Intelligent Not Artificial - the future of work and why we'll still have baristas (not coffee machines) in the year 2040

Intelligent Not Artificial – the future of work and why we’ll still have baristas (not coffee machines) in the year 2040

Presented by: Dr Stefan Hajkowicz

[Start of transcript]

Spencer Howsen: Stefan Hajkowicz, welcome to the podcast.

Dr Stefan Hajkowicz: Thank you very much. It's great to be here. Artificial intelligence is a brilliant prospect for taking out a lot of the dull, dangerous, dirty, repetitive work, but it creates a different sort of work environment. Psychological stressors get a whole lot more important in a world of artificial intelligence enablement. Plus there's ethical issues about how we go at using artificial intelligence and there's a huge amount of career transition ahead of us. People's jobs aren't all going to just disappear. I mean early estimates were saying up to half of the labour market would all be losing their jobs to robots. It's not happening at all like that. It's closer to about 14 per cent of jobs that are actually disappearing. Checkout operators and train drivers might be in the case where the job is highly automatable, but for the majority of jobs they had just impacted and changed and many jobs are improved by AI.

Dr Stefan Hajkowicz: And then there's this creation of an entirely new workforce of hundreds of thousands of people moving into AI professions. And for the time being we can see really strong wages, growth and job opportunities for people with the right sort of skills.

Spencer Howsen: So is artificial intelligence robotics, or is there more to it than robotics? The robotics we've been talking about and waiting for the last 40 or 50 years?

Dr Stefan Hajkowicz: Robotics is a subfield of artificial intelligence but not the whole thing. So there is computer vision, machine learning, natural language processing. Machine learning lies at the core of artificial intelligence and it's the ability of the machine to learn and problem solve without explicit guidance from a human being. So we can give it a task and it can build its own strategies or processes for solving that task, which is what is making it so useful because the task complexity can increase.

Dr Stefan Hajkowicz: So whereas historically machines were only able to do very specific, well-defined things, we are now able to insert them into more ambiguous environments. So robotics is all the machinery that actuates a task, the machine learning are the algorithms which help it work out what to do. Computer vision is a whole set of technologies and research about helping the computer work out what it's seeing, natural language processing and human language technologies are all the technologies associated with AI that help the technology work out what's being said to it or be able to talk to a human, either written or spoken. And human language technologies is taking big advances, knowledge representation is a subfield of artificial intelligence, which is all about the creation of a model of the environment within which the machine is operating. It's got to build a construct, and that actually is a subfield of science and research about how to do that well.

Dr Stefan Hajkowicz: So there's all these different fields of AI. It's not just a robot, it's a lot of things. In fact, a lot of the AI that we're working with doesn't look anything like a human. The whole humanoid robot thing is like a tiny fraction of a percent of robots. The vast majority of robots look very boring, uninspiring, but are doing really important things for us. And they may not be humanlike at all in what they're doing. So in all our work, we've been careful not to talk about AIs mimicking a human being. It actually is quite different to what a human being is.

Spencer Howsen: So psychological impact on the workforce you mentioned there, at the moment I'm going to guess that that is anticipation and fear, is it? As much as anything? And different psychological impacts might present themselves later.

Dr Stefan Hajkowicz: I think the number of Australians who are genuinely freaking out about losing their job to a robot is pretty tiny, actually. Significant for those who are in that situation. But it's not huge, it's more the fact that a lot of the manual tasks are disappearing and it is all very cerebral, all knowledge oriented. We're working with our brains much more. And that creates a lot of pressure on concentration gets really important but a different sort of cognitive or psychological stress I think that we're starting to understand in the workplace. So physical work can be demanding but mental work can be also very exhausting as well. And then what frame of mind you're in really matters. But I think that's what we're starting to move towards. Done well, AI is a good thing, because it takes a lot of the repetitive boring and work that humans don't like and lets the robots do it to push us into the space of work where humans do like.

Dr Stefan Hajkowicz: But there is a big adaptation dilemma there for people who might be working in a physical manual job who are then moving into an office job is, are they really going to enjoy it? And what will that be like for the individual? So the type of your day-to-day activities in an AI enabled world changes quite a bit. And that's where I think we are going to start to see what's some of the new stresses, just the information overload that we have to deal with or people who are just overwhelmed with the amount of technology, data and information that they've got to process I think is also part of it. But mental health gets a whole lot bigger in the world that we're moving towards in terms of workplace health and safety.

Spencer Howsen: Mainly for people who up until now have seen themselves as more physical workers than white collar?

Dr Stefan Hajkowicz: That's right. That's right. I think in their challenges for a lot of physical workers who might work on a mine site or somebody who drives a train doing something with their hands and operating machinery. Those jobs could disappear and they end up moving into an office job where it's much more knowledge based work, which could be good, but the transition could be hard to achieve as well. Yeah.

Spencer Howsen: So for organisations, for managers and senior management in organizations over the next few years, they're going to have to be on the lookout for this.

Dr Stefan Hajkowicz: Mental health is certainly getting a whole lot more important, and in a knowledge-based workforce it's everything. If your employees have bad mental health or are suffering depression or all sorts of challenges they face in the workforce, we know they are way less productive in their jobs and all sorts of conflict can arise as well. So in recent years we've seen this ramping up of attention to mental health issues in the workforce for productivity. So really vital for that. Being able to detect, flag and work with mental health issues a whole lot better. But the productivity of a digital or a knowledged workforce is all down to mental health and wellbeing is really important. Some tech companies pay you to sleep, for example. If you can choose to do this, but you can wear a monitor which tracks your sleep. If you meet your sleep requirements, they know you're going to be way more productive on the job next day. Like not just a bit but, but three or four times what you could do.

Dr Stefan Hajkowicz: So if we deprive you of sleep for a day or for even night let's say, the next day at work we know you're not going to get very much done. Especially if you are say a software engineer. But if we've got good sleep patterns for you, you're going to be really productive. So it makes absolute sense for an employer to get out there and make sure their employees are getting proper sleep. And I know in the States that is starting to happen where incentives are being offered to employees to get enough sleep. But I'd say lack of sleep is a simple but really significant determinant of people's poor performance at work and mental health issues that we're going to come across as well.

Spencer Howsen: That's kind of scary, the bonus for sleeping. And yet I think to myself, "No, I'm wearing wearable technology today that gives me frequent flyer points if I put the phone down before a certain time every night. So I'm already, I've volunteered to do that. It's just my boss now offering to pay me to do that."

Dr Stefan Hajkowicz: That's right. I get why you could be scared. These schemes damn well have to be opt in with no harm if you don't choose to do it, but there is also a lot of benefit to you as an individual if you get your sleep patterns working properly. It's not just a ... You'll live a better life and like you, I've got a wrist watch, well I wore the Virgin [inaudible 00:07:30], there was a competition and you wore a wristband and it did measure your sleep and all other sorts of things and it was all an anonymous so it wasn't ... But it's good, this is a data set on sleep patterns that lets me improve my sleep, so done well with privacy and confidentiality and you're the one in the driver seat choosing to do this stuff. I think a lot of people will actually opt to use these systems to improve their sleep and their productivity at the same time.

Spencer Howsen: The other spin off from all of this surely is a decrease in physical injuries.

Dr Stefan Hajkowicz: Yeah, I think that is the case in a lot of ways. So less physical, dangerous manual tasks. We are moving towards that world. Really important for the agricultural sector. Farmers have a very high physical injury rate on farm. Agricultural robotics has a promise of taking that down. Mining and manufacturing are also two sectors which due to the nature of the tasks associated with high levels of physical injury compared to other parts of the economy. So that will take down, but then there's all of the passive surveillance that helps improve safety. So for example, the forklift truck that has computer vision enabled and it sees a pedestrian or the person in the path when it reverses and stops.

Dr Stefan Hajkowicz: So recent study from the United States traffic authorities came up with pedestrian avoidance systems in cars, which are available in a lot of cars that we have today. Decreased mortality rates by 87 per cent when autonomous emergency braking was used with pedestrian detection. So that is a fantastic outcome to achieve an 87 per cent reduction in pedestrian mortality due to the use of those systems, which the computer vision technology detects it's pedestrian, puts on the brakes automatically and takes the speed of the vehicle down. So it's a severe injury or minor injury rather than a fatality or a severe injury. So that is the direction that we're heading, that's a microcosm of the technologies we use. A lot of cars and machinery also detects ... Looks at your eyes with computer vision and it can detect if you're falling asleep, which is also really important because if you're falling asleep there's a risk of an accident.

Dr Stefan Hajkowicz: But in so many ways a mine site, a manufacturing operation we can use computer vision systems to monitor what's going on, pick up safety issues and correct for them. My belief is that even partial automation of cars is going to be an absolute step change in safety on the roads. It's the biggest opportunity we have to take this horrific accident rate that we have down to much lower levels.

Spencer Howsen: Yeah. Never zero probably, that's-

Dr Stefan Hajkowicz: We should definitely be targeting zero, right? We want to move towards that world. I think there is a possibility. Some car manufacturers have that as a target that least people in the car they'll get it down to a zero fatality rate. Nobody ever killed in our cars by a certain year. So it may or may not, but we should be taking things down. There is an opportunity to use or semi-autonomous features in vehicles to substantially reduce road traffic accident rates, which are horrific.

Dr Stefan Hajkowicz: When you look at the stats and you realise how much pain and suffering goes on in Australia every year from fatality or serious injuries on the road, it's just mind boggling. Few people's lives have not been impacted from a road accident in some way or another. So we should be taking every opportunity up to do it and we're moving way too slow. I think we should have for example, standards in place that every new car sold in Australia has mandatory autonomous emergency braking at high and low speed, plus pedestrian detection and avoidance and breaking. Those things should become mandatory. Possibly lane detection and correction as well.

Spencer Howsen: It is kind of hard to fathom how we have the situation when the technology exists. It's only people who can afford it that can have that safety feature.

Dr Stefan Hajkowicz: It does exist and I think we are seeing some car models bring it down to highly affordable levels. They're not out of reach. I know it's not actually all of the very highly expensive cars that have it. And it is quite feasible to see those costs sort of ... It's sort of like seat belts. It doesn't have to cost us an arm and a leg. This technology is doable for most of the major car manufacturers that sell into Australia.

Spencer Howsen: And if you think about cars and other vehicles as being part of workplaces, that's not going to stop companies and organisations to having a policy where they will only buy vehicles that have this gear.

Dr Stefan Hajkowicz: That's why I think every organisation should only buy car fleets that have autonomous emergency braking at high and low speeds combined with pedestrian avoidance and autonomous braking for pedestrians. That should be a minimum. Then I think we should look at lane detection and correction systems as well, but that's also quite doable.

Spencer Howsen: Once we get to completely autonomous cars, and I know at the moment where we're at is still having to engage as a driver, but if we ever get to that point where the driver does nothing, then the temptation is going to be there to work more, isn't it? On the way to home from ... And then we've got a whole series of other problems again.

Dr Stefan Hajkowicz: So the autonomous vehicles at the moment, will still pull over and stop or flash if you take your hands off the wheel, or stop being attentive. They've got systems to pick that up and they ... Tesla's cars will do that. You're not allowed to take your hands off the wheel and fall asleep or do something else. You've got to stay attentive. So there's five levels of vehicle autonomy and we are probably seeing level four autonomy, which is autonomous with all of the road infrastructure and markings and everything you set up the way the car would expect it.

Dr Stefan Hajkowicz: Level five autonomy is where the car can handle any situation and that it can drive along the road and pick up, whether it's a water puddle or upturn, which is at, it can be a country road with no lane markings. Or it can even do a pull off a hook turn in Melbourne which it probably sends a WTF note back to the head office of why they thought of that. But the fully autonomous level five vehicle has no problem negating any situation on the road.

Dr Stefan Hajkowicz: Whether we get to that is a good debate. We are certainly not there. The next little bit on autonomous vehicles to get us to that is proving incredibly hard, because of the variety of situations that the car will find itself in and there is starting to be a bit of a debate about whether we really get to level five autonomy. Level four autonomy is a reality with some caveats around it. It's probably going to take longer than we thought, maybe 2030 and beyond that scenario you gave of being able to open a book up and do some work on the way to work and not have to drive is a feasible one, or to fall asleep in the car. I think it is, but for the near term for the next five years or so, that's not ... I can't see that happening on Australian roads. Car manufacturers have consistently missed their deadlines for levels of autonomy and will likely take longer because of the need to really make those systems fully safe.

Dr Stefan Hajkowicz: But the key thing is, for safety we don't need to get all the way there. We can have semi-autonomous features introduced now that make a massive difference.

Spencer Howsen: Finally then. We've talked quite a bit here about cars and that is part of the workplace, the vehicle as a tool. What's your other advice to organisations where AI is right now in 2019 and going forward?

Dr Stefan Hajkowicz: So you need to help all your staff make a transition from their current roles and duties into new ones, which are reshaped by artificial intelligence. Now, spreadsheets didn't kill off accountants, smart accountants learned how to use them and got better jobs. And that is the shift that is likely to keep on going and even speed up. But it's a big retrain phenomenon. I work at CSRO and we have an enormous retraining reskilling campaign on internally. You can do courses from a couple of days through to several months for up to six months on data science. Because no one in our organisation can really be much of a scientist by 2025 unless they've got some kind of data science basis to them.

Dr Stefan Hajkowicz: Just from the social sciences to the physical sciences we're all messing around with big data sets and we want to be able to do that really well. So this retrain is really important for my future career and I think that's exactly the sort of challenge that employers have to help that happen. Then being attentive to the mental health and wellbeing of your staff is getting really important. We have a different nature of stressors being applied to an individual. In a job like mine in an office, the physical risks are very low unless I chip up off the bottom floor of a ... Bottom thing of a filing cabinet maybe, but that's not really the ones that we would identify that I could see. I'd say they're almost exclusively mostly in the mental health space is what really are the big cost items. And as everything is cerebral and it's all knowledge based work that gets more important, so there's a huge drive around I think mental health and wellbeing in the workforce.

Dr Stefan Hajkowicz: Those are probably some of the key safety implications, but then it's ... We've talked about cars as a good way of illustrating that what's possible but just there's so many technologies that will improve the safety and wellbeing of your staff that we can start to be using. One of the things to think about is privacy and confidentiality as you apply those technologies. The safety systems that pick up whether people are driving, like truck drivers, are they driving safely or not. They're getting monitored or even at a mine site or on a factory floor, cameras can be instilled filming what everyone does all the time and working out whether or not it's safe and it's compiling a big data set on them. But that data could be damaging to individuals if it were released because you're probably catching all sorts of things about the individuals, which are within their private realm. So that is a key consideration, is the use of data and surveillance systems for safety to get positive outcomes. But you're also required to make sure you keep that data private and confidential and do it in a respectful way to your employees.

Spencer Howsen: Great to catch up with you, Dr Hajkowicz, thank you so much for being on the podcast today.

Dr Stefan Hajkowicz: Thank you very much. It's been a pleasure.

[End of transcript]