the Australian Government logo.

Aude Vignelles: Indeed. Thanks and good afternoon, everyone. I have to say, it's really interesting to hear all of you speaking and also hearing that we, the government, is gonna invest in developing in AI. And especially help supporting the talent growth. Because we, as you know, the goal of the Space Agency is to create 20,000 new jobs in the next ten years. And a lot of them are gonna be in AI and application in space. So, it's good to plan the future. I need to echo Jon, Sue, David, everything you say is applicable to space.

When I hear robots and remote operation from you, Jon, of course, when you're thinking about going to the Moon and further to Mars. We're gonna need the remote operation and robotics to prepare all the infrastructure before sending our women and men there. So, a big application there. When I hear you, Sue, talking about defect and performance. You know that if something breaks in space, you can't really send someone to fix it soon. So, having this AI technology to predict what's gonna fail and to fix it before it fails, is very important. So, I really liked what you mentioned about defects and performance.

And, David, hearing digital twins, that's also something that is being used more and more. Before sending a spacecraft in the old days, in the old space where I come from, you used to start doing a prototype model and then an engineering model and then you would build your flight model, and you will do a lot of tests on these different models. Now you have a digital twin, then you can modelise everything using AI, using machine learning. And you can almost skip all these stages and go directly to your flight model.

Now, two examples. And I agree with you, David, it's gonna be hard just to pick two examples, because there are so many. First, in advanced manufacturing, until now when you were designing a piece of equipment, especially in space, it's such a harsh environment that the constraint, the vacuum and thermal high temperature and low temperature are really harsh. You were limited into the tools in your disposition to build what you were going to design. So, especially when you think about tanks having a perfect sphere was a really nice thing to have. But how do you manufacture that? With advanced manufacturing, all these constraints disappear.

So, the creativity in designing components, systems, subsystems, is completely new. And it's a great field to be in just now, cause advanced manufacturing is opening a door to a lot of different design and pieces. In terms of AI, I need to mention that we have a couple of companies, organisation and research centre in Australia in AI who are very, very good. And they've found some application in space that they were not expecting to find. And one example is, as you know, our orbits are starting to be very, very congested. Before we were concentrating on Geo. So, 36,000 kilometres above the Earth and everybody was there, you don't move, you always in the same place.

Now, with this constellation in reverse orbits all between 800 and 1,200 kilometres, this is where you need to have a lot of satellites if you want to have a full coverage of the planet. So, we have thousands and thousands of spacecraft now in lowest observation, and that's creating a danger because they can collide. So, it's important to know where these objects are in any point in time, but it's also important to predict when they're gonna be. And space is an interesting environment. You have the solar pressure that you need to take into account.

So, you need to modelise all this. And AI is actually used to predict a collision and to try to advise on a manoeuvre to do to avoid collision. And the European Space Agency a while ago did a competition to say, who has some AI example to use to solve this problem of predicting collision? And it's actually the AI and Machine Learning Centre in Adelaide who won that competition. So, all that to say that, first of all, space is a new market for anyone doing any technology. I think any technology on Earth has an application in space.

And AI in advanced manufacturing has a great role to play. And the last thing I would say, if I look into the future, some people are telling me, you know, what we're gonna do in the future is probably just send cartridges into space because we're gonna have 3D printer in orbit or on the Moon or on Mars. And you just manufacture whatever you want and design whatever you want. You just have to send the cartridges. And why not? So, that's what I love about this job, is you can be a futurist and God knows what tomorrow is gonna look like. But there is no limit to our ambition and our visions.

Mark Pesce: Anyone who's had to fight with a printer cartridge is now living in fear because if that future... (LAUGHS)

Aude Vignelles: We'll work on it.

Mark Pesce: We have some great examples from the panellists of AI in manufacturing and all of the different ways that it will touch the work. And we've heard from the minister, there's policy, there is plan. So, let's start to connect a few of the dots here. And so, we have some questions. First one, I wanna address to Jon. After that, any of you who wanna respond, please feel free.

So, Jon, we have heard from the minister that there's money in the budget, I think it's $1.5 billion for the new manufacturing, around 150, 125 million for AI, specifically. What kinds of investments do we need to be looking at, at this intersection between AI and manufacturing? And for how long and how patient are we going to need to be before we can expect to see some results?

Jon Whittle: Yeah, good question. And before I answer that, I was getting nostalgia listening to Aude talk about space. So, I used to work at NASA back in the early 2000s. And in fact, I worked with a lot of the groups that were doing very early AI for space. So, hearing some of the applications that Aude was talking about then, I remember back then there was a really interesting application called Livingstone, which was essentially an AI model for fault detection and maintenance, that was actually applied on an actual deep space probe.

So, it basically mapped the state of the spaceship and compared that to what the expected state was. And then, if it wasn't what I'd expected, it would take corrective action. I think it's always interesting to see that there is these certain technologies that are developed. But one of the challenges we have is often actually getting them into practice because the technology, in the way, is the easy bit. But then putting that into a whole ecosystem and getting them actually used in practice can kind of take a lot longer.

But coming back to your question. Yeah, maybe, I mean, the minister talked about a number of new investments from the government in this era. Maybe I'll just talk about a couple of them that were mentioned that CSIRO is taking a lead on. So, the first of those is this National AI Centre that was announced in the budget a few weeks ago. So, this is a $53 million investment. It really consists of a number of parts. It's the coordinating centre, which is gonna be led by CSIRO, and then four digital capability centres attached to that. And they will each focus on particular themes and there will be a kind of open process for designing those themes.

But, you know, I don't know what the themes will be, but if I was a betting man, I would put my money on saying that there will be at least one of those capability centres focused on AI for manufacturing, certainly seems like an obvious thing to do. But within the coordinating centre, we're gonna be focusing on how can we build that ecosystem nationally for AI more generally, and including AI for manufacturing. And, as well as the ecosystem building activities, we're gonna be looking at various crosscutting themes that come up again and again whenever you try to develop and apply AI.

Things like, you know, how do you develop AI in a responsible and ethical way? How do you ensure that you've got diversity and inclusion in AI? And, you know, what kind of infrastructure and compute power do you need to support AI? So, I think that's a really, you know, that National AI Centre, it's a really exciting development to me. I always say that Australia has got a phenomenal capability in AI. Actually, if you look at various rankings, Australia always does very, very well in those global rankings. But we do tend to have quite a fragmented ecosystem.

And, I think, given the size of the country that we are, we do need to do a better job of just bringing all of that expertise and capability together. And that's certainly the philosophy behind the National AI Centre. One of the other initiatives that the minister mentioned is this Next Generation AI Graduates Program, which is about $25 million investment. And, in fact, there's also another 22 million for a similar program called Next Generation Emerging Technologies Graduates Program. And together, these will fund up to around 500 scholarships in AI and other related technologies. Again, not manufacturing specific, but I expect that there will be lots of opportunities for AI in manufacturing there.

And the plan behind those graduate programs is that we're gonna take a little bit of a different approach. Rather than just funding individual students to do individual projects, we're gonna try and create cohorts of students that can actually work together and work in very close collaboration with industry partners. So, you've got maybe groups of 10-20 students all working on a similar problem, working with an industry sponsor and solving real problems as part of their studies. So, I think, you know, those two initiatives in particular that we're gonna be taking a lead on, but also the other initiatives announced in the budget, I think are a really good platform for us to build on to start to mobilise the very, very strong capability that we've got across the country.

Mark Pesce: Any of the other panellists like to respond? So, let me just reflect for a moment, Jon. It sounds like we're talking about, particularly because we're talking about research. Research is amazing, but slow. So, it sounds like this is something where we're going to need to really decide that we're settling down for a decade or more, both with the investment, but also with the effort, before we really start to see this bear fruit.

Jon Whittle: I'd say yes and no to that. So, first of all, the National AI Centre is not necessarily about research. In fact, it's gonna focus more on business adoption, but more generally, the point is well taken. I think the answer is that there is some stuff that you can do now. And, you know, my colleagues on the panel have given lots of examples of real things that are happening now. But then you've got the other extreme of that, which is the full vision of you know, what we call Industry 4.0, where AI is being used in every part of the manufacturing process. And it's all kind of, it's all integrated, and it's a very holistic approach.

Now, that is gonna take a longer time. And in fact, you know, if you look at other countries across the world that are further along in that journey, they are still quite early in their journey. So, Industry 4.0 as a term was really invented in Germany about ten years ago. But there's a recent report that came out, there was survey of German manufacturing industry, and that said, they were on a maturity model of Industry 4.0. There's really only about 4% of companies that are well progressed along that maturity model. So, I think to do the all singing, all dancing, you know, AI is everywhere approach, will take some time. But, you know, that doesn't stop us taking incremental steps and doing it bit by bit and getting real value out of those in the short term. And that's already happened.

David Chuter: I think, just to echo Jon's comments, there's a recent survey in Germany across about 600 manufacturers. And they're honest enough to say only 8% of them are actually using AI, but the quarter are planning investments in AI. And so, I guess there's a pent-up demand. It's interesting, it depends which part of the manufacturing industry you're from. Because if you have been embracing for many years, lean manufacturing, you have a culture of data analytics, you have a culture of data driven decisions, then AI is low hanging fruit for you and you may get confused by terminology such as separating it as AI. It's just a continued journey of another level and another set of technology and tools that you can apply to better analyse data to make real time decisions.

The reality, however, in Australia, is that lean manufacturing is not embraced wholesale throughout every part of manufacturing. And Industry 4.0 has also been a slow take up in Australia. We don't have data yet on how AI is being taken up, but I think there's some strengths we can play to. Companies that we work with that work with universities, we've got over 50 projects we're co-funding. Many of them, these research projects in Industry 4.0 are embracing machine learning technology during the research. So, the universities are helping to educate the companies on data analytics, machine learning and AI. Another opportunity is to look at things like 5G in Australia, where we would be one of the most advanced nations in the world. Our long distances. What are the things that actually we could be world leaders in, in terms of creation of technology, because of some competitive advantage we already have? Those are the opportunities for us.

Mark Pesce: And you bring up a question that I actually want to ask Sue now, which has to do sort of where our inherent advantages are. So, Sue, we've been hearing a lot about adding AI to industrial processes, using industrial robots, whether we're sticking a camera in them so we can see defects or whatever. Do we make industrial robots here in Australia? And if not, then who are we doing this work for and how can it benefit us here?

Sue Keay: Yeah, no, we make zero industrial robots in Australia, and I find it really frustrating that sometimes we confuse adoption of AI versus creation of AI technology because industrial robots themselves are a form of artificial intelligence. And if you have a look at countries like China, they have been very strategic in their ambitions to actually have sovereign supply of the industrial robots that they then use in manufacturing. So, by last year, China had the ambition to have the world's highest robot population density, which means, that's the number of industrial robots per 10,000 employees, which if you do a back of the envelope calculation, means that they have more industrial robots in China than we have people in Australia.

Now, they haven't succeeded in becoming the world's number one country. Singapore has that title. However, they have secured sovereign supply of industrial robots to support their manufacturing industry. And I think that, you know, you could argue that's not playing to Australia's niche strengths. But I think if COVID has shown us anything, we have to be very careful about our sovereign supply of technologies like this. And I'd probably like to expand the conversation to look at what are the new industries that we really need to be supporting to help us to be applying AI in manufacturing in a way that benefits more than just the individual manufacturer, that can actually allow us to build additional industries here in Australia, and then that we can use to export that technology to other parts of the world. Because certainly, Australia has a lot of strengths in the development and manufacture of service robots, which are distinct from industrial robots.

So, those are robots that you apply in agriculture or in the mining environment, typically outside robots. There's no reason that we couldn't also be good at supplying robots into manufacturing. But at the moment, yes, a lot of our effort is about applying AI capabilities to a technology that's not even developed here in Australia.

Mark Pesce: Alright. So, if we want to then focus this effort around something that will bring benefits to the country, that will lean into our own strengths, does that mean that we need to rethink what we're doing around where we're applying AI? And if you have this vision of there's some low-hanging fruit here, some things that make natural sense. I mean, and obviously, agriculture and mining are two of them. Are there others that we can start to think of?

Sue Keay: Yeah, I think developing our sensing systems. I mean, I think sovereign capability, we hear a lot of this from the defence industry, but I don't think we hear enough of it in our manufacturing industry, arguably, which is supplying into defence. Because if you are having sensors and robots being developed in other countries and then applied here, you always have the risk of Trojan hardwares. So, yeah, hardware Trojans infiltrating the system. So, I think we'd also need to really consider, we invest a lot in cybersecurity in Australia, but we're not really investing in the development of a lot of our sovereign technology capability. And I'd like to see a bit more of a focus on that.

Mark Pesce: Would anyone like to respond, any other panellist?

Jon Whittle: Maybe I'll just make a comment. I'm not disagreeing with anything that Sue has just said, but maybe just a couple of thoughts. I mean, first of all, I think it would be, you know, AI in manufacturing, it's not just about robots. I think a lot of the examples that we've heard about today, it's the full range of what we would consider AI, whether that's machine learning, computer vision, human-machine interaction, and so forth. The other, maybe, point to make is that, you know, there are areas in Australia that we do have incredible strength in parts of AI, maybe not in industrial robots.

But if you look at computer vision as an area, for example, and the AIML Institute at the University of Adelaide is ranked second in the world, according to some rankings for computer vision R&D. And if you look at some of our own work that we do in CSIRO, actually within our robotics team, although we're not building our own robots, we're doing a lot of work putting intelligent software on robots. And we're currently ranked fourth in the world in a DARPA competition using robots to explore underground environments. And in fact, Sue led that work when she was at Data61, so she probably know it better than I. So, I think Sue is right. But let's not forget that we do have strength in other areas as well. So, it's not all bad.

Aude Vignelles: I just wanted to echo Sue's concerns. I have to say, when we are looking into where we want to be in ten years' time in space, we also, and I think COVID has made us realise how dependent we are on a lot of critical technology, AI is one. And I think the plan announced today is trying to address that. Advanced manufacturing is definitely another one. And the Modern Manufacturing Initiative, at 1.3 billion, is also trying to address that. Another example, when we were devastated by the bushfire more than a year ago, we needed to look into how to use observations that are from space to see how we could help managing that crisis with the bushfire. And none of the data were coming from Australian space assets.

So, I think there is a very thorough reflection within the government today, and realising that, you know, either we accept this dependency or we do something about it. And I think there is a lot of thinking going through this. And this two example of investment is a start, but I echo your concerns, Sue.

Mark Pesce: So, on that topic of... Yes?

David Chuter: Sorry, the challenge we have in Australia, which is more acute than most other markets in the world that are industrialised, is that the very large majority of companies who operate in manufacturing are small businesses, very small businesses by global standards. And also, we don't have large industries that typically have a big customer at the top that drives hundreds of suppliers through a tiered supply chain. In other words...

Mark Pesce: Like an auto manufacturer.

David Chuter: Like an auto manufacturer. And look, we have that. So, what we're missing is someone at the top saying, You're all gonna meet this cybersecurity standard within 12 months, or you're not part of the supply chain." And everybody jumps very high in that process. So, the challenges we have in Australia are unique in terms of how we have to address it. There is an opportunity, though, to say we've picked six primary industry sectors that the country wants to grow in. That should stimulate investment. But there are also real opportunities at the platform level.

So, think of how AI can help us solve better energy use, alright, better material use, better assurance of the providence and high quality of food and beverages that we export. You know, these are strengths and things that Australia is already globally renowned for. We can't stand still in that space. We've got to invest. And AI is a really obvious technology to take and apply at a platform type level, because if you wait for individual companies to do it, we're still waiting for individual companies to embrace lean manufacturing and Industry 4.0. So, there's an option to leapfrog with artificial intelligence. But there's also some challenge and platform approach that we need to be looking at as well.

Mark Pesce: David, you point to some of the inertia issues, You know, we are very comfortable being fast followers in this country. We're less comfortable being leaders. And what you're pointing to is that, in fact, particularly in areas where we already have established strengths in manufacturing, that we are going to need to be leaders. So, what do we do to remove some of that inertia around AI?

David Chuter: I think the first thing is demystify what AI is and what it isn't. It's very easy. We had the same challenge, I'd argue, with Industry 4.0 a few years ago, where, actually Industry 4.0 was nothing more than a collection of technologies that have been used in industry for 20 or 30 years. 3D printing, computer simulation, robotics have been around for a long time, but they were very expensive and you had to have very customised skills, and they had to be in very controlled environments. That's not the case with these technologies anymore. They're almost open source available and at a cost that companies can afford.

So, the challenge to me is, don't get hung up by the terminology. If we start, I think it's important to define what AI is versus simple data analytics, OK. But I think what we need to build is a culture of, data matters. And it's not about the information that data creates. It's actually thinking through, how can that data turn into dollars for me? Alright. And so, you've got to think through, how can I use this to save money or improve things in my business, or to create a new business model?

And I think to Sue's point, we are not going to be manufacturing everything in Australia, that is not the pathway that I think ultimately we're heading, despite COVID. But we also need to build business models that we can take to the world because we'll still be part of a global world. And it may not be necessary products or processes, but know-how, software platforms, services and smart designs that are enabled by machine learning and artificial intelligence. I think that's a golden opportunity for us.

Mark Pesce: Wow. OK. So, if we have this idea that there are easy pickings here, right? Then what we need to start, how do we start to think differently about that? I mean, is it just simply an education task? And all of you, I think, cause you probably all confront this in your work. How do you think about what it takes to remove inertia here?

David Chuter: Maybe I'll just open. I think one of the answers is collaboration. And we're generally not great at collaboration between businesses, between research organisations. It's taken a national cabinet to get some collaboration even at the federal and state level. But the reality is, AI is a common opportunity and a common challenge. It may not be actually what you end up competing on, alright. So, it may be something that clusters of companies can partner up, work with your local university, engage some data scientists. There's smart ways to do it without trying to do it all yourself. You've also got to be, I think, have an appetite to take some ambition and take on a bit of risk.

But the good thing is with things like digital twins and 3D printing, you can test things in a virtual world, in a prototype world, without actually disrupting your current production. So, all of these technologies that this Industry 4.0 and AI enables, allow you to test and develop things offline much more rapidly and much more safely than they would do if you tried to disrupt production. I think if we can raise that awareness that these are not expensive technologies, relatively easy to access, and if you could do it through collaboration, you'll find a way to get on the journey and you'll be surprised at how quickly you can move.

Mark Pesce: Aude, you are in the midst of publishing, I think you said it's eight different technology roadmaps for the space agency, is that correct? You've done the first one. The second one's coming out, there's six more coming this year. How have you positioned AI in manufacturing inside of those roadmaps?

Aude Vignelles: That's a very good question. First of all, how we picked up seven areas when we got created almost three years ago now, so we're not that young anymore, we had the choice of starting doing everything and anything in space, or say, OK, let's stop two seconds and let's pick areas where we think that we have a great future and a great role to play. And to your comment, David, when you say we're not going to manufacture everything in Australia, I agree with you. And that's the approach we are taking in the roadmap as well.

Why did we pick up these areas? It's because either there is a market, there is a need, there is a user to fill, and often people do things without really thinking, is there a market? Is there a market, is there a need, or is there a good reason as to why we need to have this capability in Australia? That's the two simple reason as to why we've picked these areas. So, they are PNTs, for obvious reason. The size of the country, observation, communication, space situational awareness, because of the number of debris around the Earth, and also because we look at the southern and eastern sky from where we are, so we have a uniqueness there.

Remote operation, 40 years of industry we have in mining, we know how to do remote operation. When you go to Perth, you can see that all the Rio Tinto, Woodside and so forth, they're operating from Perth activities that are 2,000 km away. So, we can export that to the Moon and Mars in a blink, and access to space. So, yet there is no AI or advanced manufacturing roadmap, but by doing all these roadmaps, and the first one we've issued was on communication and technology in December. And a couple were about to be published. We've identified some crosscutting technologies, some crosscutting services and some enablers that are applicable to all of the areas. And five technologies are across all of these areas.

AI is the first one, advanced manufacturing is another, cybersecurity, which is, you can't talk about AI and data without thinking of cyber technology. We have digitals that are driven as well. We have a few services, but AI and advanced manufacturing is definitely one. A few example of AI, I've given the SSA example, but when you look at position, navigation and timing, now when you use your phone to go from A to B, your phone is doing a bit of artificial intelligence to tell you this is the quickest way, this is the way without toll, this is the way where there's less traffic, so you're using AI all the time when you're thinking of PNT.

Earth observation is the same thing. You have now satellites, we've on board some very, I would even say much machine learning, maybe not even AI. They're taking a picture. They realise there is a cloud that is on the picture, and they won't take a bandwidth to send you the data of that picture because it will be irrelevant. If it's a spectral camera, so you can take different spectral bands picture of where you are, and if you saw, I need that picture in AR, they will do it automatically. So, that will be able, when the satellite is above the point of interest, be able to take this picture in the specific band that you need to receive and you need to address.

So, lots of AI, you know, observation. Comms is the same thing. When you walk around and you want communication, you should not care if the signal is coming from a tower, from a lower satellite, from a high, from a geo satellite, and which band, if it's laser, if it's K band, if it's C band, you just want your comm. So, there's gonna be some intelligence behind what is the best signal that is around you and making sure that when you move from one point to the other, you have a continuous coverage there.

So, that's for AI, really, across all the area of priorities, and advanced manufacturing as well. I mean, when you think about comms, you need some materials, you know, antennas that only advanced manufacturing can bring you. When you think about instruments to put in space, we started a bit, the cold air, you can also now design instrument that you couldn't dream of.

So, advanced manufacturing and AI are really across all these things. And we are maybe thinking of extracting all that and at some point doing one roadmap that is just about AI and advanced manufacturing for all the space application, who knows? First, we need to go and publish the six remaining roadmap, but it's definitely a very strong domain and capability that we need.

Mark Pesce: Excellent. Alright. We have time for a few questions for the audience, and you guys know that there's a Q&A bar on the side of the screen, but first one that I want to ask, and this is to all of the panellists, we're being asked, how do we integrate academia and PhD students with industry to collaborate and accelerate manufacturing? I presume AI in manufacturing in Australia. How do we actually forge those connections?

Jon Whittle: I can have a first go of that. Because this is exactly what that Next Generation AI Graduates Program is meant to do. So, you know, I think we need to get away from, say, the traditional PhD, which was individual students working on individual problems with individual supervisors in individual universities, and move much more towards a collaborative model where we've got groups of students working with industry, working with other research institutes like CSIRO and working with multiple universities.

There are some reasonably good models of that from around the world. There's the Centres for Doctoral Training, they've called in the UK, that have been running for some years right now. They've been quite successful in moving away from that kind of individual student to kind of groups of students and having bigger impact because of that. Even within Australia, we've got ARC, ITTC that have done that to a certain extent, but certainly within the Next Generation AI Graduates Program, there's an intention to take that cohort approach and put it on steroids, if you like, and really push on that collaborative element, which David was talking about earlier.

David Chuter: Well, Mark, I'd suggest throwing the net much wider. PhDs are really the tip of the iceberg, and your typical manufacturer, they haven't historically gone out and sought PhD level people because the challenges in a typical manufacturing business are probably not sizeable enough for your smart PhD student. You can work all the way back to undergrad students. I think key to this is revisiting things like internships and placements, whereby as you go through any part of your education journey, as an undergrad, as an apprentice, you are spending time both within your education institute and also with industry. That would then apply to post-grads, to early career researchers, to Master's students and also to PhDs.

So, to me, it comes back to a bit of a cultural thing where companies need to see universities and universities also need to see companies as partners in solving some of these problems as opposed to talking different languages and having different cultures. And as opposed to seeing the university system as something that is meant to provide work-ready people for me to employ, it can't be seen as a simple, throw it over the fence, to me, it's gotta be much more integrated. Hence, I come back to the collaboration opportunities.

And there are all sorts of funds out there to help companies with the cost of engaging PhDs. We've got a PhD program. We've got a Master's scholarship program. The challenge is getting industry and university just to talk. And when they do, great things happen and great ideas get shared, but it's that first conversation you've got to get going.

Jon Whittle: Just to clarify, so although I talked about PhDs in that Next Generation AI Graduates Program, it actually will focus as well on Master's students and undergraduate students. I couldn't agree more.

Aude Vignelles: And if I can add to this, I'm a great believer in competition and challenge for provoking collaboration. So, Minister Porter has announced the AI plan launch, but he announced quite a few things for the space agency as well today. And one of them was the 20 projects for demonstration program for the moon to Mars program that we are busy with. And it's 20 collaboration between industry, research. I think the key difficulty when you talk about collaboration is IP. The IP needs to be treated on a case by case basis. And I think, for me, that's the major challenge of having the industry working with research, working with students. How are you going to manage the IP? Who owns it at the end, and what's the story there? But I invite you to read this 20 demo feasibility study because there are some cool projects to bring Australia back to space.

Mark Pesce: Sue?

Sue Keay: Yeah, Mark, I think that what a lot of our manufacturers would like to see is some support in up-skilling the workers that they have. So, Griffith University has just got some funding to help with a Women in STEM cadetship that is available for a Diploma in Computing and Data Analytics where companies are actually given funds to free up people who are working for them so that they can upskill. And I think, actually being able to provide programs like that is really important because there's, I think as David pointed out, there's a whole range of skills that our modern manufacturers require. Sometimes they're picking up people from year nine. And to be able to cope with all of these AI-enabled technologies, we need to be able to bring them along on the journey.

Mark Pesce: Yeah. Yeah, alright. Last question for all of you, this is another audience question, and it's something we've touched on, but it does represent something that's slightly alongside field. Are we looking at applications in AI in what we're now calling smart design, particularly partnerships with designers? So, we're not talking scientists, we're talking designers who are trained to use AI.

Aude Vignelles: I can just, quickly, there is a lot of exercise and project being done with designer in AI to look into how are we going to design a moon base? What's the shape of the habitats, to take into account radiation and all this kind of thing. So, there is some really interesting architectural plan that are available now. And they've been done with designer with the help of AI. So, very futuristic, not so much, because we're going back in four years' time. But this is something in space that is already being used greatly.

Jon Whittle: And, actually, if you look at the, sorry, Sue, if you look at the history of AI, that design element was always one of the motivating factors that drove the development of AI. You mentioned, Mark, I think, in your earlier comments about the Google example of things being automatically designed by AI. But actually, one of the very, very early examples of AI was AI in circuit design and automatically designing circuits. But I think there is still quite a lot of interesting work to be done that there.

So, one of the kind of last frontiers of AI is often considered to be creativity, and how can you get machines to be creative. And that was often considered by some to be impossible because machines can't be creative like human beings. But there are many, many examples now of AI that has created music or has created art that has been performed or has been exhibited in public galleries. And members of the public wouldn't know the difference between that art and human-generated art. So, yeah, I do think there's lots of really, really interesting possibilities for when you bring the human beings together with new forms of creativity that can come from AI.

Mark Pesce: Sue?

Sue Keay: Yeah, Mark, I wasn't sure, are you suggesting that there are barriers to design people engaging in AI? Because following on from what Jon said, I actually think our creatives are much more likely to use AI and not even question it, just as a useful tool that helps expand the range of creativity. You know, Australia won last year's Eurovision AI Song Contest. The musicians never felt that they were being displaced by the AI. They just saw it as a useful tool to expand the range of things that they could do. And I think, in general, designers would use it in the same way.

Mark Pesce: I mean, I guess, part of that may be reflecting on the fact that design as an element has not been foregrounded in any of the discussions, and maybe this is an opportunity or an invitation to do that. David, do you want to just give us some closing words on that? We only have about another minute.

David Chuter: Yeah, look, one of the examples I opened up earlier was about this concept of generative design. I'll give you a very simple example. 3D printed inserts in the medical space to replace bone that maybe needs to be removed either because you've had an accident or cancer. The AI can work all that through in terms of being able to then 3D print something that actually replicates the structure of bone. Imagine how long it would take, or a designer to draw that up in an old fashioned way, OK, whereas the AI algorithms can do that intuitively. And this is where I think it's about bolting bits of technology together. It's not about robotics on its own, or 3D printing. It's about how AI can allow you to bring these different platforms together, to create something that you simply cannot do any other way. That's the real opportunity that exists.

Mark Pesce: Right. And that it's about new tools. Alright. I'd like to thank Jon Whittle, Sue Keay, David Chuter, and Aude Vignelles. This concludes the opening panel section of Techtonic 2.0. We're going to take a 15-minute break until we start the breakout sessions. Once again, bit of technical fluff here, before we go into those four streams, you will need to tap on the live stream button once this stream ends, which is going to be in just a moment, and that will then take you to the next session. But the next session is going to be 15 minutes from now. So, don't close any windows. Go get yourself a nice cuppa, have a comfort break, whatever you want to do, walk around a bit. And then you'll be able to join us again at 2:45. Thank all four of you very much. It has been a great pleasure working with you.

Aude Vignelles: Thank you.

Sue Keay: Thanks, Mark.

(INSTRUMENTAL MUSIC PLAYS)

Description: Against a white background there is teal-colour text on the left.

Text: TECHTONIC 2.0. Australia’s National AI Summit. 18 June 2021.  This session has now concluded. Thank you for joining.

In the bottom right, still colour images of industry. A man in a hard-hat. A woman in a field controlling a drone.

Against, the same background the text changes.

Text: TECHTONIC 2.0 Australia’s National AI Summit. 18 June 2021. This session will commence shortly.

Sanvi, a young girl in glasses and a school jumper, talks to camera. A chyron appears, reading ‘Sanvi. Year 7 student.’ Behind Sanvi is a set of posters. One reads, ‘Metal Tech.’ Live captioning appears on the bottom of the screen.

Sanvi: My question is, will the advance of artificial intelligence...?

Description: Appearing behind the same background is Anders, a boy with glasses. He wears a school jumper. The chyron reads, ‘Anders. Year 7 student.’

Anders: My question is, will AI have an effect on the development of humankind?

Description: Darcy, a young girl with her hair in a ponytail, talks to camera against the same background. The chyron reads, ‘Darcy. Year 7 student.’

Darcy: My question is, how much do you think artificial intelligence will advance in the future?

Description: Yanfu is a young boy with black hair. He speaks to camera against the same background. A chyron appears and reads ‘Yanfu. Year 7 student.’

Yanfu: My question is, could we use our consciousness to control artificial intelligence?

Description: Jane Hume wears a spotted blue blazer over a white blouse. She stands in front of a large painting of a white flower. She speaks to camera. On the bottom, live captioning appears.

Jane Hume: Hello, everyone. I am pleased to be able to deliver the closing remarks at this important event.

Description: A chyron appears, reading, ‘Senator the Hon Jane Hume. Minister for Superannuation, Financial Services and the Digital Economy, and Minister for Women’s Economic Security.’

Jane Hume: Artificial intelligence is a key feature of Australia's Digital Economy Strategy, and Techtonic 2.0 marks the bringing together of some of the best and brightest minds from around the country. A major technological transformation that will change how we live, work and play is underway as technology has become embedded in the fabric of everyday lives.

AI, along with other digital technologies, is going to play an increasingly important role in our economy and society over the next decade. As the Minister for the Digital Economy, my role is to champion the digital economy and to ensure that we are ambitious about Australia's digital future. Harnessing the opportunities that AI...

(INSTRUMENTAL MUSIC PLAYS)

Description: Against a white background there is teal-colour text on the left.

Text: TECHTONIC 2.0. Australia’s National AI Summit. 18 June 2021.  This session will commence shortly.

Hide publication menu: 
Show menu