Podcast transcript: When will society be ready to harness AI's potential?

27 min approx | 10 Sep 2019

Every single thing we do will be changed by AI in the next five to 10 years. It will touch everything.
Nigel Duffy
EY Global Artificial Intelligence Leader

Yassmin Abdel-Magied

Welcome to The Better Question, EY’s podcast series that will answer the questions that will help you lead your business through this transformative age.

What will happen when trends that we already know are underway today start to transform society?

I’m Yassmin Abdel-Magied, engineer by trade and writer by vocation.

Over the course of six episodes, we are asking some of the most probing minds around the globe about the world-changing issues that are just over the forecast horizon. And we are asking them, what’s after what’s next?

And I have here with me writer, futurist and the curator of the What’s after what’s next? Chris Meyer.

Chris Meyer

Great to be with you, Yassmin. I am Chris Meyer, and my ongoing work with EYQ, the EY global think tank, is focused on the question, what’s after what’s next?

The EYQ think tank, EYQ, has assembled the Global Fellows from all over the world. And part of my work with EYQ is to chair this board of 10 fellows.

Abdel-Magied

And in some of the many books you have written, you have tackled a lot of the topics we are going to cover in this series.
 
Meyer

Right. But we have narrowed it down to the six topics most important to leaders interested in the future today. And for every episode, I will sit down with two guests to dig deeper into the topic at hand. So, in each episode, we will have a conversation between one of the EYQ fellows and one expert with a different perspective on the issue.

And let us talk about you, Yassmin. You are here to be the representative of the listener. You are a writer, you have worked across three continents, you are trained as an engineer, and you are going to help us make sure that we do not miss the points most important to senior leaders.

Abdel-Magied

Chris, you are being very kind. I am just here to ask questions. And here is my first one: If you have only got about 25 minutes to think about the future, which is not much time at all, why spend that time thinking about AI?

Meyer

Great question. Because there is no technology being rolled out today with the same potential to increase what each person can do.

Abdel-Magied

I mean, I have heard AI compared to electricity, eventually both indispensable and invisible.

I think that is a great analogy. At the top of the show, we heard from Nigel Duffy, EY Global Artificial Intelligence Leader. Nigel said every industry will be touched by AI. So, we thought it was important to kick off with an episode on artificial intelligence to help leaders understand the potential and the pervasiveness, and ask the better question: When will society be ready to harness AI’s potential?

Abdel-Magied

Yes. That is why I am really interested in this episode and in the topic generally, because it is about how we think about technology and how AI increases human capability in virtually every dimension of life.

Meyer

And we will see if we cannot connect the dots and distill the main takeaways for the executive.

Abdel-Magied

Well, I cannot wait to hear about it, Chris.

Meyer

We are going to be talking today to Stefan Heck. He is leading an AI-based startup in Silicon Valley, and he is a member of the EYQ Global Advisory Board, the EYQ Fellows.

Heck

I am Stefan Heck. I am the CEO of Nauto. I am an entrepreneur running a startup in Silicon Valley working globally. We are an AI technology company that is focused on driver and driving safety.

Meyer

And our other guest is Eduardo Porter, who thinks not so much about the business of AI, but about the impact it is going to have on our economy and our society.

Porter

My name is Eduardo Porter. I write about economics for The New York Times.

Meyer

Stefan, Eduardo, welcome to The Better Question. We are talking about issues that are most important to senior executives, and starting with one of the most pervasive topics, artificial intelligence. So, let us get a sense of how each of your work has focused on AI, what you anticipate are risk factors, and at the end of our conversation, we will end with a rapid response section.

Let us start with you, Stefan. As the CEO of Nauto, you are one of a growing class of people known as AI entrepreneurs. So, can you talk to us a little about what you are doing now, and what does your company do?

Heck

Sure. I started a company called Nauto. We are an AI technology company that is focused on driver and driving safety. We are based in Silicon Valley but deployed globally across the US, Japan and Europe. And what we do is we retrofit artificial intelligence and computer vision to commercial vehicles, any type. So, it can be a passenger vehicle that is in ridesharing, a taxi — it can be a bus.

Meyer

So, I do not think that so far, you have used the term “artificial intelligence,” or “AI.” Where does that come in?

Stefan: What is different about our use of AI from many of the ones that get a lot of the media attention is our AI is augmenting a human driver. We are not substituting for human intelligence. We are adding the 360-view — a non-sleepy permanent attention, if you like. So, we build on the driver’s experience, but we add the experience of the artificial intelligence.

Meyer

And Eduardo, let me ask you, when you use the term “AI” in your writing, what are you referring to?

Porter

Me, I am thinking machine learning. You know, machines that can learn and improve, and kind of do something that looks like making a decision, as in driving a car, for instance, without the involvement of a human, or making a diagnostic decision based on patterns it has read in many, many X-rays or MRIs. That is my understanding of what artificial intelligence is today. Basically, pattern recognition and decision-making based on pattern recognition.

Meyer

So, let me ask you, Stefan, on sort of a spectrum, there are plenty of point solutions of AI all over the place. But when you scale up to the system applications such as Nauto, you have got to partner with the rest of the system and see how the system gets better, not just how your product gets deployed. Is that a fair summary?

Heck

That is a great summary. I will give you an example in the domain we are working. If you look at transportation today, it is actually an example of a really inefficient system from a capital utilization point of view. AI allows us to transform that as a productivity. You know, if I take a car and I drive it autonomously, I can now drive it more or less 24/7. I have to refuel, and charge, and do some maintenance, but I can definitely get much higher utilization than 4%. And so, now you are talking about a roughly eightfold increase in the throughput of every single lane of every single highway we have ever built. That is a massive productivity improvement, if you combine that.

Porter

I was hoping to just interject here with an example about how the penetration of these wonderful new technologies is kind of constrained and dependent on some of the external institutions. And whether one technology gets deployed in the most efficient way or the most dense way or not is really also embedded in our social and political systems. And so, that is going to be, I think, very important in the deployment of AI.

I am not sure what we need to do to our roads and our stop signs and whatnot for a future in which autonomous vehicles are 80% of the vehicles out there, but I am sure something is going to have to happen. And that will be political decisions.

Meyer

This is sort of a change of subject, but you have written the case for taxing robots is stronger when you consider that a lot of automation these days is not deployed to enhance economic productivity. What do you mean by that?

Porter

Well, I find that is a very interesting thought, and it is a thought that has come to me from a few economists. And the finding is that a lot of the automation that is taking place might not necessarily enhance productivity because it is really happening because of basically a tax break, because there is a tax preference to automation over people. And so, you will replace a person with a robot, even though, at a society-wide level — I mean, it will be good for your company because you will reap the tax break. But at an economy-wide level, it will actually not improve productivity much. So it will not really improve welfare much. So why do it?

So, the argument isn't really, let's go and tax robots willy-nilly. It is why do we not remove the tax preference? That is perhaps the best way of saying it. I know the headline was, let's tax robots. That is a sexier headline, but it is why give a fiscal incentive to something that is not improving our overall economic efficiency, overall welfare?

Meyer

Yes.

Porter

I mean, when I am worrying about development, I am worrying about, do we still need cheap labor anywhere in the world, or have we moved to a world in which, because of automation and so forth, cheap labor has become irrelevant? And so, that would be a problem for developing countries.

Meyer

But it sounds like you are still in a model that they have to go through industrialization, is what I hear in that.

Porter

Yes.

Meyer

I am wondering if an economy emerging in the beginning of the 21st century might look really different because they are different to begin with.

Porter

I mean, it might. We have no examples of a poor country becoming rich without industry. But again, the future is in the future.


Meyer

Yes.

Porter

And I have a question for you, Stefan. Could you tell me how to make AI work for everybody, as it were?

Heck

Yes. It is governance, tax system incentives — there is all kinds of things at play there that we have generally gone in a bad direction and made things less fair and less equal, even as we have become wealthier and more productive.

Porter

I wrote a piece a little while ago about a proposal by two guys, E. Glen Weyl and Eric Posner, who had it in a new book, which said, well, how do we make AI improve everybody's income? Well, why do we not have the makers of AI pay us for our data? And that would be a way to kind of spread income all around, since these systems need to train on vast amounts of data. Presumably this data will be valuable to the owners of these systems. And we could, in such a way, kind of distribute whatever the social benefit that we are getting from these new systems throughout the economy and throughout the population. Well, what do you think of that? I mean, does that make sense?

Heck

I do agree, and I like the direction you are going of, we need to find new tax mechanisms or other redistribution mechanisms, right? And whether it is paying you for the data or taxing the profits of the AI in a different way — I mean, the whole discussion about how do you tax Internet companies to me is an example of that as well, right? Here is an enormous amount of wealth being created. The legal headquarter location really has very little to do with where the productivity gains are and then how we choose to invest that, right? And people used to talk about the marriage tax penalty. This is now the worker tax penalty, right, where the robots have an advantage because you do not have to pay for Medicare and Social Security for robots. How do we rebalance that?

There are a lot of these kinds of changes coming. And I think we are headed for a whole wave of these. We need to revisit our tax models. We need to revisit how we distribute the profit portion that we collect as government entities, through our government entities, to redistribute to people, and that that needs more balancing.

Meyer

You mentioned your degree in, I guess, formerly cognitive science and philosophy, and before that, in symbolic systems, right? So, your whole life has been preparing you to answer this question. [Laughter] And the question is, do you see AI as a change in the human condition, like the invention of writing, that allowed us to carry knowledge forward from generation to generation? Or is it more like the next wave of IT? Where does it fall on that spectrum?

Heck

I think in the long run, we will find it is more like writing. Today, I think, we are seeing it barely show up. So it is like another just software upgrade in IT. But I think if you look 20, 30 years out over that timeframe, we are talking about like the invention of writing.

Meyer

Eduardo, I know you have some doubts about seeing productivity gains from AI so far. But that is what is being promised. And I am wondering, if those come about, do you see it more like a power loom invention that displaced a lot of weaver jobs a century ago or more, or is it more like, say, the elevator, that allowed for the creation of a lot of jobs in a more productive world? Where would you put it on that spectrum?

Porter

Well, let me take a little issue with the spectrum. 

Meyer

Okay.

Porter

I frankly think that the power loom ultimately, on an economy-wide basis, actually probably helped to generate more jobs than it destroyed. Sure, it displaced the people who were weaving the stuff before these looms came into being, but it also increased the income of the makers of cloth, and that created demand for other things that had to be made, and that eventually employed a lot of the otherwise displaced people. And that increased our income and our welfare.

And I think that that has been — mostly the history of technological change since the Industrial Revolution has had that characteristic. There has been narrow labor displacement. But overall, there has been labor growth, income growth and productivity growth attached to it. Now, where does AI fit in that? I mean, my prior would be to think that because everything else has pretty much worked that way, why shouldn't AI?

Heck

I wanted to come back to one other thing you said, Chris, which I actually really liked, which was your spectrum, the elevator example. To me, that is a good thought experiment for how to think about the impact of AI. The elevator is the most efficient form of transportation on the planet, 99.8% efficient. And so, you have got exactly the asset productivity. By any metric, the elevator enabled this amazing revolution that showed up in the macroeconomic statistics in all these other ways, not because the elevator industry itself became gigantic. And I think, done right, AI will have that kind of effect. It will enable us to combine services, to target services, to make our infrastructure more efficient.

Meyer

What do you see acting there, and what are you predicting?

Porter

AI, I think, really hasn't diffused deeply into the economy, into the labor force, in a way that shows up in our macro statistics, that shows up in productivity numbers. And so, at the end of the day, I do not know that AI will act in the same way that, say, previous rounds of automation has acted. And I am indeed a little skeptical that it will diffuse very fast. And I think that is because there are institutional obstacles to AI that I think are going to slow it down some. Sometimes I feel that the people developing the AI, the tech people, think that because it can be done, it will be done. But there are enormous societal constraints, institutional constraints, that I believe are likely to slow this down.

Can we use AI to remove unwanted biases from our systems, or will we end up reproducing unwanted human biases? And as far as I can tell, it has been very, very difficult to remove human biases from these systems because these systems are trained on human data, and the human data contains all these biases. So, if you are going to entrust your hiring to some AI, I would be very careful, because it is trained on systems where these sorts of discriminatory patterns exist.

Meyer

You are an executive in a corporation, and you are looking at this technology emerging.

Heck

Right.

Meyer

How do you think about that?

Heck

Well, as an executive in a company, the way to think about AI is that it is a deep portion of your product if your product does anything that requires decision-making, data processing, perception, classification or actuating things that change.

The places where I would stay away from AI is where human judgment of the social impacts of the human consequences are still required. So if you are designing your benefits program, by all means, use AI to analyze the options, but the decision should be made by your head of HR. If you are scanning for leaks or environmental threats, again, AI can tremendously help in the perception part of that, right? Detecting things. But then the “How do we respond to the crisis? What do we do?” — you want a human in the loop.

So, there are a lot of uses for AI to augment and support humans, and support executives in companies, but where the responsibility, the ethics, the decision should remain with the humans, where AI really can become the core decision-maker, and you automate decisions that require huge amounts of data, more than humans can handle, and pattern recognition. But they should not be the kind that have broad ethical consequences. And then if you are making a product that does those things, build AI into your product directly, otherwise within 10 years or one corporate generation, you will be out of business.

Meyer

We will close with a lightning round that we are calling ping-pong, meaning I am going to ask you a question, and I want you to give me a less than 30-second answer. Name an area of society — could be in business, could be medicine, could be home — name an area that will be significantly different 10 years from now because of AI.

Porter

Oh, health care. Immediately, health care. The whole — the diagnostic possibilities from AI, I think, are just kind of fairly easy to deploy, and I think it is going to take off there.

Heck

Government services. Same idea — being able to respond very rapidly, more tailored to the needs, without having to go through paperwork, bureaucratic forms.

Meyer

Eduardo, last one.

Porter

Energy. I think that the demand for smart intelligent energy systems is going to be paramount. They are going to play a really important role in our battle against climate change. And I expect that as electrification, as the use of renewable energy in our electric grids increases there, the demand for intelligent systems of transmission and distribution is going to increase. So, I expect a lot of investment there, and I do expect a lot of gains there.

Heck

If I may add to that one, Chris, because I think it is a great example.

Historically, we have built our whole energy system to vary supply based on load and demand that we cannot control. I think the biggest application for AI in energy is we can actually make the loads smart. Because for a lot of systems, we do not care when they charge, when they run. Do you know today when your fridge runs? You have no idea, right?

Meyer

Yes.

Heck

And so, your fridge could use energy when it happens to be available, and not use energy when it is short.

Meyer

So, Eduardo, if you were going to put one thought in Stefan's mind to consider, what would that be?

Porter

What should govern our decision to move from your current technological approach about augmenting human drivers to replacing them? Is it just a technological challenge, or should we be thinking about other things besides what we can do? And, I do not know, what are the ethical considerations?

Meyer

Stefan, what is the one thought you would offer Eduardo as a takeaway?

Heck

Eduardo, you said that we should tax data because that is the fuel of the AI economy. I agree with that. But I think there is an interesting discussion we should have, an exploration we should do, about where in the value chain do we tax that data? What is the right intervention point in the new economy to tax the profits that arise out of data?

Meyer

Thanks so much for joining us, and thanks for, really, the great material and participation. I appreciate it.

Heck

I would definitely love to follow up, Eduardo, and continue some of this conversation.

Porter

Would love to, Stefan.

Abdel-Magied

Well, that was great. It was really nice to hear a conversation about AI that was not all about how clever and amazing the technology is, but one that actually focuses on the questions and the risks that it poses instead.

Meyer

That is a fair point, Yassmin.

Abdel-Magied

And every episode, we will be doing this. We are connecting the dots together and talking about what may be best for the executive to think about. So, let me start by asking, what were your main takeaways, Chris?

Meyer

First, Stefan said in no uncertain terms that AI will be pervasive. Whatever you are making, whatever your process is, AI is going to be part of it. The question executives ought to be asking themselves is, have you begun to identify how your business will be transformed by a set of technologies?

Abdel-Magied

But is there not another risk? We are finding that AI, which is used to guide hiring and lending, maybe even college admissions someday, tends to replicate the biases of the past because, well, that is the data that trained it.

Meyer

I think the good news is that both developers and users of AI are already paying a lot of attention to this question. The importance goes beyond the areas where we worry about discrimination, like hiring. For example, if we are training an AI to recognize a medical condition that affects groups differently, it is vital to check that the training data does not homogenize the population. So, I do not think the practices needed to cope with this issue are developed yet. But again, the risk is not being ignored or minimized.

Abdel-Magied

I am a little on the skeptical side, but it is good to hear that it is something that people are paying attention to.

So, I consider myself somebody who enjoys discovering and learning about new technologies. But as the conversation around AI progresses, even I sometimes find myself not quite knowing how to handle it. And so, to hear that, okay, yes, there will be enormous efficiencies — even I cannot help think, well, how is this going to affect the world that I am in? The world that I am in currently is around media and writing and so on. What impact does that have? How do I actually learn about AI in a meaningful enough way for me to be able to be on the right side of that digital divide? How do we stay abreast of all of this, and how do we not succumb to the fear that we might be left behind as individuals or as organizations?

Meyer

I think there are a few final questions we can leave our listeners with that may help. Okay, first, take a clean sheet of paper and brainstorm with someone or some people knowledgeable about AI, what would a new company using this technology to compete with you look like?

Second, take a snapshot of your business processes end-to-end, as they are today. Where might they be improved by AI, and how would you prioritize these improvements?

Last, are you preparing your organization for the process of incorporating these new capabilities? What is the work needed to make the changes? What is the disruption that will be caused when you make them? And how will your organization be different when you are done?

And finally, our better question: When will society be ready to harness AI’s potential? And if we focus on the upside, mechanical automation reduced the average work week in industrial societies from something like 80 hours to less than 40. And in 20 years, I think the productivity gains from AI will cut it in half again.

Abdel-Magied

20-hour work weeks does not sound like such a bad idea when you put it that way. I see you.

Meyer

Yes. Right. And in fact, in our next episode on the future of work, we are going to be talking about what people will do with that time. Specifically, if AI changes how we work, what will it look like? What will work mean? What is its purpose? How will technological changes impact the workforce? And we will be speaking with two terrific guests, Kevin Kelly, the founder of Wired magazine, and Aaron Maniam, leader of the Ministry of Industry in Singapore, and also a poet on the side.

Abdel-Magied

Chris, it has been an absolute pleasure listening to the conversation and having this chat with you. I cannot wait to hear what’s next.

Meyer

And what’s after.

Abdel-Magied

I see what you did there. Well-played, sir. Well-played.

And that concludes today’s episode of The Better Question. For more on this subject, visit ey.com.

If you would like to share your better question with us, leave a review.

I am Yassmin Abdel-Magied. The better the question, the better the answer, the better the world works. Until next time.