Episode 797 11.15.22
Joshua Gans, professor of strategic management, University of Toronto, Rotman School of Management, joins Peggy to talk about his new book, “Power and Prediction.” He also describes three stages of innovative technology, why humans will remain critical of AI (artificial intelligence), and how old systems will be displaced with new and more efficient ones as artificial intelligence adoption continues to grow.
Peggy Smedley: Joshua, I’m excited. You’re coming out with a new book. You and some professors have worked diligently to put a new book out. Are you excited about it?
Joshua Gans: Yeah. We think it’s a book right for the times at the moment. There’s been so much talk about artificial intelligence, which the book is about, yet I think businesspeople are a little frustrated that it’s proving challenging to actually use it.
Smedley: Well, I couldn’t agree with you more. I think in your book, “Power and Prediction,” you look at the adoption, one of the things I was looking at about electricity, in fact I think if… correct me if I’m wrong, you talk about the period of change from slow to rapid adoption and I found this very interesting in the analogy you make about the period from which the adoption, we had rapid change to what… actually slow, I should say, to rapid change. And that’s kind of what we’re seeing now with electrification. And you are saying that AI’s light is on. I love that analogy. Are you saying that that’s really what’s happening with AI? Can you talk about that, what’s actually happening and maybe what’s going to be happening in the coming years?
Gans: Yeah, everybody gets surprised when you sort of look back and you say you would’ve thought when Edison created the electric light bulb and lit up some streets, which was it. The world had changed, but in fact it took 40 years for electricity to start having rapid adoption, especially by businesses. And the reason that historians point back to, is that really electricity allows you to redesign the factory floor. It’s been a much bigger mass production operation. Ford would not have been able to create the production line for the Model T if the economy… if only steam power were available. And so that took a while for people to really understand what electricity was doing. This has happened time and time again with transformational technologies and so when we looked at AI’s adoption, which is actually pretty weak compared to its technical achievements, we thought about whether the same sorts of things are in operation and whether we can provide some clues as to how to examine that for your business.
Smedley: You can tell you guys are professors because in my book I did the same thing. I looked at a little bit of history and I think we erase history sometimes because we’re so eager to look ahead. But in your book, you make a lot of analogies looking back and one of them you just mentioned here about Edison, but you even took a deeper dive into some of the folks most of us would not know, and you actually talked about Frank Sprague and I didn’t even know who that was, the former employees of them. Talk a little bit about that because I found that so interesting talking about them and the electric motors and how they played a role and what we’re looking at today. I think that’s really interesting.
Gans: Yeah, so the way we divided it is three stages of real innovators in any new technology. First you have your point solution entrepreneurs, these are the people who with electricity realized that if you had a factory that was sometimes running on steam that didn’t have great stable water resources and electricity could be more stable power supply. And so all that took was swapping out power sources. And then after a few years people started to realize that electric engine had a scalability property, not up but down. You could build an electric engine and tack it onto some sort of machine tool or welding process or something like that and have it just worked with that process. That was a big change because previously your entire factory had to be on or off, but here you could have just machines being on and off, which made it much more power efficient. And those are sort of application entrepreneurs.
And then finally you have your systems entrepreneurs, which I would put Henry Ford and Co., in that basket, who said now that we have these different ways of configuring the sources of power on our factory floor and we don’t need power to be next door, it could be hundreds of miles away or then we can reimagine the factory. And so, we see ourselves with respect to artificial intelligence, people have already started swapping out AI for human prediction in making various decisions in their business, those point solutions. And we see a few little wrinkles of these applications coming through things that really rely on artificial intelligence to exist. But the real transformation’s going to come when someone understands how to take AI prediction and reconfigure entire systems around that.
Smedley: And that’s an interesting point because the human has to be a part of all of this. I mean, sometimes we forget, and we look at things differently and we’re saying the humans can be out of it, but the reality is when we look at point solutions, the human’s going to be a critical part of this and I think that your point in all of this is very critical to that. Is that correct?
Gans: Yes, absolutely. In fact, humans are irreplaceable because what humans do in the context of AI is they give the AI, for one of a better term, its objectives. No artificial intelligence that we process that we currently have, can do that. The best that they can do is they can watch people and mimic them and get their objectives that way. But the real way you get objectives into AI and make a compliment is that you program it straight in. That’s what happens with our self-driving cars, for instance. We program in to make sure the car is… when it’s got an uncertain situation, it prioritizes, for instance, safety and things like that.
And that’s proved a huge challenge by the way, just being able to articulate all of that. But that’s a fundamental feature of all of the artificial intelligence as we have it in this current wave of innovation going on. Humans are at the core of it, and you can’t replace that role. Now you might be able to automate some functions and other things like that. But it’s not like we’re… it’s very hard to breed a human all the way out of the system. And in fact, it’s a fool’s errand, is basically what we say.
Smedley: Should we understand the difference when we’re looking at AI and looking at application solutions like you’re just talking about? And very carefully understand why a new device that can be differentiated when we’re thinking about this completely? I mean because I think there’s a lot at stake here when we think about AI and we think about how we have to apply it and we have to think about all of these point solutions that you’re talking about because I think it makes it very valuable to the core of what we’re producing, of what companies might have as the critical end result. Is that accurate?
Gans: That’s right. You can come and someone can sell you on some AI prediction that looks like it’s wonderful, can tell you exactly what’s going on, or it can provide very high-fidelity predictions about the near future, and you think this is going to be wonderful, we can really use that in our business. But you already create stumbling blocks and they come pretty quickly because your business was built and created and optimized for a world without that AI. And there’s a lot of moving parts to that. There are all the sets of decisions that sort of radiate out from the one that you are using AI to inform.
And moreover, there’s a whole lot of people along those same chains and in those parts of organizations, for which this can be quite disruptive. So, any transformational technology there are going to be winners and losers and what existing businesses face is that the winners and losers are people in the same company and that is going to cause friction and other things like that. And our basic message is, this is all a reason for the organizations to take a step back and think about the uncertainties they face. Before the AI appears, they can actually resolve it, so that they know what they want when it arrives because these things are arriving all the time at the moment.
Smedley: So, do they have to then take that step back and say what is going to be the application solution? What is the system solution? Is there a step by step? Is that what we’re saying, they have to look at or is it one over the other?
Gans: Well, I think it’s all of the above. So, if we take the art of history, right now, the only place you can really adopt AI in organizations easily is in point solutions and maybe sort of applications depending on what you’re doing. So, you are stuck with that. But that’s okay. This is the period of everybody sorting itself out and getting some of the low hanging fruit from these new technologies. But the point is there’s going to become a point sometime later or there may well become a point where in your industry something transformative comes along and those are going to appear first in startups because they don’t create the friction, have the political frictions, and other things going on and they aren’t as set in their ways. But what a CEO can do now is realize that those decision points for adoption are coming and make sure their organization is prepared to understand the changes that would need to be done should they arise.
So, they’ve got to sort of identify: what’s my key uncertainty? Well maybe I don’t know when my customers are going to buy my products, is there going to be a surge or not? Or I don’t know if supply chains are going to be stable or they’re always going to be fluctuating or any of these sorts of things going on. And they pick down those uncertainties and say, well what if we could resolve those uncertainties? How would our business look? And that gives you clues as to what might have to change should an AI that resolves those uncertainties come your way.
Smedley: I guess the question that comes to mind as you said that, is it the surge of uncertainty you have to be fearful of? Or is it the threat of disruption by the unknown? Is it the bad guy you have to be fearful of? Is it not knowing your customer? Is it not knowing something like COVID that occurs? Or is it all of the above? It’s something that you just don’t know that might happen?
Gans: No. Well actually, interestingly enough, those things like COVID or a catastrophe associated with climate or something like that, AI’s not going to help you very much with that. So those aren’t the things that you’re worried about. You’re worried about the latter thing, which is the sort of disruption that people talk about, which is basically when a new technology comes along, conceptually it could change everything, but existing businesses that are already organized around the previous technology find that change hard. And what we’re saying with respect to artificial intelligence when it comes in, it’s going to come in from key ongoing uncertainties that your business faces and currently sort of pushes into the background by building up walls or setting up rules. But you need to be aware of those things because the worst thing that could happen, the worst-case scenario is this technology comes, customers start clocking somewhere else and you don’t understand why.
Smedley: So basically, you’re describing as old systems are replacing or are displacing new systems and you’re not paying attention to those market shifts, is what I’m hearing you describe at this moment. And-
Gans: That’s very difficult.
Smedley: There is a shift in power. I mean, in some ways that’s what’s happening. Is that what you’re describing?
Gans: Absolutely. It’s a huge shift in power. I mean the closest analogy I have for this that might resonate is, when ride sharing came about, Uber and Lyft and so forth, that was powered by artificial intelligence and it’s sometimes not foreign. We normally think about it breaking down the regulatory barriers, meaning you don’t have to have a license to drive, but I’m telling you, if you got into an Uber, and they didn’t know where to go, and they didn’t know how to handle traffic, and all of those sorts of things, you wouldn’t use an Uber very much. It would’ve failed right out of the gate. But instead, at the same time on your mobile phones, was the power and insight of a three-decade old, experienced taxi driver that could be given to any old driver. And with that, this industry got disrupted.
Smedley: And so that’s actually the trade off because they had the insights to be able to take us to a new level of what consumers wanted. So that’s why one industry was disrupted by another, and technology took them there. So that’s where we have to see that disruption that you are describing, old systems are being displaced by new ones. That’s that disruption. And that’s why we talk about manufacturing and other industries being disrupted because they don’t see what’s coming. They kind of get rooted in not recognizing what’s displacing them and they look at technology as being bad instead of seeing technology being good.
Gans: Exactly. And of course, artificial intelligence is all about uncertainty. When you look back at how your business was designed, it was designed for things that you can’t predict. It needed to insulate against that. You needed to take out some insurance, some protection, or something like that. But the problem is that those are expensive things. If on the other hand you can sort of suddenly predict those things, you’d do things completely differently and you can forget that your organization was designed to ignore because it didn’t have to worry about the uncertainty when that uncertainty was still there, and you were paying a sort of hidden price for it.
Smedley: Joshua Gans, this was always a wonderful conversation. Where can our listeners go to get your new book “Power and Prediction?” This is fabulous. Thank you for joining me today, the professor of Strategic Management, University of Toronto, Rotman School of Management. Thank you for each time. Where can our listeners go to learn more about what you’re doing and to get your new book?
Gans: They can get “Power and Prediction” at any online bookstore, or predictionmachines.ai to buy the book.