Latest / Eh Sayers / Is AI Coming For Your Job?

Transcript

Is AI Coming For Your Job? · Eh Sayers

• Tegan: Don't tell my boss but I have a really great idea. Hey Siri, create a podcast episode for me about AI. "Hm. I don't see an app for that. You'll need to download one." Ok, maybe this is a bit harder than I thought... Welcome to Eh Sayers, a podcast from Statistics Canada, where we meet the people behind the data and explore the stories behind the numbers. I'm your host, Tegan Bridge. Depending on who you ask, AI is either one of the greatest technological transformations of our age or a bubble... or maybe it's somewhere in between. One thing that's causing a lot of uncertainty is how this new technology will affect the labour force... AKA all of our jobs. Technology making a workforce to adapt is nothing new. We even have a specific word in the English language for someone who opposes new technology, "Luddite." The OG 19th century Luddites are now remembered perhaps most for smashing machines, but they were skilled textile workers who weren't necessarily opposed to new technology. Rather, they were opposed to manufacturers who were using these machines to get around standard labour practices. They wanted the workers who operated these machines to have the training they needed to do so and they wanted them to be paid good wages. Luddite nowadays is often a bit derogatory, a bit insulting, and it’s evolved quite a bit from its original meaning, but I don't think being uncertain of the consequences of new technology is a radical stance. As long as we have to rely on work to make a living, it's not unreasonable to feel uncertainty about the consequences of a technology with the potential impact of AI. I remember the hullabaloo from a couple of years ago when AI-generated art seemed like it was everywhere. And I remember the backlash from digital artists who saw not just their livelihoods at stake, but their very craft. Is art made by a computer even art at all? It's not just the art world that’s affected by AI. Can AI write a book or a recipe or computer code? Can it make a video game? If not today, what about tomorrow? And what about a podcast? Will the podcasts of the future be made by AI? (She said with some concern.) So, that's the question today. How is the labour force of tomorrow going to be shaped by AI? And we have just the expert to explore that question with us. Tahsin: My name is Tahsin Mehdi and I'm an economist in the social analysis and modeling division of Statistics Canada. Tegan: What do we mean when we say AI, and why is everybody talking about it? Tahsin: AI, or artificial intelligence, is the theory and development of computer systems that can perform tasks normally requiring human intelligence. In other words, it's the ability of a computer or machine to mimic human interaction, decision making and thought process. Um, you probably hear the term, uh, automation a lot. Tegan: Mhm Tahsin: AI is not, uh, exactly the same as automation though. So it's important to draw a distinction between the two concepts. Automation is actually the use of technology to perform simple routine and non cognitive tasks. AI on the other hand, it's the use of technology to perform complex non-routine and cognitive tasks. What sets AI apart is that it can be trained on large amounts of data, allowing it to learn, improve, and adapt over time. Tegan: Right, so automation would be looking at asking my phone to set a clock, an alarm, every day at 7. AI would be asking it to look at my first calendar event of the day and setting an alarm for an hour earlier. Tahsin: Yeah, that's kind of the idea. So if you think about jobs in factories like assembly lines, Automation is actually quite applicable to that setting because they have to do a lot of repetitive tasks, where AI is actually just, it's mimicking the human brain, right? So AI can do a lot more like intelligent tasks, basically. Tegan: So your research is more on the labor force. How many workers do we think might be affected by AI? And what are kind of the different ways that this might impact them in their jobs? Tahsin: Yeah, this is something everyone wants to know, right? So, just how much of a disruptive force, uh, will AI be on the economy? But, this is really difficult to answer for many reasons. AI is developing very rapidly; so we really don't know how powerful it'll be in the future. We also don't know how workers, businesses, and even governments might react or adapt to AI. But StatCan's recent study, which is titled Experimental Estimates of AI Occupational Exposure in Canada, it does shed some light on the possible effects of AI on the Canadian labour market. Measuring the effects of technological change is not an exact science by any means. There are different ways of doing it. One study might say X percent of jobs are affected while another one might say Y percent. But right now they seem to more or less tell a similar story when it comes to the types of jobs that might be impacted. So this particular study from StatCan is based on a method developed by academic research as well as the International Monetary Fund or the IMF. The method is applied to Canadian data by combining 2021 census data on workers with data from a website called, uh, Occupational Information Network, or O Net, which is used a lot in labour economics research to study the types of tasks people might do in their jobs and the abilities required for those jobs. What the method in this study does is it measures the extent to which different jobs might be exposed to AI, so how good is AI at mimicking human abilities used at work like speaking, listening, writing, memorization, deductive reasoning, and so on. So results show that around 60 percent of the Canadian workforce Might be in jobs highly exposed to AI. Tegan: Highly. Tahsin: highly. Yes. So that was about 8 million workers in 2021 Um, the results are actually similar to what we see in other advanced economies like the U.S. So what's really interesting here is that unlike past waves of technological transformation like automation, which mostly affected lower educated workers doing routine jobs, AI is more likely to affect higher educated workers doing cognitive jobs because, unlike automation, AI isn't limited to just performing repetitive tasks, it can do complicated things as well. So in the past, it was assembly line and industrial workers who were mostly affected. But now with AI, what we're seeing is that professions requiring higher education, such as information technology or IT, education, healthcare, business and engineering might be affected as well. However, keep in mind that what we're talking about is AI exposure, which isn't necessarily the same thing as risk of job loss. So a job being highly exposed to AI doesn't necessarily imply a higher risk of job loss. At the very least, it could mean some degree of job transformation, we call it, meaning that certain day to day tasks might evolve as a result of AI. Tegan: So it's not necessarily that an AI is going to be making a podcast, but I might be able to employ it in some way. Tahsin: Yeah, that's the idea. Tegan: Okay. So do we know what the share is of jobs that you would say are more at risk versus those that are more complementary to AI? Tahsin: Yeah, so when we think of automation or AI, we often imagine like robots and machines taking over our jobs. But what often gets overlooked is that new technologies might benefit certain jobs. Take, for example, the invention of the digital computer. There's no doubt that computers replace some jobs. But at the same time, they also created jobs related to digital technology. So even if a job is highly exposed to AI, it might still be able to benefit from it in some way. So, an important feature of the measure used in the stat can study is that it also considers the potential complementary role AI might play in certain jobs. So 60 percent of workers might be highly exposed to AI, right? But if we drill down on that number, about half of that 60% are actually in jobs which might be highly complementary with AI. So examples of these types of jobs include teachers, nurses, doctors, and engineers. So, generally speaking, these are jobs which involve a high degree of communication skills. For example, doctors make consequential decisions about other people's health all the time. AI might be able to be used by doctors to help them diagnose diseases, perform surgery, but ultimately, human oversight will still be needed in these jobs where there's so little room for error. Teachers can use AI to help them create personalized lesson plans and content for different students based on their needs and learning styles. So we live in a world, uh, now where it's easier than ever to acquire knowledge thanks to online resources. But AI still can't replicate that classroom experience fostered by teachers, which might be critical, especially for younger students. So these are just some examples of the ways AI can play a complementary role for certain workers rather than take over all aspects of their job. Now, the other half of that 60 percent of workers highly exposed to AI are in jobs which might be less complementary. Um, with this means that it's possible that relatively more of their day to day tasks might be replaced by AI at some point in the future. These might be jobs in business, finance, and IT for example, but again, this doesn't necessarily mean that these jobs would just disappear because IT professionals like software engineers, just as an example, are integral to AI. Since there would be actual individuals coding the AI algorithms. So it could mean simply that AI might change the nature of their work and their daily tasks. And perhaps free them up to do a new set of tasks, like supervising and maintaining the AI. Now, having said that, we do have to keep some important caveats in mind, these results. So first, uh, like I said, relatively higher exposure to AI doesn't necessarily imply a high risk of job loss. Second, businesses may not replace human workers with AI, even if it's technologically feasible to do so because of financial, legal, and institutional factors, right? So they may not find it profitable yet to adopt AI. So AI may not be worth it yet for some businesses. Third, the estimates only capture what's called a narrow view of AI, which includes generative AI like ChatGPT. But they don't capture more advanced AI, which might have the capability to, like, integrate with robotics and other hardware, and which might be able to think and act autonomously, like you might see in sci fi movies, for example. Tegan: Mhm Tahsin: And of course, the last point I'll make is that technological changes don't happen overnight, right? So, implementing new technology takes a long time. It took decades for computers and mobile phones to be widely adopted across society. So, studies on technological transformation are meant to be forward looking, but they're based on the information we currently have available. So, the economy's dynamic. Like that saying goes, change is the only constant. Right? Tegan: Yes. Tahsin: So, AI's capabilities will grow over time and the task we do in our jobs, they could evolve over time as well. Tegan: In our last episode, we learned that Canada has a little bit of a productivity problem. Could AI be just what we need to boost Canada's productivity? Tahsin: Yeah, so like you pointed out, Canada's productivity growth, labour productivity in particular, has been lagging behind its peer countries, like the U.S. for a while now. Naturally, many people are wondering if AI, which could be thought of as a new form of capital now can be a force for good and boost productivity. It's certainly possible that AI could have far reaching impacts on productivity and by extension on people's income. But there's no straightforward answer at this point. So there's no clear consensus regarding how AI might affect productivity, if at all. Some analysts are calling AI a game changer, the new industrial revolution, if you will, and comparing it to the invention of the computer. But at the same time, there are other analysts who are saying that the AI bubble might burst and that AI is simply overhyped. Tegan: I’ve heard that too… Tahsin: And it might not amount to any meaningful productivity increases at all. So we just don't have a full picture yet as AI adoption by Canadian businesses is still in its early stages, with less than a tenth of businesses reporting that they use AI. We also have to remember that our productivity growth, right, it partly depends on how other countries are doing. For example, the IT boom, right, it's led to tremendous growth for the U.S. in the 1990s, but its effect wasn't as pronounced in some other countries. We could be in a similar situation with AI because adoption rates of AI will vary across countries. And even if AI doesn't affect productivity overall, there could still be distributional impacts with certain segments of the economy benefiting more than others. Tegan: Is that what distributional impact means? Tahsin: Yeah, exactly. So what I mean is… take, for example, the digitally intensive sectors, right? These IT firms, they're already in a better position right now to adopt AI. In fact, you see that in the data. So, sectors like professional, scientific, and technical services have higher adoption rates right now of AI. This might this is what I mean when I say distributional impact. Right? Some sectors of the economy could see a benefit, but that's still unclear, though. Tegan: Gotcha. Gotcha. I'm just going to throw another couple problems at you, things that we've explored in the past in our show. Tegan: What about climate change and sustainable jobs? What might AI’s role be in that area? Tahsin: Yeah, so this is a really good question. So, there's an opportunity for AI to play a critical role in not just the economy, but also when it comes to societal and environmental issues like climate change. For example, AI might be able to optimize farming practices by analyzing data on soil, weather, and crops to recommend the best times to plant, irrigate, and harvest. This can potentially lead to more efficient use of resources and reduce greenhouse gas emissions from agriculture. AI can also analyze large data sets from satellite sensors and other sources to monitor deforestation, ice cap melting, and other environmental changes. It can also help predict future climate patterns, helping governments and organizations perhaps plan and respond to climate related events. Something worth keeping in mind here is that large-scale AI algorithms can be quite computer intensive in terms of power and energy use. So while AI offers many potential benefits in mitigating the impact of climate change, it also requires careful consideration of its own energy use to ensure that it contributes positively to sustainability efforts, essentially. So, another interesting question related to climate change is whether or not AI can foster development of sustainable jobs. So can AI be like a green job creator by accelerating innovation and fields like, uh, related to solar panels and other clean energy sources? It's possible, but since AI is still in its early stages, we just don't know yet what the net impact will be on jobs, essentially. So, it definitely has potential. Tegan: In the green space we have to be very mindful about how we use AI. Tahsin: Yeah, absolutely. I think that often gets overlooked. It's these huge sort of computer intensive algorithms that can take up a lot of energy. Right? And you have to be mindful of that. Tegan: The last kind of problem that I'll throw at you is what about income inequality? How does AI… What can AI give us on that front? Tahsin: Yeah, so this is another interesting question because in the past, automation mostly affected workers with lower education and lower earnings, but AI might be different because it's more likely to affect higher educated workers now, who are also more likely to earn more. But however, higher educated workers and higher earners are also more likely to be in jobs that are highly complementary with AI. So, if a handful of workers at the top of the earnings distribution, reap the positive effects of AI and increase their earnings potential, this can possibly increase income inequality. If wage growth remains stagnant at the bottom of the distribution. Also, if, uh, if only a handful of businesses adopt AI and are able to capture a greater market share as a result, capital income could be concentrated among a few individuals and shareholders, which might also increase income or wealth inequality. But, if AI affects higher earners in a negative way by perhaps displacing them from their jobs, It could force some of them to transition to other jobs, which may not pay as much. So, there are several scenarios to consider here. And we just, at this point, we simply can't predict the net effect at the moment. Tegan: Too many big unknowns, Tahsin: Too many variables. Tegan: Gotcha. Uh, looking towards the future, of course, you don't have a crystal ball, but maybe we can give it our best guess. New technologies are constantly disrupting the ways that we work, and there are many, many, many, many jobs that don't exist anymore because of new technologies, and I had a really fun time going through these examples and finding interesting jobs that don't exist, like switchboard operators, carriage drivers, elevator operators, pony express riders, scribes, town criers, lamp lighters, and I didn't know this, but you could be a leech collector in the past. Where do you think the potential scale of AI stands? If you were to rank, you know, kind of disruptive technologies in a general way, how big a deal do you think it is right now? And if you could muse on its future potential, what do you think? Tahsin: Those are all very interesting examples for sure. And AI is still in its early stages and businesses are only beginning to use AI. It certainly can be a disruptive force. AI also has the potential to be revolutionary in some industries, like education and healthcare. So it can shape the future in ways that are both promising and challenging. It has already fueled a lot of excitement, but at the same time, it has raised a lot of concerns as well. So this mix of opportunity and uncertainty will continue to be like a focal point of discussions in the tech business world and beyond. So right now, when most people think of AI, they're probably thinking of generative AI, like ChatGPT, which is an AI system that takes your input and responds to you. This is just one type of AI though. So in the future, there could be what's called general AI, which is a more advanced form of AI that can think and act autonomously without human input. So, in ways, we are not quite there yet, but we might get there at some point. So in the future, will AI make our life and work easier? Will all of us have self-driving cars? Well, right now there are just more questions than answers, unfortunately. But what we can do right now is continue to track the adoption of AI and its potential impact on workers, businesses, the environment, so on. But ultimately, only time will tell, though, how the impact of AI will play out. But bear in mind, like, as you said, implementing new technology does take time, so it's a long term process. Tegan: For so many of us, our jobs are a huge part of our identities, and we derive meaning from the work that we do. As AI becomes more capable of taking more and more of our tasks, do you think we might lose that sense of connection to the work that we do, or is it just a matter of adapting and finding meaning in a new AI-powered landscape? Tahsin: Yeah, this is a really interesting question as well. So, what the net impact of AI will be on jobs is unclear at the moment. It might create new jobs, it might not. But even if AI were to have no net impact on jobs, it might still have a substantial impact on the degree to which workers find their work to be meaningful. So, to give you a concrete example, researchers like myself spend a lot of time thinking about research questions, finding out what's already out there, and gathering potential data sources. So, these tasks can involve a fair amount of creativity. But if something like ChatGPT were to become better than us at these tasks and ultimately take over that aspect of research, Some researchers who value that creative process may find their work less meaningful, while others, we have to keep in mind, they may welcome that change since it can potentially save them some time. So, that's just an example using researchers, so you can imagine the array of jobs it can apply to. So, there could be, again, distributional impacts depending on the type of work you do. Ultimately, it'll depend on how powerful AI will be or might be in the future and how workers and businesses harness that new technology. Tegan: So many variables. Tahsin: Yes. Tegan: Um, how can education systems… and I got I'm not gonna lie when I started researching for this, I got super distracted because I didn't realize how big a problem this is in schools right now. All these… If you… AI is a whole thing if you are an educator, but how can education systems adapt to prepare students and future workers for this AI-driven future and specifically job market? Tahsin: Yeah, so another good question. So, an education centered around soft skills is probably key. So, AI can perform many technical tasks, but skills like communication, problem solving, emotional intelligence, and adaptability are harder for machines to replicate. So digitally, uh, digital literacy remains crucial. AI is getting good at writing code and doing other tasks we might do on a computer. But it's still susceptible to making mistakes because AI is only as good as the people who develop it. So understanding the basics of digital technology, its role in society and learning to view it through a critical lens is increasingly important. The rapid pace of technological change also means that learning can't stop after formal schooling. So lifelong learning needs to be emphasized, whether in the workplace or in the formal schooling system. Lastly, it's crucial that students along with everyone else learn about the ethical implications of AI because AI can be biased. It can make mistakes. It can have legal implications. So, for students, it also raises academic integrity issues. So, learning to use AI responsibly is probably a skill in and of itself Tegan: Thank you for your time and sharing your expertise with us. Tahsin: Thanks so much for having me, Tegan. Tegan: Predicting the future is hard. As Tahsin says, there are just too many variables What we should work towards is a future where workers have the training they need to do their jobs and are paid good wages to do so, which is just what the original Luddites wanted all along. You've been listening to Eh Sayers. Thank you to today's guest, Tahsin Mehdi. If you’re interested in the future of AI and its potential impact on the workforce of the future, check out the article released September 3 called, “Experimental Estimates of Potential Artificial Intelligence Occupational Exposure in Canada.” Additionally, a follow up was released September 25 called, “Exposure to artificial intelligence in Canadian jobs: Experimental estimates.” You can find both on the Statcan website. You can subscribe to this show wherever you get your podcasts. There, you can also find the French version of our show, called Hé-coutez bien! If you liked this show, please rate, review, and subscribe. And thanks very much for listening!