The term ‘supply chain’ is relatively new, but the activities involved are not as new as we think. Rob and Jackie sat down with Yossi Sheffi, Director of the MIT Center for Transportation and Logistics, to discuss the complex history of supply chains and how technology and AI will continue to evolve supply chain processes in the future.

Mentioned

• Yossi Sheffi. The Magic Conveyor Belt: Supply Chains, A.I., and the Future of Work, (MIT CTL Media, 2023).

Related

• Stephen Ezell and Stefan Koester, “Transforming Global Trade and Development With Digital Technologies,” (ITIF, May 2023).

Auto-Transcript

Rob Atkinson: Welcome to Innovation Files. I'm Rob Atkinson, founder and president of the Information Technology and Innovation Foundation.

Jackie Whisman: And I'm Jackie Whisman. I head development at ITIF, which I'm proud to say is the world's top ranked think tank for science and technology policy.

Rob Atkinson: This podcast is about the kinds of issues we cover at ITIF and the broad economics of innovation to specific policy and regulatory questions about new technologies. If you're into this stuff, please be sure to subscribe and to give us a nice rating.

Jackie Whisman: Today we're talking to Yossi Sheffi, who is the Elisha Gray II professor of engineering systems at MIT where he serves as the director of the MIT Center for Transportation and Logistics. He's an expert in systems optimization, risk and resilience and supply chain management, topics he researches and teaches at MIT and other leading business and engineering schools. His new book is The Magic Conveyor Belt, Supply Chains AI and the Future of Work, and that's what we're going to talk about today. Welcome.

Yossi Sheffi: Thank you for having me.

Jackie Whisman: Tell us a little bit about your book and what prompted you to write it.

Yossi Sheffi: Well, I just went through the pandemic, so a lot of neighbors were asking my wife, "We know your husband is supply chain, what is supply chain? What's going on? I mean, we just hear this term all the time. We don't understand what's going on." So rather than getting all her neighbors and friends and explaining one-on-one, I started writing the book explaining what supply chain about, why they are complex, why people should actually not be mad when something is not available on the supermarket shelf. But once they understand what it takes to get something from a mine in China to a series of manufacturing tiers, suppliers, transport companies, logistics company, dozens of legal regimes, custom regimes and get something to the shelf, they should be amazed when it's there, once they understand what it takes because it's a lot, it's a very complex system.

So I started the book just explaining what supply chain about and trying to get people to understand the magnitude and complexity of it. Then I talked about the technology involved because some people still think that the supply chain is all about moving boxes and driving trucks and it's actually a sophisticated profession, the management of it, the planning, the monitoring, the management, and that got me into AI, which as so happened in November when I was finishing my book, ChatGPT just came on the scene and so timing is an important issue in any endeavor. So timing worked. I even put some comments on generative AI in the book and that's basically the yarn.

Rob Atkinson: So you mentioned the complexity of it. It is one of those things we just take for granted. We see this thing and it's there and we never really think about it. But if you think about the supply chain, it's certainly way more complex than it was 200 years ago, but my sense is it's way more complex than it was 40 years ago for a lot of different factors. Do you agree with that? And if so, why and how has it gotten more complex?

Yossi Sheffi: First of all, depending on which supply chain you're talking about. When Marco Polo was trying to bring spices from the east to London, it was we had much more complex journey than today's supply chain. Actually, I'm in the process of writing another book about the history of supply chain and I start from the bible and you see, and even before. Really early, the people who were bringing stuff from all over the old world, so supply chain are not as new as we think. The term supply chain is new, but the activities themselves are not as new as we think about them. So we talk about supply chain. What happened 40 years ago or so, 50 years ago, when globalization started in forth. That's where supply chain went from mostly regional to global. So people went to China, developed supply in China and China and other countries around the globe and developed the whole system to procure and bring goods to people in the western world.

And this was driven by many things, by the revolution in communication, the internet and the bigger thing was the container, the advent of the containers in the fifties. So it became the price of, give you one example, loading a ship, just loading a ship in a container costs less than one twentieth of loading bulk. It just changes everything. And of course the standard that the same container goes on the ship, goes on the rail, goes on a truck, and you don't have to move the actual stuff from place to place reduce dramatically the cost of bringing materials and product around the world. So yes, it became a lot more widespread complex. In the book I mentioned that companies have tens of thousands of suppliers and those suppliers have thousands of other supplier. It's an enormous web that somehow makes this thing work and it's magic.

Jackie Whisman: How is AI affecting supply chains in the future?

Yossi Sheffi: It's already affecting supply chain. I mean, as you must know from talking to people and doing your own research, it's there already. I mean there are many, many applications that are fueled by AI. One application that is now people are being implemented in large number is robotics in warehouses. Warehouses are becoming automated with robotic. Amazon robots. These are robots that take the whole stuff and instead of the person going to the aisle, the aisle comes to the person and they do what they do. But then controlling the movement of all these robots that move the aisle is all AI, deciding on the sequence of who goes where, not to run into each other, bring stuff at the right sequence to the right place. It's all large optimization problem driven a lot by AI. But AI is every, when we talk about autonomous vehicles, we're talking about AI.

When we talk about, I'll give you another example, large language models, which now all the rage, several companies that I know are using it for risk management. To find out if a supplier of some critical item is having financial trouble, the old way was look at financial reporting from at least a quarter ago, and usually two quarters ago, than at Bradstreet and the like. Today, what companies are doing... First of all, and you could do it for a handful of companies and you may have tens of thousands of direct suppliers, new language models, new large language models are scanning all the suppliers of some advanced companies.

For each supplier, they look into everything that's written about them in social media and regular media and on TV and decipher or try to see if they have problems with executive leaving a long term of payment to their suppliers with some failed project, what have you. Any sign that show that something is awry and then you send a human there to check it out. But the ability to look over tens of thousands suppliers and look at everything that's written about them is something that was not available to anybody before large language model and now they're being trained to do just that. So some examples of supply chain management. There are many others.

Rob Atkinson: That's really interesting. If you think about supply chains, at one level they're pretty simple. You have to make something, then you have to move it somewhere so that it can be shipped and that could be air, that could be a boat, it could be a train, and then it gets to some other place that has to, maybe it gets to a port and it has to go to some city, and then it has to then go to the next level of transportation to the local place, and then maybe it goes into a warehouse and then finally it goes into a store, let's say, or another factory is an input. It seems to me that one of the things that you mentioned, Yossi, you mentioned robotics and it seems to me that autonomous systems will play an important role.

One of the things now in the US for example, there's a big debate about can we move to single engineer trains? The labor unions want to keep two people in the train, but the technology's really getting pretty good where you maybe only need one. In Australia, they have zero person trains for some of the mining trains. I imagine you'll see the same thing on ships. We're moving to trucks where they're platooning, where the second truck is just following the first truck. What's your sense on autonomous systems and the future of supply chains?

Yossi Sheffi: First of all, I would say that it'll take always a lot longer until we see this system in operation and it takes a lot longer for a variety of reasons. One of them, you mentioned, labor unions. They're afraid of job losses. Another reason is regulation. For example, the Italian government is regulating ChatGPT, doesn't allow it, but every government is now looking at regulating AI development, including the US. Thinking about the FDA, like federal drug administration like agency that will not allow any new version to go into the wild without a lot of testing and clinical trials or something akin to clinical trials. So this will slow the development. And another thing that will slow the development is public acceptance. Today's aircraft modern airliner can go gate to gate with no pilot. Not too many of your listeners would go on an airline and fly at 35,000 feet across the Atlantic with no pilot on board.

It's a question of acceptance and getting used to it, and people will even have, I think, some aversion to see a large truck running at the highway speed with no driver on it. Now there'll be platooning is something else, but no driver at all. It may take a while. So first of all, it'll take a lot longer than people think, but in places it's coming, in very specific places. Like right now we do have autonomous golf cart like vehicles in gated communities. You can have this because it's small stuff, it's control. You can put sensors in the ground and control it, but there's a lot of time before we'll see autonomous cars on the road, especially because we didn't solve yet the combination of people drivers and computer drivers. And as we see more and more Tesla accidents and others, it delays. It delays because of acceptance, because of government regulatory pressures. So a lot of it will happen, but it'll take a lot longer than people think.

Rob Atkinson: Yeah, I really liked your, you had a section in the book, I think it's called Implementation will Take Time, and it really struck a nice chord because I am usually the pessimistic one. I'm the optimistic one about how great technology is. I'm usually the pessimistic one where people are saying, "Oh, it's going to happen tomorrow." And I'm like, "Wait a minute, these things take time." There was a guy at George Washington University who had a crowdsource expert futurist thing where he asked, a couple of hundred futurists asked me to be on it, and we'd go in and he'd ask you to, all right, here's this technology, how important it's going to be and when will it happen? And he ended up having to tell me that I was screwing up the model because I was too much of an outlier on certain things. It's going to take longer than you think. I feel that way about autonomous vehicles.

Yossi Sheffi: Let me give you one data point. People are worried about job losses and again, it takes a long time for job losses due to technology because you need to get the processes, the people trained, the technology, it takes time. In 1892 AT&T invented the automatic telephone exchange. Until then, there were usually women sitting in exchanges plugging stuff into holes and connecting people. By 1950, there were still 350,000 telephone exchange operators in the United States. Only by 1980, profession basically went away. It took nine decades. Innovation really work its way through the system. It takes time. There's so many hurdles. People in the technology area, and I'm as guilty as the next guy, get excited about something. Right now, by the way, with all of AI, it's already implemented, there's no sense of increased productivity yet. We don't see it.

It reminds me of Aero who used to be a Nobel Laureate from MIT who are doing the third industrial revolution, talking about computers and software, talking about you can see computer everywhere, but you don't see them in the productivity numbers. It took decades until computers started to help productivity, people start to know how to use them. It became more efficient and productivity went up. Right now, we're in that stage with AI. People are sometimes, by the way, implementing AI just because. I cannot tell you how many times I was, and this is something that every consultant has seen. I work with a board and the board started, when I first listened to a board member ask the CEO, "What's your China strategy?" And it morphed into, "What's your blockchain strategy", and now "What's your AI strategy?"

Rather than looking at what the problem is, what is the best technology or process or people to solve it, rather than start with a technology and just implement it because you want to look cool because you want to answer a board member, who knows. So right now we're in the stage that a lot of it is being implemented for the sake of implementing. It's not really solving a particular problem. So we'll see. We'll see. It'll take time until everything will equilibrate and every technology will find its right place. I should say that the general AI, generative AI and general models are really more fundamental. One can see that 10, 20 years from now, they'll have a large impact, but it's not happening in the next two to five years.

Rob Atkinson: So we should form an organization called Techno Incrementalists, I guess, because almost everybody in this debate, I see all the... I mean President Biden actually just said yesterday, this was whatever, third week of June now when we're recording this, he said yesterday that there will be more change, technological change because of AI in the next decade than in the last 50 years. That's incorrect.

Yossi Sheffi: It's incorrect.

Rob Atkinson: Yeah.

Yossi Sheffi: Of course it's incorrect. If you think about what the technology change of the last 50 years, I mean the last 50 years were pre-internet, pre-WhatsApp, pre-Google Maps, pre-Zoom.

Rob Atkinson: Barcodes.

Yossi Sheffi: Barcodes.

Rob Atkinson: Cell phones.

Yossi Sheffi: Of course.

Rob Atkinson: Computers.

Yossi Sheffi: IPhones and stuff, when you have a powerful computer in your hand, more powerful than what was on the vehicles that went to the moon, it's hard to admit. Yes, there may be breakthroughs, but part of them are not going to be AI. It's going to be if finally we solve the idea of getting power from the sun in some way, finally we solved. There'll be other technology. For example, there was an article today in Wall Street Journal or around them talking about taking CO2 from the air. People are scaling this technology, who knows? But there are other problems that may be solved that have nothing to do with AI. They may be using AI to power some of this maybe, but the problem that they're attacking are really varied.

Rob Atkinson: You mentioned the 80, 70 years before the telephone switchboard operator went the way of the dodo bird, and we did a study a few years ago where we looked at decennial change in occupations from... The government has been collecting this since 1870.

Yossi Sheffi: Oh, yes, I'm familiar with the data.

Rob Atkinson: It took 50 years before the introduction of the Otis elevator that was push button rather than the elevator operator. It took 50 years before we got rid of the sort of last, and there's still a few in New York. You can go to New York City and then you might have an elevator operator, but it takes so long for this stuff to happen. People forget that.

Yossi Sheffi: Just to add to this example, because I mentioned it somewhere, I think I mentioned it in my book, the issue of acceptance. In 1945, right at the end of World War II, 15,000 elevator operators in New York went on strike. New York came to a grinding halt, even though people always saw how to operate the elevator. It was this wheel that you move up and down, people are afraid to do it because again, acceptance. They just thought, oh, they cannot do it themselves, even though of course they could do it, but they didn't and New York had a long strike and finally had to settle with all the operators. And of course Otis then came up with the push button elevator.

Jackie Whisman: How do you view the role of government to help enable all of these transformations?

Rob Atkinson: In supply chains.

Yossi Sheffi: With caution. Look, even beyond supply chain, what will happen to many... Not too many professions will go away. Not too many jobs will go away, but most of the job will change as technology is being weaved into the work and there will be a period, and I'm not worried about people working in large companies because large companies do train their own workforce, but we have a huge number of people who work in nano stores, bodegas, gig workers who are on their own and they may have to change and upgrade and learn some new tricks. And that's where the government comes in. I think at one point, I don't know if it's three, five, 10 years, the government will have to create a fund that help these people get off the job and learn new job or give them some type of encouragement to go to school at night, to go to study online, whatever.

Because the change, if the believer, not people who are a member of Robin Mine, technology skeptics organizations, but people who are excited. But if they're right and they could be, it's the future so hard to predict. There will be some dislocation required and government can support by doing this, by helping people. I am less excited about government industrial policy, about government deciding on winner or loser. We see now that the Biden Administration is investing in chips and all this. Do you remember all the green investment of the Obama Administration, Solara and all these other guys, they all went out of business. Billions of dollars were invested. So the government is not the best at deciding win or loses. I'm not sure about that. I think one of the best thing the government can do and doing this is support basic research. And whether it is the National Science Foundation, the National Institute of Health, the Department of Energy, Department of Transportation, they all have research arms and they support whether it's university or national labs, they support research. That's important.

And of course what government can do is help solve problem rather than create them. If we have a reason, now talking about the US, a reasonable immigration policy, if we have a reasonable healthcare system, if we have a reasonable infrastructure, I go to China a lot and I come back always green with envy because everything works. The trains are high speed, the roads are all no potholes, and then the weather is not much better than New England. They invested a lot. For some reason we are unable to do it. If you know price of the United States tried to build high speed train in California. The idea from LA to San Francisco, they built a few miles and stopped because build a mile of high speed train in California cost nine times more than France. And France is not a low cost country. France is a high cost country. We go nine times more because of bureaucracy, because of corruption because it was never investigated why it costs so much. So one of my next book, [inaudible 00:21:19] Why it costs so much to do stuff in the United States.

Rob Atkinson: That's a great topic. We need to wrap up. A couple of things. I think we probably have a different opinion on "industrial policy". We like the CHIPS Act, but I do think to your point there, one of the things that I think is critical for government to do in this space is to enable these innovations to both develop and deploy. For example, there's a law that the Biden Administration is trying to implement that would freight trains from going to one driver, even if they can show they're safe. We have unions at the LA Port that are trying to limit automation. We don't have the right roads coming out of our port, so we take forever to put in place new infrastructure as you just pointed out. So I do think there's a really critical role. Let's get out of the way. We had Andreessen Horowitz on that and they have a whole thing that they now pushing called Let's Build, America Needs to Build.

Yossi Sheffi: Yes, absolutely, absolutely. And by the way, I'm sure you are in favor of the CHIPS Act, but the CHIPS Act in fact had some hurdle built into it. There's union protection, there's all kinds of social program that have nothing to do with CHIPS Act. Just let the private sector build it. But it's not, there's some social intervention and all this. That's the part that I don't like about the CHIPS Act. If you want to build ships, building in the most efficient way, end of story. But we also, other things that we need, we didn't talk about the education system in the United States.

TSMC is building a ship plant as you know, and they say that in the United States, ships are going to cost 50% more than in Taiwan. And it's not because the price that they pay for engineers, they pay close to as much in Taiwan as in here. The problem is they don't find engineers. They have to train their own engineers. So even our education system is falling behind and you see all the statistics about math and science in schools in the United States, we're falling behind other nations. That's another government push that should be there.

Rob Atkinson: Yossi, I think we could go on for a long, long time. This is a fascinating topic and as I said, I've only read about half your book, but the parts that I read I really, really enjoyed. There was one of the things that I thought was valuable about that book is the detail that you go in a very accessible way about how is this all evolving, what's going on, what are the opportunities? So with that, I want to thank you for joining us. This was great.

Yossi Sheffi: Well, don't forget that all my wife's neighbors and friends have to be able to read it. That's why it's successful.

Rob Atkinson: That's the key. That's the key.

Yossi Sheffi: That's the key.

Jackie Whisman: And that’s it for this week. If you liked it, please be sure to rate us and subscribe. Feel free to email show ideas or questions to [email protected]. You can find the show notes and sign up for our weekly email newsletter on our website itif.org. And follow us on Twitter, Facebook, and LinkedIn @ITIFdc.

Rob Atkinson: We have more episodes and great guests lined up. We hope you'll continue to tune in.

Jackie Whisman: Talk to you soon.