Episode 129: AI & Econimic Disruption with Avi Goldfarb

Announcer:
This is The WealthAbility® Show with Tom Wheelwright. Way more money, way less taxes.

Tom Wheelwright:

Welcome to The WealthAbility Show, where we're always discovering how to make way more money and pay way less tax. Hi, I'm Tom Wheelwright, your host, founder, and CEO of WealthAbility.

So what if you could actually predict your revenue, your workflow, the needs that you have for additional workers, and even how a worker, and whether they would fit well within your company, how would that change your life as a business owner? Today we're going to learn how AI can actually do that, that it can give us the predictability that we've been wanting without all of the headache and guesswork that comes from it, but not take away our judgment. And today I have a very special guest with me, Avi Goldfarb from the University of Toronto, he is an expert in this area, and it's great to have you with us Avi, welcome.

Avi Goldfarb:

It's fantastic to be here, thanks Tom.

Tom Wheelwright:

And so Avi, if you would, just give us a little bit of your background.

Avi Goldfarb:

Sure. I am a professor at the University of Toronto, the Rotman School of Management, I'm in the marketing department. I am trained as an economist, so way back in the 1990s there was this new technology called the internet, and when I was doing my PhD I decided this was something nobody knew anything about, and so it was the kind of thing that was worth studying and trying to get my head around. And so I spent the first 15 years of my career trying to understand the economic impact of the internet, focused on things like understanding online advertising, understanding online market power, and the challenges with respect to privacy.

Then about 10 years ago, we started this program at the University of Toronto for science-based startups called the Creative Destruction Lab. And in our lab in the very first year we saw this company called Atomwise, it was way back in 2012, that said they were using artificial intelligence for drug discovery. Put yourself back 10 years ago, that just seemed crazy. This idea that you're going to use artificial intelligence for anything wasn't on most people's radar, and for drug discovery, that just seemed like science fiction. And what they were actually doing was using a new emerging technology called deep learning, which is a branch of computational statistics, but they call it artificial intelligence, to try to predict which molecules bind with which proteins to figure out which drugs are going to work.

Tom Wheelwright:

Interesting.

Avi Goldfarb:

And then the next year we had a handful more, and the next year we had this flood of AI companies coming through our lab. And my co-authors and I, who were running a lab at the time, decided this was worth getting our heads around, and so we moved on from studying the impact of the internet to try to understand this new technology. So that's why I'm here, since 2012, and especially since about 2015, I've been focused on trying to understand the impact of AI on the economy and on business, and that led to our first book, Prediction Machines, and our new book, Power and Prediction.

Tom Wheelwright:

I love it. We did a podcast on AI just a few years ago, and I know it has just changed drastically since then. I've long been a believer that AI combined with blockchain technology was actually going to change the business world, frankly, and that one is predicting it and the other one is actually auditing it, effectively, and making sure it's accurate. So let's talk about AI, if you would, drill down a little bit, make sure everybody understands, what are we talking about when we say AI? Because I don't think we're talking about robots here, I think we're talking about something different.

Avi Goldfarb:

Yeah, the reason we're talking about artificial intelligence in 2022, and it wasn't really on anybody's radar 10 or 20 years ago, it's because a very particular branch of computer science called machine learning has gotten much better. And machine learning is prediction technology. So it's not artificial general intelligence, it's not like the robots you're going to see in science fiction, it is the ability to take data you have and to fill in missing information.

But it turns out prediction is a really big deal in all sorts of businesses, because anytime you're trying to fill in missing information, that's prediction. So if you're trying to fill in forms, that's prediction, if you're trying to diagnose in medicine, that's prediction. If you're trying to assess what the right decision is, a key component of that is situational awareness, knowing what's going on, and that's fundamentally prediction. The key point though is that prediction's not everything, so if you have a prediction you still don't know what to do. You need the prediction, to decide you actually need some judgment, you need to know what matters, what do you value in order to make a decision. And so we can think about that in all sorts of ways.

There's this old movie I, Robot. I don't know if you remember.

Tom Wheelwright:

I remember it well, Will Smith.

Avi Goldfarb:

Okay. So it's a classic science fiction movie based on a classic science fiction novel, and the protagonist of the movie, Detective Spooner, he hates robots. Why does he hate robots? And there's this flashback scene where he and this little girl are in a car accident, and their cars are both sinking into a river, and it's pretty clear that both of them are about to drown. And then a robot comes along and saves Detective Spooner, saves the adult, and not the girl, and that's why he hates robots. And he says, “Well…” Oh, because it was a robot, and actually this is really important in terms of, you said, auditing with blockchain, well it's also true with auditing with AI, because it was a machine he could audit it he could say, well, why did the robot save me and not the girl? And the robot predicted that he had a 45% chance of survival, and the girl had an 11% [inaudible 00:06:01], but it's actually very meaningful in the context of understanding today's AI.

So those predictions, 45 versus 11, actually didn't tell the robot what to do. Someone programmed the robot to say, well 45 is more than 11, so save the adults and not the child. And then the protagonist, the detective goes on to say, well 11% was more than enough and a human being would've known that, the robot should have saved the girl. Well, that's a different statement, that's about judgment, that's about what do you do with a prediction, and it turns out these tools are amazing at prediction, but the judgment of what do you do with those predictions, that remains inherently human. And so all over the place, when we're thinking about using artificial intelligence and business, the AI's just giving you those numbers, it's giving you the predictions, but you have to decide, what do you do with them?

Tom Wheelwright:

So let me give you an example, and lets explore this, because I'm a business owner, most of our listeners are business owners, and we're looking at, boy, if we could predict revenue, if we could predict customer behavior, if we could predict workflow, it would have a huge impact because then we could make decisions, like you say, judgment, as to when to hire somebody, who to hire, at what time do we hire them, we can make judgments as to how do we expand? Because obviously if we can predict our current revenue we can say, okay, well if we did this, what would happen? And then if it could give us a predictable result and we say, well, okay, well if I spend $5 in this aspect of marketing, I do $5 on a Facebook ad, what will that get me? And it predicts that, well, it'll get you $20. Well okay, what if I put in $500,000, will it get me 2 million? So the same type of a thing, is that the type of prediction we're talking about here?

Avi Goldfarb:

So yes to some and no to others. So the investment predictions, that's less so because ultimately if you're using a prediction tool to try to figure out how to invest, and so is everybody else, it's all going to wash out the way the market works out. But for something like what's demand going to be, what's customer demand going to look like next period? Giving consumers coming to my website, or to my business, what are their interests? What do they want? That kind of filling in information, that kind of prediction, is very much in the AI's wheelhouse, today's AI's wheelhouse. So you can do things like predicting inventories, predicting demand, predicting whether a new applicant is going to be a good fit in the organization, things like that.

Tom Wheelwright:

So let me give you a… Oh, I like that last one, that's a tough one, right? So let me give you an example. So in my business, I'm a CPA, and what we're always trying to do is predict workflow throughout the year. We kind of know what customers have done in the past, but presumably with AI we could absolutely know and it could predict, okay, well we're going to get this much work in June, this much work in March, and this much work in October, and this is how we need to staff up.

Avi Goldfarb:

Absolutely. So to the extent that those things are predictable year in, year out, or at least they're based on factors that you can anticipate, then absolutely. So you have a sense maybe in April it's busy, but exactly how busy, and you could do those predictions at the customer level. So what's the likelihood that given my set of customers, each one is going to be, how many hours is each one going to require every month? And using your historic data, and historic data from other CPAs that maybe a vendor could provide, you could then predict the demands customer by customer and then you can aggregate it and figure out how to allocate your time and when you need to bring extra people on, et cetera.

Tom Wheelwright:

I like it. So one of my big questions though, for the smaller business, we understand that in our industry, for example, Ernst & Young, yeah they're probably using AI, they've probably been using AI for years now. But what about the smaller business? How does a smaller business actually find, I mean, even figure out how to use that information, where do they even go for that?

Avi Goldfarb:

So for most smaller businesses you're not going to be developing in-house, you're going to be going to a vendor.

Tom Wheelwright:

Exactly.

Avi Goldfarb:

But a lot of your software vendors, for lack of, will have some AI built in, you just have to look for it and think about the relevant ones. So if it's an e-commerce business, Amazon and Shopify will both have AI related tools that can support what you're doing already, and if you're, you can think about other dimensions, the tax software will have some AI in it already. So there's lots of opportunities to use it, but I don't want to exaggerate the payoff, so I just want to be a little bit cautious here.

So if you have your existing workflow, this is the theme of our new book Power and Prediction, which is if you have your existing workflow and all you're doing is taking an AI to improve your existing workflow, but not really changing anything else, so oh, you know what? This is a costly human process, it takes lots and lots of hours, let's use an AI tool, a machine learning tool to make that a little bit better, the upside is necessarily going to be limited. And so you might save 5, 10, 15% on that particular process, maybe even a little more, it's not going to have any transformative impact in the business. So you got to decide, for 5, 10, 15%, is it worth it to invest your time to learn the new tool, and all that.

Where the real potential lies is once you can figure out what are the major bottlenecks in my business in order to deliver valued my customers, are any of those bottlenecks driven by a lack of prediction, you don't have good information. And if that's the case, maybe you can build an entirely new kind of business, or entirely new business line, is maybe the best way to put it, where you take advantage of that better information in order to better serve your customers. My favorite example of this is the airport, which is those airports that are rated the best airports in the world, like Singapore and Seoul Incheon, and there are fewer in the United States and Canada, but there's some.

Tom Wheelwright:

That's true.

Avi Goldfarb:

But if you think about these great airports, what makes them so great? They have restaurants and they have shopping and they have museums and art galleries and theater, and all this. What do you actually want to do at the airport? Well, now let's look at what the super rich do. The private jet terminals don't look anything like those spectacular airports.

Tom Wheelwright:

That's right.

Avi Goldfarb:

I've been told they're effectively sheds, right?

Tom Wheelwright:

They are.

Avi Goldfarb:

Because no one wants to spend time at the airport, you only spend time at the airport because of a product fail on the part of the airlines.

Tom Wheelwright:

Hey, if you like financial education the way I do, you're going to love Buck Joffrey's podcast. Buck's a friend of mine, he's a client of mine, he's a former board certified surgeon, and he's turned into a real estate professional. So he has this podcast that is geared towards high paid professionals, that's who it's geared towards. So if you're a high paid professional and you're going, look, I'd like to do something different with my money than what I'm doing, I'd like to get financially educated, I'd like to take control of my money and my life and my taxes, I would love to recommend Buck Joffrey's podcast, which is called Wealth Formula Podcast with Buck Joffrey. I hope you join Buck on this adventure of a lifetime, exactly.

Avi Goldfarb:

Your goal is to actually just get out. And so imagine if you had a great prediction about how long it would take to get to the airport and through security, then you don't need shops, you don't need restaurants, you don't need any of that stuff, and the airport and the airlines would actually be delivering much better customer service. And so in your own business, just think through this, where are the constraints? I'm spending all this time, I have all this architecture, I had all these rules and standard operating procedures, not to serve my customers, but to actually compensate for the fact that I'm not serving them as well as I could.

Tom Wheelwright:

Interesting. And I do get to travel private from time to time, because I have buddies with airplanes, and seriously, you park close to the terminal, so the parking is an important part, and you walk in and get on the plane, and that's it. I mean, there's actually no security at a private airport, there's no security at all because you're getting on a private plane, so we're not worried about that, we know everybody who's going to be on that plane. But if you could literally go through security in five minutes and get on the plane five minutes later, that would be amazing. But we know you can't do that because you got to queue up in line, you got to queue up for getting through the security, then you got to queue up again to get into the seats, and then you got to find out, am I on standby? And all these kinds of things. So you're suggesting that boy, you could actually use AI to predict all that and actually solve for that if you used good judgment and figured out how to solve for it.

Avi Goldfarb:

Absolutely. And so the process is thinking through what are those things, like the shops and restaurants at airports, in your business that you're providing to your customers basically because you can't really deliver excellent customer service, because there's some constraint that makes it impossible.

Tom Wheelwright:

Yeah, it'd be like the TV's in the doctor's office. Why there TV's in the doctor's office? Because you're waiting for 45 minutes to see the doctor. That's a fail.

Avi Goldfarb:

Exactly, exactly, exactly. And almost every business has examples of those kinds of fails, some smaller scale like TV, some big scale, like the entire multi-billion dollar architecture of an airport. You see it in insurance, you see it in accounting too, you have these constraints and you make your customers wait, and do all sorts of things, just because it takes time to fill in those forms and do things, that fundamentally are predictable in prediction tools.

Tom Wheelwright:

Exactly, exactly. But you distinguish the judgment from the prediction. So I always tell people that the job of a CPA, for example, is analysis, which is to help you make a judgment, right?

Avi Goldfarb:

Right.

Tom Wheelwright:

So how does that differ? I mean, you mentioned the I, Robot movie, but in practical terms, and in typical business terms, how do you combine that judgment with that prediction?

Avi Goldfarb:

There's all sorts of places, so it depends on a particular context. My favorite example, I'm going to use a sports example, what is the business, it's the business of sports. Michael Jordan in his first season, this was a long time ago, and he was injured, and he wanted to play, but the doctor said, “There's a 10% chance that you'll never play again if you play.” And so the owner of the team went to him and said, “Michael, why do you want to play so much? Imagine that you had a headache and you could take a pill that might cure you, but there's a one in 10 chance it would kill you. Would you take the pill?” And his response was, “It depends how bad the headache is.” That's the essence of judgment.

So you think about these risks, it's a risk reward trade off often in many businesses, which is you have the prediction, and then you have to decide what matters, what do you value? So are you willing to take, in that context, Michael Jordan wanted to take a 90% bet that he'd be able to play in the playoffs, willing to take the 10% risk, the one in 10 risk that his career would be over, the owner of the team was not willing to make that bet, and pretty much didn't, and basketball history ended up being made as a consequence. But maybe basketball history would've started a year earlier, with the Bulls winning the playoffs in his very first season. Who knows. And there's every single decision you're going to make.

So a restaurant, you have to decide whether to set up your patio on a given day or not. There's a prediction on rain. Well, how bad is it if you set up the patio and it rains, versus if you don't set up the patio and you forego all that money. There's predictions on, in any business, on how much demand you're going to have, how many customers are going to come in to your store, or how many clients are going to come in to your business this week or this month. And then you have to make hiring decisions. So the prediction just tells you how many people are going to be there, the judgment is what's the consequence of hiring too many people, and having to pay a little extra, versus you're not hiring enough and having to turn customers away.

In every business, frankly even by industry, every business owner is going to think of those costs differently. That's the essence of the judgment. So the prediction can just tell you, oh, you know what? On Friday you're going to have 20 extra people in the store. And the judgment is, well what do we do about that? Do we hire more? Do we turn people away? How do we think about those risks?

Tom Wheelwright:

That's really interesting. So one of the things that every business owner thinks about is what's the value of my business? Not just what's the income currently to the business, but what's the value of the business? And when you look at AI and predictability, do you see AI actually being, in the future, being able to predict, here's the value of this business based on what's going on now, and if you change certain parameters, here's the value of the business if you made these changes?

Avi Goldfarb:

That's going mad, I'll put my economist hat on and say it depends, which is for the path that you're on, if you don't anticipate very major changes, then you can say yes. If the past is like the future you can use AI to fill in missing information and provide some value to the business. Once you're starting to think about, well if I do something different, what's going to happen? Well now we're actually no longer in the world of prediction, it's a different AI tool called causal analysis, that hasn't had the major advances that we've seen in prediction tools.

And the reason it's so hard is… So look, I'm a professor, I come in and I say, “Look, I'm going to teach you all about AI,” and my students come in, they're like, “Wow, we learned about AI.” And I say, “Well imagine that because you've learned about AI from me and you've read my books, imagine your future where you're spectacularly successful professionally and personally in every way you can imagine, and you can connect it to your knowledge of AI. Can you then say it was the right call to read my books and show up to my class?” And the students are like, “Of course.”

And I'm like, “Well actually, unfortunately, as much as I'd love that to be true, the answer is no, because you don't know what would've happened to you had you not read my books and not taken the course. It's possible you took them because you were interested in AI in the first place, and all of that success would've happened anyway.” And because of that we don't know what happens in the counterfactual, so we don't know what would happen if you did something different, so it means we don't have data on it, which means the AI won't work there. So trying to think about what would happen if is challenging, unless you think you have a business that you can simulate, and you can design a simulated framework for understanding those different opportunities.

Tom Wheelwright:

Interesting. So the AI, when it comes to predictability, is really taking the information that we have and it's just assembling that information, well, based on the information we have, this is what's going to happen, but we always know that past performance is not a predictor of future results, right?

Avi Goldfarb:

So it's not just… The prediction should be fine in terms of past performance predicting future results as long as you don't change your strategy. But once you change your strategy, we don't have data on that, we don't know what will happen if you change your strategy. And so that's where it's going to break down. And of course it can tell you in many contexts if we stay along the same path what will happen, but what you want to know is, uh-oh, okay, I see if I stay along the same path what will happen, I want to know what happens if we do something different, and that, at least for now, in most contexts, is not an AI problem, so there we need to remain [inaudible 00:24:00].

Tom Wheelwright:

So not taking over humans anytime soon.

Avi Goldfarb:

Not taking over humans anytime soon at all, no, not at all.

Tom Wheelwright:

Well, that's good news for most people. Now you've said that you think that AI will actually make business better, be more successful, that you think that this will have a powerful impact on business, and really could completely disrupt certain industries. Can you just take a few minutes, go into that a little bit for us?

Avi Goldfarb:

Sure. So it's a version of the airport example I just gave you, but those happen in all sorts of business. So many businesses have SOPs, standard operating procedures, and for the most part those standard operating procedures are rules that are there to accommodate the fact that sometimes you can't make the perfect decision every time, and so instead what you go for is reliability, and to become dependable both to your customers and your suppliers, but also to the other people in your company, so person A and person B, the two people, their decisions are coordinated. So bring in AI, and the first way we've seen that so far is, okay, let's add it to our workflow, and it's had some impact, but it hasn't, as I said, it hasn't had that extraordinary impact. To have the extraordinary impact you want to break your SOPs, you want to change your standard operating procedures to think through, how can I deliver value in a way that I hadn't before?

When I talked about the airport you had think about insurance, you can do the same thing. So we'll start with, imagine going to your doctor, and your doctor looks over your symptoms, looks at your blood work and says, “There's a 5% chance that you're going to have something catastrophic happen to you over the next year.” If that happens we'll give you $100,000, see you next year. That's not what, the doctor's supposed to give you some treatments, right?

Tom Wheelwright:

Right.

Avi Goldfarb:

Think about what your insurance company does, they don't give you treatments, they say, oh you know what, there's a 5% chance, they price the risk. So basically they're saying there's a 5% chance that something catastrophic is going to happen to your house next year, and if that happens we'll give you some money. And they claim that they're giving you peace of mind against catastrophic loss. That's not really peace of mind, peace of mind would say, you know what? There's a good chance that something disastrous is going to happen to your house, we think it's going to be because of an electrical fire, for example, and we can then help you reduce the risk from electrical fire. So not just compensate you if something goes bad, but actually use better prediction in order to reduce the risk and deliver a much better product to customers.

10 years ago that was impossible, the insurance industry's predictions just weren't good enough, so they couldn't do that. But now they're getting there, the predictions are good enough that it's not just at the aggregate level, at the sub parallel level they can say no electrical fire, or leaky pipes, or whatever else, and they can help you reduce that risk. And there's all sorts of industries like that where the fundamental nature of the way you serve customers is related to your standard operating, and the way you build your standard operating procedures is because you don't really know how to deliver value, and with better prediction, you can deliver value better, you can-

Tom Wheelwright:

Interesting.

Avi Goldfarb:

And I think there's a real opportunity, there's a real risk for disruption because if you don't do it your competitors might, but there's also a real opportunity.

Tom Wheelwright:

So from a practical standpoint, I remember, well, back in the 1990s when the internet was new, and you spent $15,000 to develop a website, and that was a basic website, and now you have plug and play and it takes 15 seconds to do that. Do you see that coming with AI, do you see where the AI tools will be such where an average business owner can say, okay, CPA, for example, here's this AI tool that we can use, and let's sit down and do an analysis because we've got this plug and play predictive tool.

Avi Goldfarb:

Yes, I think there's two ways that small businesses can think about the opportunity. So way number one is we're moving toward plug and play. We are for various applications, and that plug and play, that you'll purchase through vendors, will make your businesses more productive in various ways. It's not going to make a huge difference to the bottom line, but it's going to be part of the standard improvements that you make every year. There's various ways that you invest and improve the way your business operates, and one of them will be to use AI, use machine learning tools for a particular context, as purchased from a vendor. And I would [inaudible 00:28:53] percent or more of small businesses are going to be in that category.

But a handful of them are going to see much bigger opportunity, and say, you know what, if I can get these predictions right, then I can actually create a new way, a new type of value for my customers, or even capture a whole customer segment that I was never able to capture before. And those are the ones that are really going to make the headlines and transform the way we operate, but also be the ones that think about AI as central to their business. So there's going to be most, and I imagine most of your listeners are going to be in the category of this is useful, just figuring out a new way to save on electricity costs, energy costs is useful. And then a handful are going to say, well now that I can do everything differently, I can develop a new product, and that new product is going to be central to the business going forward.

Tom Wheelwright:

Got it. So at the beginning of this discussion I suggested that it was AI combined with blockchain. Have you looked at that as to that combination of AI and blockchain and how they work together?

Avi Goldfarb:

A little bit. So I think there's, to the extent that we can think about blockchain as verification technology, so that's the essence of it. Now it's online verification technology, we know very well that you also need to check things offline, because the internet, cyberspace isn't a real place, you still need to verify that everything's really happening. But there are opportunities to take the online verification tool that's in blockchain and combine it with prediction technologies to develop some better services, particularly in financial services, there's opportunities there. But in both of those contexts the opportunity relies on you remembering that we still need to do verification in the real world, and we have to recognize that just because we can track the transactions across the blockchain, that doesn't mean that those transactions themselves are legitimate in terms of what's happening offline.

Tom Wheelwright:

Awesome, thank you. So to sum up, what would you say are maybe two or three things that a business owner or investor ought to be looking at over the next year or two when it comes to AI?

Avi Goldfarb:

Okay, the first thing is when you hear AI, don't think the robots from science fiction, think prediction technology. It's software that gives you predictions, and predictions are useful because they help you make better decisions. So that's the central point. In terms of, then number two is for most businesses, particularly most small businesses, the gains will be useful but incremental. Now if you're thinking about, especially if you're thinking about investment opportunities, what you want to be looking at are, if you're focused on AI, are opportunities where the prediction allows you to overcome a bottleneck and generate a new way of creating value to some set of customers. And so where the predictions allow you to change your workflow, do things differently, and do things much, much better. There's going to be lots of opportunities like that, but they're harder to find. It's easier to look for the easy wins, but there are big wins out there for rejigging the organization and thinking about disruption. And that last point is the theme of our new book, Power and Prediction: The Disruptive Economics of Artificial Intelligence.

Tom Wheelwright:

Awesome. So thank you so much, Avi Goldfarb from the University of Toronto, really appreciate you being here. If they want more information besides reading your book Power and Prediction, where else could they go for more information?

Avi Goldfarb:

My website is avigoldfarb.com, I've got all my work on AI up there.

Tom Wheelwright:

Awesome, thank you so much for being with us. And remember, when we look at these disruptive technologies, it's either an opportunity, or it's a challenge, because either you're going to embrace it or you're going to get left behind, and if you embrace it, and this is why we do this in The WealthAbility Show, you're going to end up making way more money, and as we get to using AI and blockchain in the tax world, pay way less tax. Thanks everyone, see you next time.

Announcer:

You've been listening to the WealthAbility Show with Tom Wheelwright. Way more money. Way less taxes. To learn more, go to wealthability.com.