Where Are We At With Autonomous Vehicles, with Professor Michael Milford.

First published December 2025

Transcript

Hello, and welcome to Where Are We At With…? The podcast updating you on everything... One topic at a time. I’m David Curnow.

Few things altered the physical and cultural landscape of the 20th century like the automobile. In rapidly urbanising countries like America, Germany and Australia , the car gave freedom to the masses, opened the horizons to all, and shaped the design of cities and communities all over the world.

Yet right from the beginning, futurists saw a world where drivers were freed from the machine of freedom. Picture the iconic image in the 50s of a car with its four passengers sitting around a table and chatting, while the vehicle drove itself.

More recently, movies like Minority Report and I, Robot, imagined the personal car being simply too fast and powerful for mere humans to control.

Scene from "I, Robot", 2004 by 20th Century Fox

"Manual override." "What do you think you're doing?" "Driving." "By hand?" "Do you see me on the phone?" "You can't be serious, not at these speeds." "John please just send the backup."

And yet despite the decades of promise, and constant assurances of the driverless car being just around the corner, where are we at with autonomous vehicles?

Someone who spends quite a lot of time pondering just this, mostly while helping to create the technology capable of bringing those cars to the streets, is Brisbane based Professor Michael Milford. He’s the Director of the QUT Centre for Robotics, an Australian Research Council Laureate Fellow,, QUT Professor, Microsoft Research Faculty Fellow and a Fellow of the Australian Academy of Technology and Engineering.

Among his numerous awards, is being named the Australian Research Magazine’s top robotics researcher by citations for 4 of the past 6 years. I assume any tie is decided by a robotic rock-off.

In addition to the research work he has a STEM-based publishing company that’s put out more than 20 books, including his “The Complete guide to autonomous vehicles for kids and anyone else” So today, “Where are We At With…” Autonomous Vehicles… I headed into the QUT Centre for Robotics, and a quick warning, while it’s brilliant for design and technology, the offices are a little ordinary for sound quality. Sorry.

Interview

David Curnow (host)

Professor Michael Milford, thanks for joining me today.

Professor Michael Milford (guest)

Thanks for having me on the program.

For many people being in control of their journey, their destination in life, it's kind of important. It's something we tend to rely on. How is there an attraction then to relinquishing?

I think this really depends on the individual and you're right people have a range of attitudes towards relinquishing control. I think once you've had control for plenty of time you start to see the benefits of perhaps not having control. For example napping in the car. Napping in the car, definitely better if someone else is driving than if you are.

I suppose if we look back, let's start at the beginning, the early hominid sleepwalking, probably the first autonomous vehicle, then perhaps jumping up on a horse that was a bit strong-willed. Those were relatively early ones. But after that we're talking what, 1600, 1700, something like that, when people started imagining the idea of things that are controlled and using them. Leonardo da Vinci, for instance.

Yeah, those Leonardo's automation machines, but those also in religion and mythology, there was the idea of Golems for example, which you can think of as a precursor to some of the concepts that were popularized in robotics in the 20th century.

Indeed. And it's interesting when you look at some of the fiction of writers like Terry Pratchett or like who used the Golems and their autonomous working and in a sense articulated some of the concerns as well. We might get to that in a little bit. Let's talk about then more recently, the 20th century was where the big steps started to be made. What is an autonomous vehicle? Because when I get into a train I don't see the driver, I'm assuming it's a human, I get off the train I don't see the driver, all they're really doing is is making it go faster or slower with greatest respect to train drivers out there. What makes a vehicle autonomous?

I think we need to go back one step to the idea of a robot and then and then go on to autonomous vehicles. So the working definition that a lot of people have used for a robot for many decades has been something that senses the world, thinks or reasons about the world in some degree and then acts. Now the problem with that definition is if you take it at strict face value, you could imagine something like a washing machine or an air conditioner you would call a robot. And so people have added additions to that. For example, it needs to have some degree of physical movement in the world, whether that's moving around on wheels like an autonomous vehicle or a robot arm that is moving around and picking things up. And so that's a reasonable working definition of what a robot, what lot of people would consider as a robot. And then the autonomous vehicle is a bit more nebulous... because it's what is a vehicle versus a robot. The biggest area that that plays out is whether you would consider the automated robots and vehicles in logistics warehouses, which are one of the widespread uses, whether you'd consider that an autonomous vehicle and you'd split the field on that. I would generally talk about autonomous vehicles as not including those. I would talk about those as robots. I would talk about autonomous vehicles being on-road and off-road platforms that vaguely resemble the form factor of vehicles and trucks that we're so familiar with.

It's a hard thing to define because you're right, if something is effectively lifting something up and moving it over here and putting it down, that is something we've seen in factory assembly lines for decades now really so it does change a little bit. What were the big breakthroughs in technology which allowed effectively what we think of now as autonomous vehicles, what were some of the big breakthroughs that allowed for that to start to be possible? Are we talking optic sensors? Are we talking just even the powerful processing capabilities?

So at a higher holistic level, it was the sustained interest and very large amount of sustained investment of resources over an extended period of time. Really the modern birth of autonomous vehicles, I would argue, happened around the 2004, 2005 DARPA Grand Challenges. So DARPA is that Defence Research agency in the US, and they ran a challenge in 2004 and 2005 where semi-autonomous vehicles had to drive about 200 kilometres across the desert. In 2004 all of the vehicles failed within I think the first 11 kilometres and then in 2005 I think four or five or six vehicles made it to the finish line so there was a radical improvement and more importantly it was the first time that robotics and autonomous vehicle technology really captured the wider public imagination and so there's been 20 years by some accountings about a hundred billion dollars of direct and indirect investment in autonomous vehicles it has been enabled by things like hardware, break-through sensor improvements, large advances in computers and AI, but it was that sustained level of widespread interest and investment that was really the single most critical factor.

Those were some of the things that enabled movement forward. What were some of the, I suppose, roadblocks, the bottlenecks to the development of some of the technology?

In the autonomous vehicle space, people should and always look at precedent. So how did other technologies unfold when they succeeded or perhaps when they failed? One of the issues with autonomous vehicles is I think in the early days, people took that precedent a little too literally and a little bit too fast. So there are aspects of what we've learned from other robotic technologies and aerospace that are very informative for autonomous vehicles. But autonomous vehicles are not the same. They don't have the same problems and the same opportunities. One of the key aspects that was really challenging and continues to be challenging with autonomous vehicles is unlike other commercial uses of technology you can't deploy a half-baked works most of the time solution especially on a general public that has never signed up to be experimented on so other technologies for example voice recognition when that first came out it was horrible, horrible experience but that was a first step and then they refined it they improved it they made it so it didn't necessarily have to send your data off to the cloud as the computer became more powerful on your smartphone, for example, you can't do that with autonomous vehicles. You need it to be pretty much near perfect when you first deploy it in the real world. And so hitting that 99.99 % the percentage is of reliability, that's arguably one of the biggest sticking points and it has a number of on-flow challenges around things like corner cases. So those one in a billion encounters that nevertheless your autonomous vehicle will need to safely navigate and solve.

When we develop some robotics, some machine activations, we often look to nature for inspiration. It took human centuries to realise that to fly like birds you didn't have to try to be a bird because, well, we weren't birds. You did something similar, you took inspiration, but then you changed. Are there challenges for autonomous vehicles where we're trying to make them, or have tried to make them too much like a human driver when they are not a human driver?

Yeah, so this is the debate, one of the ongoing contentious debates. It occurs at a couple of levels. So first of all, autonomous vehicles are enabled by what we broadly class as artificial intelligence. We call it driving intelligence or driving artificial intelligence. And the underpinnings of those systems, especially in the last 13 or so years, have been very heavily based on what we call neural network or artificial neural network based approaches. And these are systems implemented in software or the cloud where they at a very, very, very loose level emulate some broad, some not all broad characteristics of how the networks in our brain work. And so some people would say that's biologically inspired. Some people would say it's very tenuously connected to the biology. And so that's one level of the debate in the autonomous vehicle space, the more specific area of the debate with regards to taking inspiration from nature. And in this case, it's taking inspiration from people because dogs don't drive cars, which I wrote in a grant application at one stage. And that is simply that we know people, generally speaking, drive cars fairly well, not perfectly. And they primarily use their sense of vision. They use auditory, they use vestibular systems to sense acceleration and so forth. But primarily we're using vision and we're pretty good at it. And so one of the big contentious debates at the moment is how much we should try and emulate that behaviour, because there's a proof of concept that people drive cars fairly well and how much we should come up with a pragmatic engineering approach which I guess exploits technological advantages where it makes sense. The way this is playing out in particular would be the area of, and we can talk about this forever, of cameras versus lidar sensors. Lidars are a type of range sensor that detects how far away objects, people, pedestrians are around the car. It's an amazing sensor. There's really no analogue in the natural kingdom so bats use echolocation but you could argue that the top lidar sensors now are so far beyond what's available in nature. Should you use this sensor on an autonomous vehicle and the effects on how that makes how an autonomous vehicle drives at least somewhat different to how a human would drive a vehicle. And again, humans use vision partly because that is effectively one of our main and best senses when you compare us to other creatures out there is that our vision happens to be particularly good compared to some of the others. And that nice brain that's behind that vision as well.

Doesn't hurt, depends on who you are of course. Okay so we're talking about the different senses, copying some in nature not others. How much of this is being driven by the push for industrial automation and how much of it is for domestic individual use primarily as in which is driving which?

It's a great question so we can go back maybe two decades for a little bit of history. So originally a lot of the revolution in a lot of robot navigation type advances was brought about by the availability of range based sensors. So these are sensors that emitted some sort of active radiation into the environment and were able to tell how far away things were. Most of those sensors were never designed for robotic applications initially. A lot of the laser range finders were initially developed for things like surveying and also for automation to stop detect if someone's hand entered a forbidden area. And then roboticists as they do co-opted these sensors for robotic tasks. The same thing happened a little bit later in gaming. So the Microsoft Kinect was this range sensor that worked a little bit differently that was revolutionized the robotics and computer vision fields. Again, not designed for robotics wasn't really as far as I know its primary intent, but it transformed the robotics computer vision fields

It's interesting because in the automotive world a lot of the changes are driven by the high-end luxury cars trying to come up with a new trick. think the S-Class Mercedes with a new sensor or a new something like that and then eventually it'll trickle down into your Kia Rio but in this sense we're effectively aiming for the stars and trying to get it into everyone which is a little bit of a different approach. I suppose we've all got sensors these days in our car and even governments are demanding that we have them otherwise you don't get a five-star safety rating or something like that. The science of it is one thing. What about the legality and I suppose the liability when it comes to rushing ahead with sensors that can tell you whether you're going to hit the wall or not. If it's wrong, I hit the wall.

Yeah, and again, this is a messy field. This gives a lot of headaches to regulators, to safety agencies who have to certify this sort of technology and is very expensive to do properly. So one of the things is you really need to break it down into a lot of different components. there's how good a particular piece of technology like a sensor is, like how reliable it works and the conditions under which it doesn't work. But then there's also the, I guess what you call the specificity with which you can define the exact nature of the conditions in which you shouldn't trust that sensor. And that's not always completely clear either. A sensor doesn't have to be perfect, but you need to know when you won't be able to rely on it. So you can use other sensors or have some sort of fail safe. So it's all very, very messy. And for example, if you're a safety ratings agency and your job is to certify aspects of this, can't, you don't have the resources to really exhaustively test it over the tens of billions of kilometres a little bit counterintuitive and a bit challenging to deal with because let's just say hypothetically you're just about to choose from company A and company B and company A had an autonomous vehicle that was demonstrably much safer than people and company B had an autonomous vehicle that was five times worse than current human drivers. You could put both of those cars on the road and drive millions and millions of kilometres and not be able to see any difference whatsoever. So this is one of the challenges in ensuring that these technologies are sufficiently safe before they become widespread.

Professor Michael Milford is our guest today on Where We Are At With. We're looking at where we are at with autonomous vehicles. Professor Milford is the director of the QUT Centre for Robotics. I wanted to very quickly bring in social and pop culture as it were. Every time we see movies about the future, particularly in the last 30 or 40 years, they featured forms of autonomous driving. We think minority report, we think iRobot, things like that. How much of those were right? And were on the right track, I suppose, so to speak.

Science fiction portrayals in books and movies are fantastic and they are often remarkably prescient about what will happen. One of the scenes that I don't think is as appreciated that featured autonomous vehicles recently was the movie Logan with Hugh Jackman and there's a fantastic scene in there where they are near a highway and they're helping corral some horses that got loose and there are autonomous trucks and they obviously talk to someone to get some insight into to what those would look like, because the autonomous trucks don't have windshields, because you don't need a person in the vehicle anymore. They don't seem to be responding very well to normal sort of interaction cues with the other vehicles. I think they force their car off the road. And so a lot of these depictions of some of the aspects, including some of the dystopian potentials around these technologies are remarkably insightful. And I think famously, I don't know if this has actually been verified, things like authors like Arthur C. Clarke who depicts that, I think satellites also even potentially caused issues for commercialization of those technologies down the track. Look I know that most of the movie makers do consult quite widely in terms of... don't always listen but they do definitely consult.

"We have consulted! Tick! Righto." "We've taken inspiration from" you put it that way something like that.

Let's talk about the history of effectively what is now becoming usable, autonomous vehicles, the places where you can actually have them on the road. And I know that's the case in a number of places in the world. We think of Tesla as being one of the first, but there were other companies working in the space, like Google for instance. What was driving that, so to speak? What was behind the push to do that?

So again, a lot of this comes back to that DARPA Grand Challenge series in 2004, 2005. There was also an urban version of that Northern Desert, but in the city in 2007. So back then it was university-led initiatives with some sponsorship from car companies. What happened is then some of these tech companies like Google was one of the front runners in this area. They had their own in-house autonomous vehicle teams. They were automating these Prius cars. So I got a ride in one of those back in 2012, I think it was, which is quite a while ago. And I should also mention there were unis doing autonomous vehicles on highways in Europe and America and Asia back in the 1980s as well, of course. What really happened is it shifted from being a university-led initiative to being something that only sort of billion and trillion dollar companies could do. And so that's how the tech companies got on board and Apple, Google, Amazon, all of those big tech companies and some of the Chinese companies have all had a hand in this and really are leading the charge as well as traditional automotive. But you could probably argue, and this is no disrespect to traditional automotive, that they were sort of a second tier player in trying to enact autonomy. And they were looking at intuiting as like a feature value-add to their product offering.

Which to be fair, lot of the technology thus far has been, whether it's parking sensors or these days lane assist, adaptive cruise control, things like that, or the ability to literally press the button and have the car park itself while you're sitting there. These are effectively features which have become standard. Yes. It's a big step then to go from that to having an app in your car while it's driving.

Yes, and I think some of the early narrative was perhaps a little bit misinformed in that people thought we would just steadily, incrementally work our way towards adding more and more features until suddenly, presto, there was a fully autonomous, privately owned car that you could buy from a major automotive company. And that hasn't really happened. They've added all these nice features that have reduced rear-ending, for example, but taking the next step is such a big one and that is why there's been a big consolidation of players and arguably there's only probably a half dozen credible players in the autonomous vehicle space right now leading the charge.

Is it a brave move for a government, for a local authority to allow that sort of thing to go ahead earlier than others?

This is an interesting question. So I have a lot of respect for the my government colleagues because they have a unenviable job of being stuck between the race to innovate and deploy new technologies and trying to protect the citizen... citizenry, right? And you see this play out a number of different ways. So in the US, there was very much a largely self-regulatory type approach. Other countries are being more cautious about rolling out these technologies. One of the other things that's become clear relatively recently is a lot of these agencies are somewhat limited in how much they can front run in enabling or not enabling technologies. And a lot of it is done on a responsive basis. So when there is an incident, they can then investigate and they can take quite punitive measures and there's a lot of debate around whether that is a fit for purpose system when all of these new technologies are potentially going to be deployed on a roads at widespread scale.

As you're deploying something nobody's ever imagined it's hard to know what the rules for it are until the thing goes wrong. That's an interesting point I want to come to in a moment, I know Tesla's had some problems with that, but let's talk humans because we have computer-like programs. We can run, we can operate, we can create But we also have a psyche, and sometimes that can be a little weird. Like for instance recently when Waymo in San Francisco, a Waymo, an autonomous taxi, you get into the car, it drives itself and takes you to your destination. It killed a cat. And the uproar over that was enormous. There was something like three people killed within the next hour by regular people drivers, barely got to mention. But the cat caused a problem.

(Sidenote)

Ok, so just a little side note on the Cat-killing cab in the US.

Late in 2025, a popular San Francisco neighbourhood cat was killed by a WAYMO taxi, which is owned by Google’s parent, Alphabet. The death of the domestic short hair, “Kitkat” was international news… It apparently ran under the wheel of the parked cab and then after a while, the car pulled away and ran over it. The Waymo controversy is interesting because of just how emotional it shows most of us to be. San Francisco is of course one of the most liberal and free thinking places in the US.. possibly the world. So when disruptors like Uber and Tesla launch, they find open arms in the steep and varied streets. So too, Waymo. Those who’ve ridden in the driverless taxis describe wonderment and childlike glee in the experience… although there is still the battle over the choice of radio station.

Onboard video of Waymo

Child: Can we like put on our own songs?

Adult: Mmmhmm

At the end of 2025, Transport NSW confirmed they are speaking with Waymo, and they will be hoping to conduct trials for what’s termed level 4 autonomous driving in a specific area of Sydney. Level four means no one needs to be sitting in the driver’s seat, but a driver can take over if there’s an emergency.

(Return to interview)

It speaks to our psychology, in a sense our fear of these sorts of vehicles. How important is that?

It's very important and the mistake that I think all technologists make is sort of not all a lot of technologists make is in just sort of brushing that aside or making sweeping statements like of course there's a 1 million plus people that killed in automotive related incidents. Why would you not adopt this technology as quickly as possible? And of course we've learned from bitter lessons in the past that you should always take the words of those predictors that technology is going to solve all problems with a fair grain of scepticism. It's really important. I think we should emphasize that it probably shouldn't be overblown. So I don't want to equate, for example, streaming services with autonomous vehicles but there have been a number of technological introductions. And I'm not saying that these introductions have all been good, where people, large swathes of population were... horrified conceptually initially and then it was introduced, it became widespread and now it's just a sort of matter of fact. I think the whole idea of not owning your own movies and not owning your own music was and continues to be abhorrent to a lot of people but it's become very, very widespread and now it's just a sort of matter of fact. Autonomous vehicles also have a precedent in ride sharing. So for example, when Uber first came in, I'm not saying the way they came in was fantastic in any way, but it was transformative. It was a big disruption, but now it's just a matter of fact in a lot of cities around the world. So these concerns are extremely important to pay attention to and woe be the person who doesn't pay enough attention to them, but they aren't the only factor that will determine whether a technology rolls out.

There's always going to be somebody raising the trolley problem in terms of our decision making when it comes to vehicles such as these. Again, is it important or is it just worth keeping an eye on and humans don't make the best decisions either?

So the trolley problem is really important as long as everyone's aware of its limitations. One of the limitations I don't think is talked about enough is humans don't do well with absolute hypotheticals because they always know there's some level of doubt. The other thing a lot of autonomous vehicle people will say off the record is that a lot of the current technologies are still not sufficiently able to distinguish between between a child and a small adult 100 meters away. And so this theoretical decision is purely hypothetical anyway. So it is important. I think there are other there's a whole range of other issues beyond the trolley problem around social impact, the impact on how cities are laid out, the impact on the jobs landscape, all of which are equally important. The trolley problem does seem to get an extra-ordinary amount of attention even today.

I suppose partly because we don't know what decisions we'd make and a lot of people struggle with it internally, let alone whether a machine is making that decision. It does bring up one of the problems I know that has caused headaches for both drivers, pedestrians and car companies and regulators. Tesla particularly has been noticed for the fact that one of the best ways of making your self-driving car better is by finding where it's done it wrong. And in a car that can be an accident and often death. In other words, a stupid Tesla fault actually is good for the company even though it might, sorry, kill you. How much of a problem is that that the best way of learning or one of the big ways of learning is through accidents?

So you'll get a number of opinions on this, right? So a lot of people would push back that that is actually an assumption that will have to happen in order for the technologies to reach game time. People do use increasingly sophisticated test tracks, increasingly sophisticated simulation. The fact is, of course, that at some stage, if the technology were useful, you'd have to deploy it in the real world at scale. I think one thing to keep in mind, which doesn't absolve any of these companies of their responsibilities, is that even if autonomous vehicles become, say, widespread in the USA, and let's say they're five times safer by whatever metric you used than people, there'll still be 8,000 fatalities every year involving autonomous vehicles with varying levels of fault of the autonomous vehicle. And that would be a technology where most people would argue that overall it's beneficial, at least in terms of safety. You would have to check that the 8,000 people that are being killed is not some perversely focused subset of the demographic with all these technologies. They always have to check. But there will be the technologies is not going be perfect. physical world, tires blowing out, is never going to be perfect. And so we will have to deal with that. And part of the reason it's been so challenging to deal with is because they're not deployed at scale, perhaps rightfully so yet, every incident does rightfully get a lot of attention, unlike the regular human-caused fatalities which continue to go on.

Possibly should get more attention. And again, part of the problem too is, and by problem I mean it's also a benefit that the more of a particular piece of software is being used, that's where it's going to get better. So that's why for instance, self-driving, I know in America, the more Teslas are using self-driving, that area is a better place to use your self-driving car because they've ironed out some of those kinks and anecdotal evidence is that California is the best place by far and away to drive your Tesla because they know to stop at that stop sign or this particular corner looks a bit funny but it's okay do this.

Yeah, so one of those aspects is they do annotate and correct maps manually to deal with identified problematic cases. When I did a test drive a few months ago in Brisbane for full self driving premieres, I got a preview of it. It made a minor mistake of falling in a vehicle in front of it into a car park instead of taking the next right turn. And someone else had a very similar thing happen. So they're very open about saying that was something they were going to annotate and fix. I guess one of the things to be cautious there is whether you can scale that process up and actually annotate and correct your way manually to a sufficiently safe service is completely up in the air.

How did it feel when it made a decision that you felt wasn't right?

I knew it was probably going to make that decision because we know that a lot of these learning-based systems do have a tendency to queue off the vehicle in front. And to be fair, humans have a tendency to queue off the vehicle in front. So if people were regularly driving off cliffs, some tiny fraction of people following behind them would also be driving off cliffs. And so this is a problem that's common to both humans and the artificial systems. The ride quality felt pretty good qualitatively. Again, I could sit for thousands of hours in these vehicles and I would not be able to say anything quantitatively about how safe it is.

One of the problems, and I've said that a lot, I'm gonna put it a different way. A challenge for introducing autonomous vehicles is not necessarily just the vehicles themselves, but it's the fact that you've still got humans behind the wheels. That Venn diagram of human drivers versus robot drivers trying to work out what human drivers are doing. Is there a way of solving that without just saying one's allowed and the other isn't?

It's given the amount of resourcing and attention being devoted to this problem. There are likely ways to find a way to introduce autonomous vehicles into mixed environments with human drivers that is adequate. It gets a little bit tricky if the human drivers are adversarially trying to screw with or exploit the autonomous vehicles. So we saw this, I took a bunch of Waymo rides in Los Angeles back in the middle of this year. And there was at least two occasions one vehicle came overtook us into incoming traffic because they were, I guess they were impatient and another vehicle pulled out aggressively in front of us possibly intentionally and brake checked our Waymo and so there's that dynamic of how people change their behaviour that's the big wild card here.

Yeah that's one thing to have it driving around the streets of a capital city it's another to be on the Bruce Highway with a tradie on his way home and try and interact with those particular drivers.

A tradie, a lawyer, Yeah, could be any profession though. There are bad drivers from all walks of life.

Not necessarily saying tradies are bad drivers, just there are some dangerous roads in Australia and they are used by some different people. Let's talk about your work because recently particularly you've been doing a lot of work on Visual Place Recognition. What is it? Why is it so important? I mean it sounds a lot like me on a golf course. Where am I? How did I get here? What's it about and why is it useful?

So anything that moves in the world, could be a robot, a drone, an animal, a person, a shipping container or a hospital bed. There is a general interest in knowing where it is and tracking where it goes. And so that's really what this field of visual place recognition comes down to. It's a very technical term, but basically the idea is that at some stage you've created what we call a visual map of the environment. So maybe that robot or a camera has previously been through the environment. It's stored a map of what different parts of that environment look like and the task of visual face recognition is reconciling or matching what your robot currently sees to things in that map and trying to work out where you're located. It's useful for navigation, it's useful for tracking assets, useful for anything where things are moving around.

It's interesting because that then parallels very neatly with, again, the animal world, the way a lot of creatures navigate, insects for instance, ants. It's very much that style of imprinting on itself where it is and trying to learn where to go from there. Let's look into the future because as we know, driverless cars are here, autonomous vehicles are here. They've been in industrial settings for decades. We know that even in the mines there's at least 400 odd of the big Komatsus here in Australia. Where are we likely to at least notice them first. Here in Australia at least.

So I gave a talk literally three days ago at a conference and I actually made a specific prediction about that very question for the first time. So journalists have generally been frustrated with me because I've given that typical academic sitting on the fence question. But right now, unless there is a major catastrophic setback for the technologies, for example, there's some series of tragic incidents which rightfully grounds these companies for 18 months, not just their company, but perhaps other companies as well. I'd be very surprised if we don't see centrally controlled autonomous ride sharing fleets in some of our major metropolitan cities in Australia in the very near future. Like we know that they have had a number of conversations with transport agencies and government. We know that they're operating at scale in San Francisco, LA and obviously in Chinese cities. The only real showstopper has been whether those companies have had sufficient interest to come and deploy in Australia. All the other problems are addressable with enough motivation and so I would be very surprised if we don't see them in reasonable numbers in for example Sydney, Brisbane or Melbourne in the very near future.

And very near future means?

Next couple of years I think there'd be a reasonable chance because these companies have announced very aggressive scaling plans to a lot of the other metropolitan cities. For example London was announced recently and it's not just one company it's usually multiple companies announcing it. So unless that doesn't roll out as quickly as possible there's really no reason that you wouldn't see them in somewhere like Sydney in the nearest future.

And as somebody who has actually not only worked with the technology involved but sat in these vehicles quite a lot in various places, when one becomes available for the personal driver will you be jumping on board?

So as a day-to-day thing I need to get around for work to lots of locations so I would be very willing to try for example a Waymo in Brisbane if it was available.

What about buying yourself a car that could then do it for yourself?

I think it's very unlikely or much less likely we'll have privately owned autonomous vehicles in the near future for a number of reasons. I think the initial foray will be the centrally controlled ride sharing fleets, which are operated by a Waymo or perhaps or perhaps a Tesla or perhaps another company like Wave.

I suppose you don't have to insure somebody else's car.

There's all sorts of issues with that. They can control the environment, they can shut down the fleet when they want private fully autonomous vehicles apart from the wild card that is Tesla where currently you have to fully supervise it so it's not really autonomous is probably not on the horizon.

Okay so we are still not there but we're pretty close.

We're there with centrally controlled autonomous vehicles, private ones a little ways off.

Okay Professor Michael Milford thank you very much for your time today.

Thank you.

Professor Michael Milford, the Director of QUT’s Centre for Robotics. The technology is developing pretty rapidly these days, check out your news provider of choice for when you might see an autonomous vehicle near you. You can see the transcript for this episode on our website, www.wawawpod.com that’s w-w-w dot w-a-w-a-w pod dot com. You’ll also find links to Professor Milford’s QUT page and his personal site with his books and other endeavours. Feel free to leave us a note if you have any questions, or want to suggest a topic to explore for our future episodes of “where are we at with”. Music for the podcast is created by Michael Wilimot, production assistance from Claire McMillan… I’m your host, David Curnow, thanks for listening.