Markets and Economy

Market Conversations: Is artificial intelligence coming for our jobs?

Graphic of a hand

Investing is all about possibilities. And artificial intelligence (AI for short) presents many. This powerful and misunderstood technology can help companies and their workers be more productive across the economic spectrum. However, thanks to Hollywood portrayals, some fear AI could lead to the end of humanity, or at least the end of their job. People fear what they don’t understand. Most technologies are resisted until they’re embraced.

Ashley Oerth, a Senior Investment Strategy Analyst at Invesco, joins this episode of Market Conversations to add context to these concerns and shed light on what AI can and can’t do. She talks about three types of companies that may benefit from AI: enablers, adopters, and responders. What are the obstacles? What are the opportunities?

Here are some of her thoughts:

  • Artificial intelligence is a nebulous concept that carries different meanings for different people. In the most basic form, it mimics human intelligence and decision-making by processing and categorizing data to respond to a prompt. This narrow AI is designed to do something specific.
  • With the rise and release of ChatGPT, generative AI seemed to take the world by storm. The technology has been around for a while. But more computing power and better data have led to incremental improvements. These newer systems give the impression of understanding and creativity.
  • Generative AI is essentially a prediction model. In terms of text, a chatbot can predict the words that flow from a question. It’s trained to respond with the most likely words. But the answer won’t necessarily be the same every time. The tools are probabilistic rather than deterministic, meaning the sort of dice roll that happens each time leads to different responses.
  • Generative AI systems have proved capable of creating in mediums besides text, including images, audio, and video. The technology has moved beyond science fiction to real-world applications that the public can experiment with. The tools are flexible enough to respond naturally in many different scenarios. This has created excitement and apprehension.
  • It’s not clear if AI will ever be able to do any of the more intellectually challenging jobs. The technology can learn, and it can create increasingly convincing facsimiles that give impression of thinking. But AI doesn’t have ideas. It can’t critically analyze a problem. The technology recognizes interconnected topics based on training data. Rather than taking over for knowledge workers, the tool will more likely be paired with them as part of their regular workflows.
  • Ashley described three types of companies that may benefit from AI: enablers, adopters, and responders. Enablers are companies with products and services to help build and train AI models, such as producers of semiconductors, mega-cap tech names, and owners of large stores of data. Adopters are companies that can integrate AI models to create efficiencies in their businesses. Responders are companies that react to AI’s implications in society.
  • Companies involved in the AI craze have created some euphoria in the market, illustrated by earnings and valuation. In terms of earnings growth, mega-cap tech names have been marked up about five percentage points, which Ashley thinks look reasonable.1 Valuation has moved from 36 times earnings earlier this year to 51 times earnings.1 While that’s a bigger jump, it’s not necessarily outlandish, Ashley said.

Transcript

Brian Levitt

I'm Brian Levitt.

Jodi Phillips

And I'm Jodi Phillips. And we're talking artificial intelligence today. Ashley Oerth is here. She's a senior investment strategy analyst at Invesco. So Ashley will be here to make sense of the optimism and the fear surrounding AI. So Brian, which side are you on - excitement or fright with AI?

Brian Levitt

Yes. Is that okay?

Jodi Phillips

Yeah, sure. Great answer. Probably most people would echo that, but yeah, no, I think that's pretty common.

Brian Levitt

Yeah. I'm not sure if I even know enough yet to be excited or frightened, but yeah, I'm still trying to get my head around it. I think everybody else is as well. I can look to certain things, like if you were to ask me, am I excited about autonomous cars that get safer and safer over time, then yeah, sure, of course I'm excited about that.

Jodi Phillips

You sound like the father of teenage girls.

Brian Levitt

Yes, no doubt. No doubt. And my oldest one will be 16 next year, so that'll be really front and center in our minds. Very real.

Jodi Phillips

Yes. Well, as the mother of teenage boys, including a 17-year-old, I 100% agree with you that safer cars would be an amazing, amazing development. And look, beyond that, really exciting possibilities. You think about the medical field, robotics and hospitals, predictive software that can diagnose diseases earlier. That's just amazing.

Brian Levitt

Yeah, exactly. And look, so the possibilities can boggle the mind. And I know deep in my soul, everything about history tells me that I shouldn't fear technology. And so I'm very much, Jodi, pushing back against any instinct to be frightened. You look at history, it's always hyperbole. It's always overblown.

Jodi Phillips

Yeah. Well, it's like the quote I read the other day, right? "Once a technology rolls over you, if you don't get on a steamroller, then you're part of the road."

Brian Levitt

Yeah, it's so true. It's so true. So you want to lean into it. Although I will tell you, I'm not going to watch the Terminator again anytime soon.

Jodi Phillips

Oh, yeah. We were warned about this back in 1984, and we didn’t listen.

Brian Levitt

We were. But look, as we're saying, almost all technologies are initially feared until they are embraced. And typically what you see, standards of living generally climb as a result of it. Concerns of mass unemployment have historically not materialized and of course-

Jodi Phillips

No, in the US what is it, 3.6% unemployment?

Brian Levitt

Yeah. So technology's not killing all the jobs and the human race persists. But I think what investors want to know beyond-

Jodi Phillips

Thank goodness.

Brian Levitt

... all of this is how do they prosper from AI? How do they identify the types of businesses that will benefit from this?

Jodi Phillips

Absolutely. And that's why we're so happy that Ashley's here. She's going to put this all into the proper perspective for us and help us to think about the investible opportunities. Ashley's got a framework to help us categorize companies that are directly and indirectly involved with AI. And I think that's going to be really helpful to wrap our arms around all of this.

Brian Levitt

Ashley, welcome to the show.

Ashley Oerth

Thank you so much for having me.

Brian Levitt

Yeah, I promise you that I'm not a cyborg from the year 2029 sent to hear your best investment ideas.

Jodi Phillips

Well, that's a movie pitch right there. I'd watch that.

Brian Levitt

How far away did 2029 seem when we first watched The Terminator?

Jodi Phillips

All too fast.

Brian Levitt

All too fast. So Ashley, why don't we start, what is all this? What is artificial intelligence? What does it mean to you?

Ashley Oerth

Sure. So artificial intelligence, I think it's one of those words similar to so many we've heard in the not too distant past of metaverse and cryptocurrencies and all this, that it carries a lot of meaning, but we don't really know exactly what that is. So artificial intelligence, it's a pretty nebulous concept, but in its most basic form, it's really about mimicking some kind of human intelligence or decision making. It's really about helping us process and categorize data, make decisions based on available data, or even create new data, as we're seeing today, based on some kind of prompt. So really what we have today, it's not the Terminator, it's not HAL from a Space Odyssey, it's really what we call narrow AI. It's task specific, it's designed to accomplish something in particular.

Brian Levitt

How did we forget a Space Odyssey?

Jodi Phillips

Yeah, that's a classic reference for sure. But Ashley, so what's driving all the excitement now? We're making all these old school references and we've been talking about AI since, I don't know, what, the '50s or so? So what is it about today? Why is it all of a sudden, or at least it feels like all of a sudden, everywhere you look?

Ashley Oerth

So we're excited today because of generative AI. It's really this topic that has taken us by storm since the release of ChatGPT late last year. Really, this tech has been around for a while, but really through this combination of incremental gains and computing power, greater data availability, better models over time that have really just been incremental improvements, we're now able to have these generative AI systems that are able to match human capabilities in natural language and a whole host of other possibilities.

So what we have today are these systems that are able to, for example, pass the bar exam or score well on the LSAT or the GRE. And we have similar systems as well, not just for text, but also for images, for audio and video, all sorts of capabilities that are cropping up and the capabilities are impressive. So I think that's why people are excited is because suddenly we have these tools that they're not the stuff of science fiction, they're the stuff that we can go online and play with at any given moment. And I think that the possibilities are boundless, but also I think there's a great deal of fear that comes with that. So possibilities plus fear, I think is excitement, right?

Brian Levitt

Yeah, exactly. And is this different than Deep Blue beating a chess master in the 1990s? Or Jodi, do you remember when IBM Watson was on Jeopardy and-

Jodi Phillips

Yes.

Brian Levitt

... and was doing quite well? Is this all that different? Have we made huge leaps and bounds since then?

Ashley Oerth

So in those cases, I would say AI was really purpose built. So you mentioned the examples of Deep Blue and of Watson. So these tools were really designed for that task at hand. They were within that context of narrow AI that I mentioned, they were even more narrow than what we have today. So things like these large language models that we've been hearing about and have been able to play with since late November, these are exciting because they are quite flexible. They're able to understand and respond in natural human language. And it is something that I think seeing is believing. You're able to play with these things and they're able to write you a poem or write you a paper or summarize a document or all sorts of everything from menial tasks to things that are more, I think, intellectually demanding.

Jodi Phillips

That's right.

Brian Levitt

It's pretty remarkable.

Ashley Oerth

It's amazing.

Brian Levitt

A friend of mine was having a religious service for his daughters, and we asked for a speech and it spit out a beautiful speech for him. I don't know if he used all of it, but it was almost too lovely to use all of it. But Jodi, you're a writer. Are you using the shortcuts now? Is the great American novel by Jodi Phillips coming from ChatGPT?

Jodi Phillips

No. No, not at all, although I am mindful that the more I write, apparently that helps ChatGPT get smarter. And Brian, you have a monthly column, Above the Noise, and you occasionally do a segment in there that I really like where you ask ChatGPT a question and kind of critique its answer compared to how you would answer it. And I think in most cases it was maybe a little off base, not quite the full story, so it's got a lot of room to improve. So Ashley, when does that happen? When can ChatGPT just write the whole column or write my whole book?

Ashley Oerth

So you know what they say, Brian, right? Prediction is very difficult, especially if it's about the future. And really to me, it's not clear if it ever will be able to do our jobs. I think we can create increasingly convincing facsimiles of our jobs with AI that can sort of give the impression that it's able to think and learn, but ultimately there's not a whole lot of deep thought that's going on here. In other words, AI can learn, but it can't think, it doesn't have ideas. It can't really critically analyze a problem. It can really, at best, give the impression of ideas by recognizing interconnected topics based on the training data it was initially trained on.

So that said, data science, it's really a field that's been developing at a breakneck pace for quite a while now, and predictions about future capabilities are often exceeded, and the timeline of them is something that maybe will say, oh, this will happen in five years, but it ends up happening in two, or maybe nothing happens for a decade, but then suddenly everything happens in two years. So I think that the most likely outcome right here, is that AI, I think, will be used as a tool paired alongside knowledge workers as part of our regular workflows, rather than something that really just takes our jobs. That's my prediction, but of course, we could all be very wrong about what the timeline is here and what its ultimate capability is.

Brian Levitt

I love that Ashley assumes that there's deep thought going on here or in the rest of the US workforce.

Jodi Phillips

Very optimistic point of view.

Ashley Oerth

I have aspirations for our lives.

Brian Levitt

Do I need to know how it works or am I just going to be harnessing the... I don't really know how the World Wide Web works. I'm not really sure I know how my telephone works, so do I need to know how it works or it's just that these are going to be tools that I'm going to harness?

Ashley Oerth

So it's similar to what you just described with the phone. It's something that you can appreciate how it works, but it doesn't necessarily change how you interact with it. So I think that when we're talking AI, everything that we're talking about today is really centered on this generative AI topic, and I think there's a lot that's exciting here. So when we think about exactly how it's working, it's essentially a prediction model. If we're using the example of text, if we ask one of these chat bots a question that it's able to predict the series of words that flow from that. So if you ask it, how are you? It's going to, based on the training data it's seen before, sort of throw at you what the next most likely words are from that. So I think what's exciting about what's going on here is that these models, they're not just giving you the same response every time, in the parlance of the space, they're not deterministic, they're not always arriving at the same output given some kind of prompt or input.

So in other words, they're probabilistic. There's a sort of dice roll that's happening every time there's a new word that we're getting new content that flows from that. It gives us the impression almost of creativity. So if we take this idea and apply it to a model that has been trained on a mindbogglingly huge amount of data, we get this sort of large language model that's able to understand natural language, understand topics, and provide intelligent sounding replies with variety. So if we can ask it to write a screenplay or write an academic paper or whatever, each time that we do that, we'll get a different output because it is probabilistic, which I think is one of the really cool things about this generative AI craze, is that there's so many things that can come from it that, again, it can feel like creativity and we can make use of that.

Brian Levitt

Now, I've heard that ChatGPT may have been getting dumber. Is that true?

Ashley Oerth

Well, I think that there's a lot of fervor to question what's going on here to try to cast doubt on the capabilities, so maybe look at that with a grain of salt. But so far there have been some studies that have suggested that because some of these models are live models, in other words, they're learning over time, given how people interact with them, that then maybe that's a commentary on society.

Brian Levitt :

Yeah. It's idiocrasy.

Ashley Oerth

… that it’s gotten dumber over time, which go figure on that. But it has been documented that on certain tasks that performance has degraded in certain categories, but in others it's actually improved. So maybe this is a challenge for engineers to figure out how exactly to wrangle how exactly these models we're learning.

Jodi Phillips

Putting it in that context, it's a tool, it's a predictive model, what it actually is versus what people either hope or fear it could be, understanding that, do you feel like the market's become too excited about it in that context? Is the excitement that the market is showing, do you feel like that's appropriate for the potential or how do you view that aspect?

Ashley Oerth

So that's a tricky question for sure. I think that from what I've seen year to date, I'm feeling like the euphoria is there. I try to look at any sort of tech trend or anything that's driving the markets from two perspectives. So on the one hand, how reasonable is the growth that we're pricing in? So what are the earnings estimates of the companies that are pushing up the markets? And then two, what price am I willing to pay for that? So we've got earnings on the one hand and the valuation we're paying for it on the other.

And so from the earnings growth side of things, we have seen companies that are involved in this AI craze be marked up about five percentage points. If we look at some of the mega cap tech names since the release of ChatGPT, which that's not too crazy. So that's five percentage points over the next three years. That's a compound annual growth rate there. So again, seems reasonable. And then on the valuation side of things, if we think of it from the price to earnings perspective, we've really moved up from about 36 times earnings earlier this year to 51 times earnings on a trailing valuation perspective. And then on forward PEs, we've also moved up from around 32 times earnings to 37, which-

Brian Levitt

And that's on the mega cap growth names?

Ashley Oerth

These are the mega cap tech names.

Brian Levitt

The mega cap tech names.

Ashley Oerth

The typical FAANG names that we like to pick on. And so you have to ask yourself, do you believe that earnings growth, and again, if so, are you willing to pay for that? And I think the earnings growth has been marked up, but so is the valuation. So I think that from the sort of perspective, yes, it has moved up in price and I think that it's gotten quite expensive, especially if you look at these names, but it doesn't seem too outlandish.

However, in the context of all of this, we've got rising interest rates, we've got a backdrop that's macroeconomically speaking fairly weak. So at this stage I'm sort of thinking, okay, maybe that's a bit expensive to get it on this trend, but maybe you could say that it's just been priced in.

Brian Levitt

Now I'm old enough to remember the craze around the dot coms and the original launch of the internet. And of course some of those businesses were famously overvalued, and some of them of course did disappear. But yet there was a lot of way to profit and a lot of ways to take advantage from this new platform that was going to connect billions of people around the world and change really how we do everything. So regardless of cyclically whether it's expensive, how do you think about the structural investment opportunities and what type of businesses should investors be watching?

Ashley Oerth

Yeah, so I think that this is always tricky to think of who's going to win, who's going to lose, and over what timeframe. If you go back to the tech bubble days, a lot of the ideas that were at play there eventually did play out. It's just it was a bit ahead of its time, that the rest of …

Brian Levitt

Right. You had to own Amazon eventually, not pets.com.

Ashley Oerth

Yeah, exactly.

Brian Levitt

But I've heard you categorize the types of businesses in this space. I'd love to hear you talk through that.

Ashley Oerth

So the sort of buckets that I put all of these investment implications, if you will, there are sort of three categories of business that I think broadly speaking would benefit from this AI craze. So on the one hand, and I think we've already seen a lot of this, are the enablers. So this is everybody from, if we go to hardware, so the hardware that's used to train these AI models, so if you think semiconductors, those are in the sorts of enablers or picks and shovels approach, if you will. And then we also have those companies that are building the models themselves. These are often, again, the mega cap tech names that really have the development capabilities to make this happen. And also companies that have large treasure troves of data. If data is the new oil in our information economy, then you're well positioned for being somebody who can develop a differentiated AI model.

So those are the enablers that I see behind this whole AI craze. The second bucket would be sort of the adopters. So those are the companies that are able to use these AI models that have been built and integrate them into some kind of part of their business, whether that's their product or how they actually run themselves. Maybe it's internal efficiencies that they can gain. There's a long list of possibilities of how exactly this can be applied and in different sectors. And then on the third bucket, I sort of view this as responders. So AI brings all sorts of new threats that society must address, and we have companies that can also use AI themselves to respond to that. So I think that's a third bucket that can perhaps benefit from this AI trend. So there you have it, you have the enablers, you have your adopters, and then you have your responders.

Jodi Phillips

When you're thinking about those buckets, are there any that you think are, I don't want to say better than others, or just a better position to be in than others? Or are there buckets that you're watching particularly closely to see how either the adoption plays out or the enablement plays out? What are your thoughts in terms of that?

Ashley Oerth

Yeah, so I think that from what I just laid out there, you can sort of view it almost like a timeline. So in this theory I've laid out, the enablers would benefit first, and I think we've already seen a lot of that in the price action so far. Then the adopters would be those companies that are able to actually make use of AI. And I think that we're seeing the beginnings of that, although it's still early stages.

Brian Levitt

And that could be pretty much anyone in any sector or industry. We started this talking about autonomous vehicles or robotics and hospitals, that could go to even some people like us analyzing markets, writing emails.

Ashley Oerth

Absolutely.

Brian Levitt

Yeah. So that's a broad bucket.

Ashley Oerth

That's right. And I think the broad bucket, broadly defined like that, it's done that way for a reason. We have so much that AI can impact that it'd be a mistake, I think, to just focus on one particular sector. I would say though, from studies I've seen on automation and in general, they tend to focus more on the information economy and less on more manual tasks. So perhaps those adopters are those that are less manual labor and more information economy, which is, in the US at least, some a hundred million jobs. So it's a pretty large chunk to sift through.

Brian Levitt

Now, let's go with an FDR (Franklin Delano Roosevelt) quote here, "The only thing we have to fear is fear itself." Is the only thing we have to fear, fear itself? How fearful should we be when... You hear some of these people who have worked in AI over the course of their lives say, look, we got to slow this down. There's big challenges that face humanity here, and you already have the Biden administration speaking with some of the leaders of those mega cap growth companies that you had mentioned to try and put some parameters around this. Do you have concerns?

Ashley Oerth

So I do have concerns, and I think that my concerns are less focused on what people normally talk about, which is this going to replace me? And it's more about what its implications are for society at large. So my biggest fear is really about how AI can be used for malicious purposes. So you've probably heard of things like deep fakes, voice mimicking, image manipulation, automated code generation, and all sorts of threats that are brought on by generative AI.

And there's already been examples of this. We had last year, deep fakes of Ukraine's President Zelenskyy. This year just in May, we had a faked image of an attack on the Pentagon that briefly moved markets on a morning late in May. And these threats are real. I don't think as well that our tools as a society are really evolving fast enough to appreciate and tackle those problems. We're already struggling with how to handle the internet, cryptocurrencies, misinformation, and all sorts of other challenges. And I think these problems will only add to that pressure.

Brian Levitt

And this whole idea that the machines will rise up, is that just science fiction nonsense I joke that I'm not a cyborg from 2029, but I do get questions from investors about the fate of humanity. Are you unwilling to even go there in your mind?

Ashley Oerth

I'm not worried about the fate of humanity. Maybe we could all live in a WALL-E world, maybe the good parts of the WALL-E world, maybe not so much the other side of things, but I think that the risks to what this means for humanity, it's not like we're going to have something that's in control of the nuclear codes or something like that, that we have some kind of tool like HAL that's able to go rogue and cause all kinds of mayhem, rather these tools are really built for a particular purpose. Their abilities are limited to a specific set of functions. It's not like we can just give them free rein over whatever they want to do, right?

Brian Levitt

Right.

Jodi Phillips

So Ashley, tell me, we've talked a lot about the capabilities and what AI is and isn't, but what excites you the most? What are you most looking forward to watching develop as time goes by?

Ashley Oerth

Yeah, I think like we've been talking about, there's a lot that people I think are nervous about, but there's a lot I think to be really excited about. So for example, if we're thinking of generative AI involved in our day-to-day work, this could mean faster summarization of content that we're looking to read but don't have time to get to. It could mean helping us with task prioritization. In essence, we're kind of getting this personal assistant that could be embedded into our work streams that really helps alleviate distractions and enable the kind of knowledge work that our livelihood center around.

There's this author Cal Newport who's really written at length about this idea that he calls deep work. And his argument really is about how in today's knowledge economy focus is a sort of precious commodity, but really what we see happening is evermore notifications and things that are distracting us that pull away from our ability to do real quality knowledge work. So in other words, every interruption costs us, whether it's our phones or an Outlook email or other distraction. And if AI can be integrated into our work streams to help minimize those sorts of distractions and alleviate menial work, take care of those rote tasks and allow us to focus, I think we can all be more productive. So that's what I'm excited about with AI, that it can really make us more productive in our day-to-day jobs and help us grow the economy and ideally our livelihoods.

Jodi Phillips

So Brian, how are you feeling on your own personal fear/excitement scale? This helps?

Brian Levitt

Yeah, of course it helps, and I love listening to Ashley, and like I said in the intro, I'm pushing against my instincts for fear because I really liked your quote. I can't get it exactly right, but I'm either going to be steamrolled into the road on this, or I'm going to get on board and I'm going to get on board, right? I'm going to look for opportunities to best take advantage to make my life and my career more efficient, and I'm going to look for opportunities on how to invest and take advantage of what I think is a strong long-term structural theme.

Jodi Phillips

Yeah. And I might try to write my book a little faster just in case. Ashley, thank you so much for joining us and help putting all this in perspective.

Brian Levitt

Ashley, thank you.

Ashley Oerth

Absolutely. So great to be here. Thanks so much for having me.

 

Important Information

NA3070198

Recorded date: July 26, 2023

Image: d3sign / Getty

Some references are US centric and may not apply to Canada.

The opinions expressed are those of the speakers, are based on current market conditions as of July 26, 2023, and are subject to change without notice. These opinions may differ from those of other Invesco investment professionals.

This does not constitute a recommendation of any investment strategy or product for a particular investor. Investors should consult a financial professional before making any investment decisions.

Should this contain any forward looking statements, understand they are not guarantees of future results. They involve risks, uncertainties, and assumptions. There can be no assurance that actual results will not differ materially from expectations.

All investing involves risk, including the risk of loss.

Discussions of specific companies are for illustrative purposes only and should not be considered buy/sell recommendations.

In general, stock values fluctuate, sometimes widely, in response to activities specific to the company as well as general market, economic and political conditions.

Many products and services offered in technology-related industries are subject to rapid obsolescence, which may lower the value of the issuers.

All data sourced to Invesco unless otherwise noted.

Commissions, trailing commissions, management fees and expenses may all be associated with mutual fund investments. Mutual funds are not guaranteed, their values change frequently and past performance may not be repeated. Please read the simplified prospectus before investing. Copies are available from Invesco Canada Ltd. The opinions expressed are those of the presenter, are based on current market conditions and are subject to change without notice. 

These opinions may differ from those of other Invesco investment professionals. Forward-looking statements are not guarantees of performance. They involve risks, uncertainties and assumptions. Although we make such statements based on assumptions that we believe to be reasonable, there can be no assurance that actual results will not differ materially from our expectations.

Invesco is a registered business name of Invesco Canada Ltd.

Invesco® and all associated trademarks are trademarks of Invesco Holding Company Limited, used under license.

© Invesco Canada Ltd., 2023

FDR (Franklin Delano Roosevelt) was an American statesman and politician who served as the 32nd president of the United States from 1933 until his death in 1945.

The US unemployment rate was 3.6% in June 2023 according to the US Bureau of Labor Statistics.

LSAT stands for Law School Admission Test.

GRE stands for Graduate Record Examinations.

Information on tech company earnings and valuations is from Bloomberg, L.P., as of July 25, 2023

P/E stands for price-to-earnings ratio, which measures a stock’s valuation by dividing its share price by its earnings per share. Forward price-to-earnings ratio is calculated by dividing the company’s current share price by its expected earnings, usually for the next 12 months or next full fiscal year.

FAANG is an acronym that stands for Meta (formerly known as Facebook), Amazon, Apple, Netflix, and Alphabet (formerly known as Google).

Footnotes

  • 1

    Source: Bloomberg, L.P., as of July 25, 2023