Episode Transcript:
Marcus Johnson:
This episode is made possible by Walmart Connect. If you expect your media investment to deliver clear, measurable results, Walmart Connect can help you get there. With powerful analytics and the reach of America's number one retailer, their closed-loop measurement uses Walmart's proprietary customer purchase data to track the impact of your campaign on sales not just on Walmart's site and app, but in-store as well. For some campaigns, they can even provide rest of market data that tracks the impact on sales at other retailers. Visit walmartconnect.com today and see how they can help make your media spend meaningful. Walmart Connect. More than media, meaningful connections.
Jacob Bourne:
It might come down to not so much what AI can do and what it's capable of, but do we actually just want people in certain roles versus an AI. Is there something valuable about just having a human do a certain job? And I think it's going to come down to that eventually.
Marcus Johnson:
Hey, gang. It's Tuesday, April 23. Gadjo, Jacob, and listeners, welcome to the Behind the Numbers Daily. It's an eMarketer podcast made possible by Walmart Connect. I'm Marcus. Today, I'm joined by the same two gents who hung out with me yesterday. I'll reintroduce them just in case you weren't there. We start with our senior technology analyst. He's based in New York. It's Gadjo Sevilla.
Gadjo Sevilla:
Hello, Marcus.
Marcus Johnson:
Hey, Fella. And we also have one of our technology analysts, but based in California. His name is Jacob Bourne.
Jacob Bourne:
Hey, Marcus. Happy to be here.
Marcus Johnson:
Hey, fella. So gents, today we're talking about how gen AI could disrupt jobs this time. Yesterday, it was the economy. In part two, we're talking jobs.
But we start with the fact of the day. Which musician has made the most money ever? Yesterday, it was athlete. Today, it's musician. Any guesses? I'll give you a hint. Gadjo, you are close to their birthplace. Actually, maybe you're not. I don't know exactly where you're... But you're closer than Jacob. Hip hop artist. V, want to bail them out?
Gadjo Sevilla:
Is it Jay-Z?
Marcus Johnson:
Bang. Doesn't need it.
Jacob Bourne:
Really? Really?
Marcus Johnson:
Jay-Z.
Jacob Bourne:
Wow.
Marcus Johnson:
$2.5 billion. His wife-
Jacob Bourne:
Interesting.
Marcus Johnson:
... Beyonce, is also in the top 10. She's eighth, but she does have him beat in Grammys. Uh-oh. She got the most ever, 32 to his 24. I'm sure it's the point of contention in the household. Rihanna is the musician who's made the second highest most money with 1.4 billion. And Sir Paul McCartney of the Beatles is third with 1.2. Taylor Swift is 1.1, and Sean Diddy Combs rounds out the top five with a billion. Anyway, today's real topic, gen AI's economic impacts on jobs.
In today's episode, first in the lead, we will cover how the rise of AI will impact jobs. Then for In Other News, we'll discuss Apple's gen AI strategy and whether AI has a measurement problem. But we start, of course, with the lead.
Christopher Mims of the Wall Street Journal and others are convinced that we are now entering a new kind of industrial revolution that folks are calling the cognitive revolution. The automation of knowledge work. And we read a ton for this episode, but one of the pieces that jumped out was from EY-Parthenon. Chief Economist Gregory Daco recently asked whether fears about AI-driven mass unemployment are warranted. So a small question to kick things off. Gadjo, are they warranted? Is AI going to put a lot of people out of... Most people, I guess mass, that's what that means, out of work?
Gadjo Sevilla:
I think in certain industries, maybe more than others, we're already seeing mass unemployment attributed to AI especially in customer support, programming, and various services as they're replaced by AI chatbots. So companies are trying to justify their huge AI investments. And it varies in industry, but we're seeing press releases of large investments followed weeks later by announcements of layoffs from the same companies. So it's not only happening now, but it's happening quite quickly. So I think that will change. There'll be some realignment at some point, but yeah, it's just what we've seen so far.
Marcus Johnson:
It definitely depends on the... Depending on the industry is such a good point. There's a Goldman Sachs report. Close to half of office and admin support was exposed to AI automation, and the most of any employment field compared to just 6% of construction, 4% of repair jobs, and 1% of building maintenance.
Jacob, industries shift and so some folks might be saying to themselves, "Well, agriculture went from half of the US economy to less than 10% in about 60 years from 1880 to 1940. So maybe these fears of mass unemployment driven by AI are overblown." Where do you land?
Jacob Bourne:
Yeah, I don't think that they're overblown at all. And like Gadjo said, we're already seeing evidence of it. Now, of course, the job market is still very robust in the US at least. But over time, I think the thing is that yeah, it's going to vary by industry, certainly. But the more advanced AI gets, the more jobs will be exposed to it and then the higher the risk for more different types of workers. I mean, even some CEOs are now starting to worry that their jobs could be at risk due to AI and its ability to make decisions.
So I think we hit a point where there's, like Gadjo said, a realignment, a changing of perspective on the issue in terms of how are we going to implement this in a way that actually makes sense. It might come down to not so much what AI can do and what it's capable of, but do we actually just want people in certain roles versus an AI. Is there something valuable about just having a human do a certain job? And I think it's going to come down to that eventually. And there might be some government intervention as well in terms of there could be a taxation on replacing workers, for example-
Marcus Johnson:
Oh, interesting.
Jacob Bourne:
... with AI or robotics.
Marcus Johnson:
Yeah, yeah. Could just see pushback from the labor market from people saying, "This isn't acceptable." And so some companies might respond to that by saying, "Okay, we're not going to go as far as we'd like to to save money." And you've got to try and find that balancing act.
A lot of this depends on, you mentioned industry, also the job. AI can do the easy jobs better than people for now. Maybe it can do the other ones better later, the harder ones. But the US Council of Economic Advisers' annual economic report for the president estimates 10% of US workforce jobs fall into this category of simple tasks which an AI can perform more efficiently than a person does.
Last summer, YouGov asked, "Which tasks do Americans think AI will be better at than professionals?" Number one was translating languages. That was over 50%. And over 40% were folks saying predicting weather patterns, providing tech support, creating software, designing buildings. Of the 30 jobs listed, there were only four jobs where the majority of folks thought AI would do worse than humans. But to your point, Jacob, they were in fields where you probably want to stick with a human even if it could, in theory, do it better than a regular person. Ethics, therapy, emotional companionship, and providing child and elderly care. Most of the other ones were pretty balanced between AI doing it well and a human doing it well though.
Mr. Daco, who, from EY-Parthenon was saying that technological innovations can affect labor in three main ways, which we touched on in yesterday's episode. Job creation. So the rise of the internet brought web design, digital marketing, et cetera. Job displacement. So some of those routine tasks are affected, so assembly line work or data entry are susceptible to automation. And then job transformation. So new tasks might emerge. Bank tellers handling less cash because of ATMs instead talking to clients about new financial products. So that's how you could see things. It's not just that robot appears, someone gets traded for that robot. There are a variety of ways that this could affect the labor market. But if we zoom in on the ad space specifically, Jacob how will ad industry jobs change in your opinion?
Jacob Bourne:
Well, I think the ad industry is one of the industries that we're seeing the most immediate effect from generative AI in terms of job losses and just pretty rapid adoption of technology. And for the jobs that do remain in that industry, I think that using the technology will become a basic requirement of the positions. You can create campaigns and execute them just much faster with AI than without. But I think there's a caveat there.
And the thing about generative AI to remember is it's still very, very new, at least on the commercial front. So how do AI-generated ads perform relative to non-AI generate ads? Well, there might be some sense of that, but to my knowledge there haven't been studies done to examine that. And so I think over time, we're going to see a lot of research on generative AI's effect on things like advertising and a great many other things. And I think that that could steer adoption one way or the other.
Marcus Johnson:
Yeah. I mean, Gadjo, to Jacob's point about gen AI being a basic requirement of doing an ad industry job, there was some research from Basis Technologies, which was in an Adweek piece, cadence for using gen AI as part of the advertising workflow. You've got a quarter of people say they never use it. You've got a quarter saying once or twice a month, but you've got 11% of people saying every day and you've got 15% of people saying at least three to four times a week. So it does seem like it has found its way into the ad industry rather quickly, and folks are using it more so than maybe we would've expected given this technology officially, or unofficially, is a year and a half old. What do you think of how it's impacted, and going to continue to impact, the ad industry in particular?
Gadjo Sevilla:
I think for creative aspects of the ad industry, it's really very compelling because it could speed up a lot of the processes. And if the campaigns require quick turnaround, if you need to spit out content for social, for different mediums, it could be applied for that. And the key demo for AI is content creation, so that includes audio, image, video. We're already seeing technologies like Sora. And now even the influencer economy, which is a huge market, that's starting to be challenged by AI influencers on TikTok as well as Instagram's chatbots, right?
Marcus Johnson:
Yeah, you mentioned content creation. That same Basis Technologies study saying digital marketing process expected to be impact by AI. 55% said content creation way out in front. Second place was analysis with just 14%.
But looking at where the gen AI is already replacing people, nearly 12% of C-suite and VPs said they either already replaced human employees with gen AI or had plans in progress to do so. That was according to that survey from Basis Technologies. They surveyed over 200 marketing professionals. But I think that gents, that's probably more folks leaving or being let go for whatever reason and not being replaced by a person and companies saying, "Okay, this person left. We're going to try to make up the shortfall using AI in some way, shape, or form." 10% of marketers said their organization has replaced human tasks with AI in the past 12 months.
Let's look, finally, at how AI could change a person's workday. I want to see where you guys land on this, because how much of your day could be automated if you really thought about it? There's a few different estimates here. One suggests that by 2030, tasks that represent up to 30% of all hours currently worked in the US could be automated by AI. That's from McKinsey. That's one. The second one is AI has the potential to automate 40% of the average work day. That's from research company Valor. And the third one is another estimate suggesting that current gen AI and other technologies have the potential to automate work activities that absorb 60 to 70% of employees' time today. That's another one actually from McKinsey. So that is yeah, just same company, different piece of research. Where do you guys land on how much of a day could be automated or... today or in a couple of years?
Jacob Bourne:
I mean, again, it comes back to how advanced AI will get. If we're looking at 2030, well there's quite a few people in the tech industry and AI researchers who think that we could very well achieve artificial general intelligence by 2030. In which case, potentially close to 90%, 100% could be automated.
But I think we focus in this conversation on a lot of the negatives, but I think there is an ideal here in terms of automation, and that is that the things that people don't want to do could get automated, like the mundane things, the things that suck away time from what you really want to do, the creative aspects of your job. And I think AI could do that. I mean, it could vary in terms of percentage based on the position. I think AI could also make that kind of automation more personalized. So everybody has an AI assistant which knows your habits and knows what your schedule is and knows what your priorities are and knows how you work, and it caters to your needs. And that would be, I think, an ideal outcome of all of this.
Marcus Johnson:
Yeah. So Gadjo, before you answer, just to set the table a bit more, Valor asked workers what share of their day they spent on 13 different work tasks. And so this is how the pie chart looks. Average amount of time spent on different work tasks. So 11% of their day was spent reading and responding to emails, same share for talking on the phone, same share for entering data. Then it was 10% for analyzing that data, and then it was 8% for in-person meetings, 8 for editing documents, 7 for managing their schedule, 7 for research and reading, 7 for virtual meetings, 6 for creative thinking, 4% for coding and software, and 9% other. So folks think through those different shares and think about how much of their day could potentially be automated. Maybe it's more than they expect. But to Jacob's point, maybe that's a good thing. What do you think of how AI could change a person's work day?
Gadjo Sevilla:
Yeah, I totally agree with Jacob. I mean, the idea that you could have maybe a virtual version of yourself that's programmed into an AI, scours your emails, it submits a cohesive summary of important and actionable tasks, something that can populate calendars, set up calls, meetings, and eventually help compose emails and documents in your voice in an authentic way, I think that will definitely make workers more efficient.
Marcus Johnson:
Mm-hmm. I mean, to your guys' point about a lot of this is being viewed as negative, or how it's going to impact things negatively, looking at things positively, Peter Cohan, founder of Peter S. Cohan and Associates, points out that airline pilots perform a blend of simple and complex tasks, and that even though autopilot systems have been in use for more than a century, airlines still have pilots.
One thing we probably will notice a lot more, gents, though, is how much of your next position's job description will mention using AI. So people looking for jobs now, two thirds of S&P 500 companies have posted job ads mentioning AI, according to research firm, PredictLeads. Yeah, more people requiring some kind of literacy with AI to some extent.
That's all we've got time for for the lead. If you want to listen to how gen AI will affect the economy, you can listen to yesterday's episode, part one. We now, though, move to the fourth quarter of the show, today in other news. What is Apple's AI strategy, and does AI have a measurement problem?
Story One. Is Apple right not to push headlong into generative AI? The Economist points out things haven't been going fantastically for Apple as of late. Its market value is down 9% this year. It's lost its top phone maker spot to Samsung as iPhone shipments failed in Q1. Microsoft, thanks to gen AI, has vaulted it becoming the world's most valuable company. The article says folks are hoping that at this year's developer conference in June, Apple will announce whizzy gen AI upgrades, proving it can join the chatbot hype fest. But that is not how the company does things, nor should it be, the Economist says. But Gadjo, what do you make of Apple's AI or gen AI strategy?
Gadjo Sevilla:
Yeah, so if you look at Apple, they've been dabbling in AI for over a decade. They use machine learning with their cameras. Siri was the first real virtual voice chatbot. I mean, the parts have always been there.
It surprised me to find that they bought more AI acquisitions, new companies, than their big tech competitors. So they bought all these small, boutique type AI companies and they've also refocused 2,000 workers from the canceled Apple car Project Titan project to work on AI. So that's very much front and center in their mind. The issue with them, though, is that Apple's brand now is about privacy and safety, and that's not really compatible when you think about AI-
Marcus Johnson:
Good point.
Gadjo Sevilla:
... and its potential to hallucinate or just go off the rails. For a company that's priding itself in the amount of control it exerts on its product, the user experience, AI could be considered a bit of a loose cannon.
Marcus Johnson:
Yeah, yeah. One point The Economist was pointing out, it's got a massive head start because of its bank account. The Economist noting that at this stage, the vast sums needed to train AI models favor deep-pocketed incumbents over scrappy startups, which will work to Apple's advantage.
Story two. AI has a measurement problem, suggests Kevin Roose of the New York Times, asking which AI system writes the best computer code or generates the most realistic image. He says that right now, there's no easy answer because AI companies aren't required to submit their products for testing before release. Mr. Roose questions, which AI tools should folks use for a certain task? Does ChatGPT or Gemini write better Python code? Is DALL-E 3 or Midjourney better at generating realistic images of people? Jacob, does AI have a measurement problem, and what will be done about it if so?
Jacob Bourne:
Yeah, I think this partially depends on the use case. I mean, coding and image generation are two very different things. And I think part of the reason why this is difficult is because so much of what generative AI does is creative. And it's subjective, in other words, in terms of how you measure its capabilities.
I mean, I think there's some parallels to human intelligence as well. I mean, there are IQ tests. And just like there are IQ tests for humans, there are benchmark tests for AI models. But intelligence is a very broad, dynamic category. And so when you have this technology that can do so many things, it's hard to just have all these tests that are really good at measuring the different capabilities.
I think we will get there. Eventually, there will be some more industry-level testing that will come out to really give people a better idea of what tools are best. But I think ultimately, if you look at other technologies and how it's played out with consumers, I think AI is going to play out in a similar way. For example, there are search engines that perform differently, and we don't necessarily look for test results in order to decide what search engine we're going to use. We just use it and if it works for us, then we continue to use it. And I think that that's going to be the same for AI.
Marcus Johnson:
I mean, the article is pointing out there are some tests out there. The original AI test being the Turing test devised in 1950 by English mathematician, Alan Turing. That tests whether a computer program can trick a person into mistaking its responses for human. But then you've got the massive multi-task language understanding, or MMLU test. It's a SAT for chatbots or a general intelligence test released in 2020. It's a 16,000 multiple choice question tests covering a bunch of academic subjects. But even the guy who created that said it's getting too easy for AIs. That's the other problem, is you devise a test and AI figures it out so quickly, you have to update it very regularly. And there's a few others out there, TruthfulQA, HellaSwag, but they only capture a narrow slice of an AI system's power. Mr. Roose was saying none of them are designed to answer more subjective questions like, is this chatbot fun to talk to? Is it better for automating routine office work or creative brainstorming? And how strict are its safety guardrails?
I know in the past we've talked about this, but I've mentioned other benchmarks, like there's the level of vehicle autonomy scale for driverless cars going from zero, no driving automation, to five, full driving automation. So I wonder if we'll see something like that emerge, but for AI. We'll see. That's all we've got time for for this episode. Thank you so, so much to my guests for hanging out with me yesterday and today. I'm breaking down a very complex topic. Thank you to Jacob.
Jacob Bourne:
Thank you, Marcus. My pleasure to be here.
Marcus Johnson:
Thank you to Gadjo.
Gadjo Sevilla:
As always, a pleasure. Thank you very much.
Marcus Johnson:
Yes, indeed. Thank you to Victoria who edits the show, Stuart who runs the team, and Sophie who does our social media. Thanks to everyone for listening in to the Behind the Numbers Daily, an eMarketer podcast made possible by Walmart Connect. You can tune in tomorrow to the Reimagining Retail Show for the April edition of the top eight most interesting retailers of the month list.
%20scale(0.100000,-0.100000)%22%0A%20%20fill=%22%23000000%22%20stroke=%22none%22%3E%0A%20%20%3C/g%3E%0A%3C/svg%3E%0A)
