Today’s episode features an interview with Yves Caseau, Group CIO at Michelin. Yves also previously served as the President of the ICT Commission at the National Academy of Technology in France. Yves is a global thought leader on organization theory, social networks, and computer-mediated communication. On this episode Yves talks about the importance of keeping humans involved in AI systems, adaptations Michelin has made to its supply and distribution chains amidst the COVID-19 crisis, innovation in France’s start-up culture, and much more. So please enjoy this conversation between Yves Caseau, Group CIO at Michelin and your host, Steve Hamm.
Today’s episode features an interview with Yves Caseau, Group CIO at Michelin. Yves also previously served as the President of the ICT Commission at the National Academy of Technology in France. Yves is a global thought leader on organization theory, social networks, and computer-mediated communication.
In this episode, Yves talks about the importance of keeping humans involved in AI systems, adaptations Michelin has made to its supply and distribution chains amidst the COVID-19 crisis, innovation in France’s start-up culture, and much more.
--------
How you approach data will define what’s possible for your organization. Data engineers, data scientists, application developers, and a host of other data professionals who depend on the Snowflake Data Cloud continue to thrive thanks to a decade of technology breakthroughs. But that journey is only the beginning.
Attend Snowflake Summit 2023 in Las Vegas June 26-29 to learn how to access, build, and monetize data, tools, models, and applications in ways that were previously unimaginable. Enable seamless alignment and collaboration across these crucial functions in the Data Cloud to transform nearly every aspect of your organization.
Learn more and register at www.snowflake.com/summit
Steve Hamm: [00:00:00] Hi, Yves Nice to meet you.
Yves Caseau: [00:00:04] hello, Steve. Very nice to meet you
Steve Hamm: [00:00:05] we like to start Yves by talking about kind of how people came into the technology industry or in the technology business.
So if you could start by telling us how you initially got interested in technology and why you
decided to make it your career. .
Yves Caseau: [00:00:21] I went, I dipped typed into, uh, into into computers. I was a teenagers. I was also very good at math, so I, I easy to beat in math. And computer science, but, at last moment, I, the chance to, to do a little bit of both. So I would say the first step I wanted to do, artificial intelligence who took part, at least on my, my advisor told me it was a dangerous topic.
So I moved to, into compiler designs and we'll be systems. So I would say first 15 years of my life, I was in, R and D actually, software research. And then I moved to, to it and, corporate carrier that they have now.
Steve Hamm: [00:01:03] Yeah. So why did your advisor say that AI was dangerous?
Yves Caseau: [00:01:08] Remember, it was the eighties. There was a light here we bought a couple of years . It was the first winter of AI.
Steve Hamm: [00:01:16] Oh, I remember the AI winter.
Yves Caseau: [00:01:18] I moved into object oriented programming, which better choice because, became extremely popular next 15 years.
and then I moved to operations research, which was another way to do smart systems without having the AI name
Brendan
on it.
Steve Hamm: [00:01:33] Now most people think they know what business Michelin is in, but it would be really helpful if you describe it. So we understand the full dimensions of the company's business. And then talk about your
role there as CIO.
Yves Caseau: [00:01:46] So as everyone knows, main business is tires and, in every possible sector from really small things like, bikes up to we launched earth, moving mining trucks. We are the leading brand with the focus on performance, what we call total performance and innovation. And when we mean. Total performance, the combination of driving performance, but also durability energy savings.
Everything at the same time, I'll share purpose is a better way forward. Which means that besides besides tiles, we have several address and businesses. One of them is services and solution for such as fleet management, with connected tires and connected devices. We also operate in travel experience and, such as the world famous mission guide and also high performance materials because we leverage our deep knowledge of materials and chemistry.
I joined Michelin, mid 2017 as the group CEO, Michigan has built over the years worldwide. It was hurting France in us, India and China, and a large number of smaller teams around the world. It's a global organization with shell infrastructure and business solution deployment. And it's also a hybrid organization where our entrepreneur architecture enables local agility.
And when I say locally, that means. Local country or specific business line. And my job as the group CIO is to develop the best technology platform for mission businesses to run their services and to leverage technology such as artificial intelligence and machine news. Pretty. We'll talk about that later on into their own strategy.
Steve Hamm: [00:03:21] Yeah. Now you've been there for about three years. What was your main goal when you arrived
and what's your main goal now?
Yves Caseau: [00:03:29] So three years ago, uh, would you find the strategy as, as a mass loop, Jeremy, that is every layer being important, but the bottom is the most important and every layer. Quotes on top of each other. The first layer was quality of service, deliver our services more regularly with less downtime, which is really what our customers and partners do expect softer bottom layer.
On top of this, there is simplification removing technical debt. We're moving obsolescence, and this is really critical for the next century. This is a, we need to reduce the dead weight. So that we can accelerate faster and more or less, we need to work on modern software tools to increase the rate of change.
Well, there's certainly, yeah. Which is pretty, what we're going to talk most about is, is building the foundation for data driven mission. And, uh, we have always gathered lots of data at Michigan. It's part of our culture deciding based on that as part of our culture, well, the data driven challenge is to make data circulate everywhere.
Uh, to break the silos and to make this much faster, we'll do the fourth layer of the strategy is really system to system integration and in, okay. Expose all data in services and work with sell Papas, but that requires the first three things to work well because otherwise, if you expose your services are not good enough to join no problem.
And the last part of the strategy is really. Becoming innovation enabler for Michelin, which is also something we'll talk later on and improve the quality of experience that is leveraged. So the UX design skills that we, that are really important in this day and age. So to go to come back to your question, but my number one priority when I joined three years ago was recorded of service.
We've made tremendous progress with, uh, the downtime was divided by effecting sector of fall during the last few years. We should have some room to grow. We're not as good as I wish we were, but we've made big progress. So now what what's, this it's really a simplification renewed removing our legacy systems, uh, and making sure that Michaline can run on an up to date, height, availability, uh, software system.
Steve Hamm: [00:05:39] You mentioned shared services that suggests that you're sharing data and, knowledge with some of your
partners. Do I understand that correctly?
Yves Caseau: [00:05:48] We are, we are not as much as we should all as the shoe, they wish we would, but yes, we are already. We have EDI and electric that I entered in their exchange. We have APIs, we are sharing that data. With all, uh, distribution partners with Soma for customer business customers. And he's clearly a trend of the future.
If you look at what's happening with COVID, the growth of e-commerce also, as far as the system to system integration is more and more critical to our business.
Steve Hamm: [00:06:18] Yeah. Now how important are data and data analytics to
Michelin?
Yves Caseau: [00:06:22] So for us, it's, it's very
important. we see this as a, journey. It was a number of steps, uh, and, and learning curve that is there. First, we need that out too, to better see understands, uh, I would say predict and automate our business, that I'm following the framework from the German academia of engineering and, uh, All of this has been true at Michigan for a while.
We use that out too, to better run out on our R and D, but the new changes of digital manufacturing. This is becoming more and more important to classical example of prediction is predictive maintenance. We use it in our factories. We also use. I use it for it. Operations, AI for ups, it's really predictive mentioned applied to 22 operations.
And then once, uh, we can foresee with enough accuracy that we can start to automate the reaction. And we moved to worlds okay. Adaptive and autonomous systems, which is what the, the four steps of the data driven journey. And when we are good at this internally, because all of this so far has been to help me better run our own businesses.
Then we can start share this with, uh, our partners and customers, which was exactly your previous questions and moving to data services. That's what we do. For instance, we do fleet management and fleet management is data driven service, and then even move to data products where you send that data to someone who is going to use the data for that owner benefits and purposes.
So there's a. journey for us. I could give you a number of example from R and D, where we use that to accelerate, uh, the, the knowledge extraction from tests manufacturing. We talked about, uh, supply chain. We use AI to improve the efficiency of the supply chain. We also use data and AI to fascinating the customer interaction, which machine, but.
The, the thing that is even more exciting, as opposed to simply using that analytics to make each function perform better is when you look at the bigger picture and you make the test circulate so that you can treat couplings for sexual size, designing R and D with manufacturing or coupling supply chain with demand management.
And this is where you were the . Enterprise skill, the time management becomes important. It's not only function by function optimization, but. Creating a new picture of the global Michigan.
Steve Hamm: [00:08:39] Yeah. Yeah. That's interesting. A couple of questions. First in your manufacturing, it sounds like you're using AI and data to gonna monitor the manufacturing process and to optimize it kind of on the fly. Is that what's going on? Are you able to actually improve the way things are going or modify them or reconfigure them without actually humans getting involved?
Yves Caseau: [00:09:03] There are humans involved , for two reasons. First it's a journey and a win. One of the great quotes from, uh, who's the CTO of Accenture is that AI, it's a learning system with embedded humans. And I really liked this idea that you have embedded human because to train those systems, you need people with , the business skills with the engineering skills, with the manufacturing skills.
And then once you have created the smart system, no, we are not a delivered why it could operate on the fly. However, it really depends. Actually , when we start to use AI to do defect detection, for instance, we hope to automate lots of these. When we use AI to do predictive maintenance, it works together with the human it's going to create an alarm and say, Oh, you should really take a look at this machine.
And then maybe the technician was a yes. You're right. Let's let's take a look. No, no, I know why you have this alert. It's I'm not going to do it. So I would say that's the Mo the more common case we like to speak about augmented intelligence at Michigan, where we see AI as a way to augment the sadness, the intelligence of the operator in the factory.
Steve Hamm: [00:10:05] The other thing you mentioned was the fact that you, actually sell some of the data you gather. And I wondered what kind of data fits into that category. And what's your kind of your sales channel for
that.
Yves Caseau: [00:10:16] what we,
yeah, because we don't sell the data about our own, what we share with, partners, our owner, I would say supply chain management, uh, inventory management, and also what we know about the market.
So there's a. Certainly we'll have Asians there. Uh, what we are starting to sell is occupy BDT also, too, to tell some of our, uh, partners, what we know about the business from, from what we can observe. So it's kind of a quid pro quo. each of us, we know something about the other based on what we can observe in all that time.
We want to exchange these type of things.
Steve Hamm: [00:10:57] So it sounds like it's more of an exchange than a sale.
Yves Caseau: [00:10:59] I think that we are not at the sale level yet. That is we have a few maybe on, more on a proof of concept type, type of it's mostly it's mostly exchange.
Steve Hamm: [00:11:09] So are you familiar
with snowflakes data exchange technology?
Yves Caseau: [00:11:13] Yes, we've been looking into it. it's part of, our vision of where we want to go. Uh, so we, we see, uh, Suffolk as a, as a possible, I would say tool in the toolbox there, uh, two distribute to share that, uh, across difference. Uh, with a high level of motivation. Interesting. Uh, part of snowflake is it's a good example of what I will pulling AI for ops Belial is automate, uh, get scalable, distributed sharing of data. It was a service which is as suffer, no maintenance as possible. So it's for Michelin is still preliminary. We have a pub going on. We are evaluating snowflake as a performance accelerator for either some of our business processes. And here I'm using, I'm talking about using snowflake as a, as I see that as a, I just smelled that, that sharing platform, but also related to our business intelligence process.
And, uh, the capacity of snowflake to share. Uh, sure. Very fortunately to many different places fits very nicely though. The requirements of business intelligence and especially distributed business intelligence.
Steve Hamm: [00:12:28] Yeah, I get that. I get that. I want to take a step back though. Yeah. What's your view on cloud computing for data management and data analytics? Are you migrating your data to the cloud or are
you thinking
King about it or are you already there?
Yves Caseau: [00:12:43] So when we think
about our data infractions infrastructure,
we think about two things storing and forward and moving forward, and we need to store that, that share and for future use. And we also need to move that out. Uh, for mostly, for real time processing, uh, as far as store is concerned, uh, we have moved to the cloud.
I would say most of our new Jewish metallics are on the cloud. All our advanced that the store for the advanced analytics platforms, they are on the cloud. So we are already there. Why? Because we get most . And we get more agility, better tool. It's a, it's a, it's a newer, date's a newer software stack that gets refreshed constantly.
As far as forwarding data, moving data from one place to another, uh, the cloud gives access to better tools. Same same argument. And I would say access to the current flow of data management. And it's very striking. I would say, I would say that snowflake is a very good example, uh, distributed data management.
It's, it's a domain where there are new technologies or technology improvements most of the time and, uh, leverage leveraging cloud computing is a way to make sure that you can add anytime, get the best services that are available.
Steve Hamm: [00:13:53] So I understand that you're kind of kicking the tires with the snowflake
technology, correct?
Yves Caseau: [00:13:58] Yeah, we're making,
our first assessment. I will. One of our key transformation for, for Michelin. I would say four for the DJ to come it's a longterm transmission is to become a reactive company. That is a company that reacts who events, according to the customer's timeline, as opposed to our own time today.
We still manage most of our processes in batch mode and, uh, during the night with our own schedule and really want to evolve towards even driven architecture, continuous processing. And that really requires improving our game at managing data flows and leveraging the best that that account management has to offer.
And our vision is to move towards something that used to be called the lung, the architecture, where you combine cold analytics on the test tools and hot analytics on that, that flows. And we want to see how we can leverage a snowflake as one of the technology that will enable us to go towards that vision.
Steve Hamm: [00:14:52] So right now, when your, your analysts want to do a major query or set something up, they actually have to wait for it to run overnight and then get the
results the next day. Is that true?
Yves Caseau: [00:15:03] , no, it's not true. BI is
no longer when I was talking batch processing. I was thinking for instance, about sales and operation planning, the planning operations, I would say it's a big business processors. No, fortunately the business intelligence processor much more reactive already today.
Steve Hamm: [00:15:19] Yeah, I got that. So could you be a little bit more specific about the snowflake stuff? You're, you're running a pilot program. what are you testing? what are you looking for? What
kind of results are you looking for?
Yves Caseau: [00:15:29] So we are, uh, as I was saying, we are looking at snowflake too, either accelerate some of our business processes and, and she I'm using snowflake as. Uh, data distribution technology. So let's do the first two, the first area. The second area is, as I said earlier, we have, uh, well, we call it a hybrid model. We have one central set of the lakes.
We have. Uh, that, uh, intelligence platforms, we have three or three of them for each of the zones, which is kind of a shared platform. And then we have fluff, uh, data engines locally for some of the business units and for some of the countries. So we have this sort of multiple scale of, of systems. When you have this, this multiplicity of skill, it gives you more agility, but it also creates.
Chris a question or problem of, of synchronization and sharing. And we have been using a number of technology in the past, and we found that, uh, probably a snowflake would do that faster, better, but mostly with, as I said, uh, we slay separation and motivation. So, so if I want to, to summarize what we expect to learn from testing snowflake is to see whether the technology fits.
Into our needs. And if we can leverage the, the embedded AI too, to a, I would say, leverage automation and make us, uh, more agile, faster, and as always with a better quality of service, as soon as you can automate a destruction function, you reduce the probability of errors.
Steve Hamm: [00:17:04] And you talked about the fact that you have these different zones and you're, so you're sharing the same data, very widely , is one of the attractions of snowflake. The fact that you don't have multiple copies of the data that you just kind of have one version of the
truth.
Yves Caseau: [00:17:17] We all doing
two things at the same time. One of them is to have, I would say. Global data that that has to be shared everywhere. And the reason we have multi-scale BI business intelligence architecture is to also combine that data with local data and local means local to a zone or to a country or local to, to a, to a specific business line.
Uh, and yes, you're right. The, the, the value of, uh, of snowflake in that architecture is to help us. Make sure that we have only, uh, one single source of truth, because it's very important. We want to do two things which are, which look different on the, on the one hand, we want to make sure that, uh, global sales, for instance, they are global.
There is everybody agrees on the same figures. There is nothing a multivariable for a company to have marketing meetings where people have different figures. So you want to make sure that on some of the requests we have. Very standardized way of we standardized data. On the other hand, many of the sales practices, they are different in China, in Brazil and in Europe.
So it means that the BI tools, they have some specificity. And which is why we came up with this hybrid approach. And so the goal with using a snowflake is to, to leverage boosts that is yeah. GDT of local customization together with a back burn where you have single point of fruits and a seaplane version of the truth.
Steve Hamm: [00:18:44] Now, as we speak, the world is in the middle of this carbon crisis. We were a couple of months into it. Who knows how long it will last. How has Michelin
been impacted by the COVID crisis?
Yves Caseau: [00:18:55] So Michelin being a worldwide company, we have been forced to shut down factories around the world and it started in China, then Europe and then us. The biggest impact actually was the drop in, uh, we styled demand and, uh, Oh, She won results. First trimester showed the, that Marsh was banjos with minus 21% in volume and truly is worse at the same time.
Now we are in may, so we are starting to see the recovery in Asia. So that's, that's one of the value of being a worldwide company is that we started getting hit early in January in China, but now we see that the recruiter is there. Major impact from where I stand for for the CIO is the use of data and technology.
Uh, and. There has been two major areas of concern. Firstly, the use of the degree for remote working and secondary use of business intelligence to monitor the sales very closely, close to real time. And we're back to this BI architecture, as far as remote collaboration is concerned. It turns out that two years ago.
Remember I told you my, my number one priority was quality of service, and we decided to invest into our software stack from, from the network architecture, to who we, I made a significant. Investment in cloud solution from Microsoft, from office three 65 to teams, and that strategy has proven success and our cloud based solutions have scaled very nice, very, very well, very nicely.
And, uh, yes actually has never been as popular as it is today. I don't know, as it has been for the last few months, thanks to the they're really good performance of, of those new tools. As far as business intelligence is concerned, we have reinforced all teams and processors to implement what we call. Oh, TrakCare level of service because, uh, when the business depends so much on how fast and how serious the recovery is, we are making sure that we deliver the best operational excellence and that all our, since monitoring tools are working, uh, with absolutely zero defects from a really to data quality.
Steve Hamm: [00:21:03] So obviously supply chains, distribution chains have been disrupted all around the world and it's a really calamitous kind of experience. This is not, this is not minor. Is it making Michelin kind of re-examine the way it. W w you know, where it builds things, where it gets supplies from where, where, how it distributes those kind of fundamental business questions.
And if so, are you using data analytics to kind of figure out what's going on
and figure out how you might do things differently?
Yves Caseau: [00:21:36] So let
me address the supply chain first and the strategy there as far as our supply chain is concerned, Michigan, as I said, is widely distributed around the world. We have factories in every five continents and most large. Major markets. Uh, I would say so far, our strategy, uh, for, uh, has worked properly.
That is we, we have enough redundancy and we are, we've been able to operate. We have reached out to a number of the factories earlier. So that's some of the specialty products, because I would say that traditional tires, the regular tires, as I said, the demand went down. But for the specialty tires, that there was a little more.
Needs a bit more resilience and we've been able to operate whatever flows of products we needed. Uh, so we considered that, that it was the, the, the originals of the suppression was there. You may have noticed that I'll see you is very keen on describing our strategy as being everything's sustainable.
And, uh, what we've seen, uh, tend to reinforce this week. We see that there's a tension between, uh, I would say the post COVID world and the necessities for co two emission, uh, reduction. Uh, there are a number of example, people. Seems to want to move from private public to private transportation. And the social distancing is not the friend of, of transportation efficiency that that's clear.
The increase of eCommerce means that, uh, well, fast delivery means usually less fuel efficiency because trucks are less efficiently men manage. We have a number of, I don't have a crystal ball. Nobody knows exactly what 2021 is going to be, but we see this tension between, uh, Could be the Curry than in the need for a new green world.
And when there is a tension, there is a room for innovation. And this is exactly where Michigan wonderful wants to be. Uh, we think that we have to innovate so that there is no choice to be made. We can deliver both health protection and ensuite reduction. Safe transportation and returning the planet to a more sustainable level of, of greenhouse gas emissions.
So that that's really where Michelin wants to be, uh, solving those, those, those tension with product innovation and that is needed here for four, for three reasons so that we did even more value because at the end. Uh, this is all about efficiency and energy efficiency, more innovation and more agility.
Because as I said, I don't have a crystal ball and we don't no exactly where the market and where those opportunities are going to be. So if I look at what is important to get agility from data, it's really about that architecture, having a common data model and. Cloud is an accelerator. You asked us, can we put all the data?
Oh, okay. Some of the data on the cloud and the reason is the answer is yes, we are putting that on the cloud because it makes us more agile. It will. Help us to adapt to whatever the, the opportunities are going to be. Data for innovation is really about circulating the data and it's a data driven ambition.
I was mentioning making sure that that, that, uh, moves around that that's how we can treat services and solutions for customers. And they're even. Getting more value from that data. It's about artificial intelligence and machine learning. So my job at Michelin is making sure that we have, I provide to my businesses.
I would say data friendly software ecosystem so that each of them, they can leverage AI and ML to be more efficient at what they're doing, whether it's developing new tires for electric cars that will, uh, consume even less fuel or mini. Many of the, sorry. The opportunities that we will find along the way.
Steve Hamm: [00:25:15] Now you talked about the tension between sustainability and some of the lessons we've learned and the adaptations we've had to make to the COVID. are there ideas that are already popping out in the company that could lead to, methods for mediating between
those
two points of
view are those two values?
Yves Caseau: [00:25:34] So the
different categories,
they read the product innovation and getting more performance out of tires. And that I can talk about because of it's public knowledge. We do, we're working very hard on, on long lasting tires with less. Impact on the environment. So as I said, better fuel efficiency, then there is the service and solution where you better manage your fees.
Remember what I said about eCommerce and the need w we need as a society to find a way to manage all the fleets of trucks, even better so that we can. You too, enjoy the fixed indigent convenience of, of, uh, uh, e-commerce without spending. So, so, so much energy. And then there is the deeper inhibition.
Steve Hamm: [00:26:18] before you move on. So you talking about optimizing delivery routes and things
like that.
Yves Caseau: [00:26:24] Yes, that has been around for a
while, but the, the, the, the, the progress in, in two, in mid with sensors, we with, with, with captors, uh, connected objects and connected tires, the connected tires they already exist for, for the last large trucks, because the law, when the tire is big enough, it makes sense to add the cap.
We inside the tire and then you can drive more value out of your tires. So what's going to happen is the miniaturization trend. The fact that sensors are becoming cheaper and they're becoming smarter so they can sense more, more signals. Uh, we're having to improve studies. There is nothing new about a smart routing of your fields, but.
What we're going to be able to do in the years to come we'll, we'll be better. We'll be more, more efficient. And the third dimension for us is really it's material science. It's inventing new solution based on all deep understanding of chemistry, because at the core we are a chemistry company, but that I cannot talk too much about first.
I'm not, I'm not an expert and this is very proprietary.
Steve Hamm: [00:27:33] you know, these days passenger cars have these have, uh, sensors in the tires that say with their underinflated and things like that and send you alerts, but you were talking about some kind of, much more sophisticated sensors for the big equipment tires.
What, what do you, I mean, this, obviously isn't a data question, but I'm just kind of curious, what kind of sensors do you have in
there?
Okay. Well, one of the things that,
Yves Caseau: [00:27:57] that we can measure is the temperature of the tire. So for instance, we have a service school, Ruth connect, where we for hypothalamus calls, we have a small sensor that we put inside the tire that not only gives the pressure, but it gives the temperature. It gives more detailed readings.
And if you want to get the best performance for your risk, it actually helps a lot. Uh, we can also measure. The, the wear of the, of the tire. So that's, uh, there are a number of ways we have connected objects that you put outside your truck or outside your car, and you will walk, uh, near by or on top of them.
It, it would give you a reading of your, of the, of your, where, and that that's important. You, you can replace the tire just at the right time. You can also did you, uh, the wear of the tire from the vibration, from, from, from those measurements. So. There are a number of measurements that you can make that will, you should have the knowledge about the geometry and the, how the tire is built, which is exactly what machine is very good at.
You can, uh, infer from what you hear offered. You measure how the tire is behaving. This is really where, where technology is going.
Steve Hamm: [00:29:04] No, that sounds great. I mean, the people have been talking about the IOT revolution for years, but it seems like it's finally here and it's translating into a lot of
knowledge, a lot of data that is really actionable.
Yeah, for
Yves Caseau: [00:29:16] instance, in the world of mining, it's already there. All the mining trucks, they are amazingly a fitted with skeptics and sensors and the way they are handled and manage is already very optimized.
Steve Hamm: [00:29:29] Yves you have a leadership role within Michelin, but you're also a member of Francis national Academy of technology and the president of the ICT commission. So you kind of have this much broader national purview. What exactly are your roles there and what are you trying to
accomplish?
Yves Caseau: [00:29:46] So the national Academy
of technologies, the French equivalent of the, any national Academy of engineering in the United States, actually, I have cool organize the, the frontiers of injuring concerns with within a couple of years ago, that that was precisely about big data. So. National Academy of technology, uh, work like any Academy works on questions or issues, which are either given by the government by, by institution, stake stakeholder, or that's our emergence, which has self safety clear or self selected by the academia.
And we work, we produce reports and how do we work? We use the collective intelligence. We have our own network of experts, but mostly we listen. We listen to. Uh, technology vendors to startups too. Uh, the best scientist in France, we, we, uh, talked to, uh, companies that are using the technology and we produce reports as most Academie do, uh, with recommendations.
And our goal is to contribute, to help friends, to better leverage technology as a society from, from the citizens to the corporation. So that's really a, the, the goal of the Academy. The the, it has a motto and the motto is about progress. We want the technology too, to reflect progress that is shared and understood by everybody.
To give you an example. Uh, the past few years we have issued two reports. The first one was on, on big data as a paradigm shift at the sec. Second one was about artificial intelligence and machine learning. And the goal was to educate, uh, our stakeholders, uh, both companies, but also, uh, the governments about what is necessary to make those new technology create value for companies because France is a paradox as a country.
France has. Uh, lots of talents. Top are scientists starters and French citizens. They are very fond of products and solution, but digital production solutions. But most of them come from American companies. That is where we are not eating our own dog food. And, uh, what you could find in our report looks a little bit like.
Thankfully last book, AI super poor, but is it emphasizes the role of software and the rule of ecosystems. And it's not only about science, is that the, the, the software ecosystem do matter to create value from artificial intelligence?
Steve Hamm: [00:32:03] Yeah. Yeah, but we talked to the founder of debt. I CU a couple of days ago and I hadn't realized they were in, in Paris. Um, so is there a pretty good, healthy, uh,
tech startup
Steve Hamm: [00:32:16] industry in, in France these
days?
Yves Caseau: [00:32:18] Yes,
the tech,
the startup ecosystem in Paris has grown a lot in the last 10 years. And now it's, uh, uh, it's fairly sizeable and France has a tradition of mathematics during the teaching during school and so on. So we have. Brick metal metal medical school. And I would say the level of, of, uh, of, of training of most a computer scientist or, or, or engineers is such that we've always done well in, uh, developing these types of new technologies.
If you look at Silicon Valley, there are lots of French, uh, software engineers trained in the French nearing schools. And that's now, uh, visible in Paris itself. That is we have in France, lots of stuff. Which is why the question now is what do you need to do as, as, as a country to make sure that that ecosystem grows and create values, but value locally, uh, as opposed to being exported abroad.
Steve Hamm: [00:33:15] You know, two of Snowflake's founders are French and, uh, but you know, back in those days, of course, to succeed in the tech startup industry, you basically had to go to Silicon Valley or people thought that, and they did that and
they, and they, and they, you know,
work their way up
in Oracle and
have done very well for
themselves.
Yves Caseau: [00:33:34] Yes, I know.
Yeah. This
is exactly what I did by the way
that is after I finished my
PhD in France, I moved to the States and I worked there for a couple of years because I figured that. As a software as a scientist, the best place for me to exercise my talent was the U S so it's a, it's part of the, of the culture.
Steve Hamm: [00:33:50] Yeah. Yeah, but we're democratizing data. We're democratizing technology and Redux democratizing AI. I mean, it seems like there's a tremendous kind of trend toward distributing the tools and the power and the knowledge global really
these days.
Yves Caseau: [00:34:08] Yes,
absolutely. This is true, but there is also the know how you need a critical mass of injuring, no house, somehow imagination, creativity, or, or deep data science skills. They are no, you have great people in Switzerland, in China, in you can in Israel, for instance, France. So many other countries that are extremely good.
But, uh, there is something about the software culture, which is still, uh, where the U S has an advantage. And this is really where, okay. At least for friends, we have to emulate and we have to develop that, that Suffolk culture.
Steve Hamm: [00:34:43] Right, right. Cool. Hey, we've talked several times about AI and I noted that you participated in a singularity university program a few years ago, and that's the. The institution set up by Ray Kurzweil and in Silicon Valley. And so I I'd like to know kind of what's your view on how AI will shape the future of business and society, and even the experience of individuals.
Yves Caseau: [00:35:08] So this is the big question. So let
first, I, I do believe that AI will change the way we run our businesses. So I will the first part of the question. Have something to say with a good level of confidence. And then I will address what the society question. Uh, first I, as I said, I see software AI as a software modality.
That is, there are those two great quotes. The first from the countries and software is eating the world. And then the CEO of Nvidia said an AI engine software. All of these means that AI is going to intensify the digital transformation. Because it increases the value that company can extract from data exactly what we were discussing actually.
So it will also increase the gap between the digital savvy companies and the digital legato, so to speak the ones that are, uh, so it's, it's creating a very interesting, uh, uh, Competition too, to look at. If I look at what I've learned at the national Academy, by interviewing the best in class first future, that that has more value than past data.
So it's really about treating flows and growing, uh, that I injuring as a practice, as opposed to simply no learning from the past, uh, more that give you more insights than most services. And if you have more services, you have more data need to, it's a positive circle. Uh, so at the end, what you're creating is you are creating a reinforcement cycle with that, and you're using AI as a way to accelerate the knowledge acquisition there, which is back to what, uh, Dakota was giving from Paul Dougherty.
That AI is a learning loop with ambulate humans. Uh, and, and for me, if I still in the context of, of the. Business of the future. I see. Yeah. Uh, as us yielding a new way of working, which is the, the, the big, the biggest change I, that I can see coming is the rise of the cognitive assistant. So sure. Today we have Google search and we have Siri boats.
I think what's coming, it's so much better and it would change the way we work. We also change the way we collaborate, because whenever we collaborate, we have to share context and do more. We can use advanced techniques to, to, to say compress the context too. Uh, then the faster we are. So I, I, I believe that, uh, yeah, uh, AI, he's going to change the, we weren't all businesses.
As far society. It's a more difficult question. I see that. Yeah. He's bringing the continuous pass towards more automation and it's a long pass. I don't have a crystal ball. There are many challenges that, that are hard, that, that science and technology will have to address. And you pick the low, I think, longer than what we think, but still I think the direction is very clear.
Everything that that could do to me, it will, at some point sooner or later, and that that's, it gives me two trends. The first one is that. We will move our focus more towards human interaction. That is the rest will be done by baby machine. So what, what is being too become valuable in the future is human interaction.
It means focusing on emotions because emotions are sourcing, which you unique to you been at this for the time being, we value more local interaction. And that's interesting because it fits each for, uh, jobbing down the green gas, the greenhouse gas emissions. The green and in a, the more human interaction business society that good, they go together.
Uh, and we're going to value probably more. You mentioned and personalization because. All of these big, this possible, this new world. And, and second to come back to what I was saying earlier, I think that AI is going to open new boundaries on human activity because it's complexity. When you have a complex situation, you can use AI to help you solve the complexity.
I was talking about digital manufacturing and what we do at Michigan, our factories, and the reason we're using yeah. In machine learning is because we have. Complex manufacturing processes. And by using AI, we can, uh, manage that, that complexity better. Okay. If the AI helps you to hide that complexity, then, then you can look at new frontiers using things that could be I've been overcomplicated or over complex.
And that now become possible, which is also, uh, a message from a, uh, human plus from . Uh, th the use of AI is going to. Creates new edges, new boundaries, new frontiers of what we can do in the future.
Steve Hamm: [00:39:33] True. Yup. Well, um, that was really interesting. I mean, I, you know, I think one of the, the things that we haven't gotten to in AI yet is general artificial intelligence. And that's one of the things that Kurzweil talks about and the people. Kind of contemplate with a mix of excitement on one hand and kind of concern on another.
Do you, do you have a sense of what are the guiding principles for how we should look at AI and how we should position it in our
society?
Yves Caseau: [00:40:06] So first I
should say that the way. If I had to, to make, yes, it's a guess, but I think system of your STEM is going to wait to create, it's going to be the way to create really smarter system that is today. For instance, with deep learning, we are able to be extremely good at recognition, sound recognition, voice recognition.
Uh, but there is not one single ticket. I think that that solves all problems and it's very hard for me to imagine. Okay. Uh, General AI, the clergy, uh, I've read as you probably have many of the arguments and it's a right. It's how to, to grow value from it's too speculative. Whereas on the other hand system of systems, they are already there that if you look at the smart, uh, like two days robot in Japan with the boat too, to enter the Japanese university competitive exam, she uses a combination of technique.
Uh, It's very clear to me that we are going to build smart systems that will combine, uh, different types of AI for reasoning. I, for planning. Yeah. I, for building a small memory, uh at that system of systems scale, the question that you're asking about the ATX about the control. It's easier to address that is if you tell me we have formed a wonderful mathematical equation, this represents the way to create.
Uh, general artificial in intelligence. I don't know how to tell you how to control it on that. It's a speculative question, but if we say tomorrow, we'll have a smart dog and the small dog will do so many smart things or smart robots at home. Uh, but it would be a combination of subsistence. Then we have, we have discipline.
We have scientific discipline too, to a, to test to control, uh, maybe to validate, to say, If the robot decides the color of my, uh, of my room. I don't care if it's two 30 feeble. Yeah. You said if the robot is going to decide to close, open the door, there should be an algorithm that proves that is correct.
That it will never miss with the fingers of my kids and, and so on. And so, uh, I think we will, we will evolve from having those discussion. I would say, you know, theoretical matter about algorithms too. A discussion about the system to have sustained injuring. And it's, it's easier. It's easier. It's way it's a, there are, yeah, there are frameworks to start addressing those issues.
And then, uh, as, as we make progress, we'll, we'll find you with difficult problems to solve.
Steve Hamm: [00:42:34] Yeah. Yeah, no, I like the point you're making about systems of systems. I mean, we, these days we understand just how complex the world is. And even as we get more and more and better tools for understanding it. You know, we still, in fact, one of the things that makes us do is appreciate just how many kind of interdependencies there are and unintended consequences and things like that.
So it seems like that is a wonderful use of artificial intelligence. It can be a tremendous aid to humans and the all that focus that we've had at various times in the, in the history of AI on Oh, trying to make an AI. Thinks it's as smart as a human or it can do all. It can think the way human thinks.
It seems like that's kind of, an academic exercise or, or an exercise in imagination, but practically it's not very, it's not as useful as these other things that we're doing with AI and that we're making lots of, lots of progress with. So I think that's very encouraging.
. So Yves, I want to thank you so much for your time today.
I think it's really, it's been really interesting, you know, both to hear about what Michelin is doing and how you're the directions you're going there, but also to hear some of these higher level thoughts. I mean, you're, you're, you're really a thought leader, not just for the nation, but for the world. And I, and I think, uh,
it's been, it's been fascinating to hear what you have to say.
Thank you so
much.
Yves Caseau: [00:44:03] Thank you very much, Steve. It was a very interesting
conversation,
indeed.