- 21 minutes ago
Category
🤖
TechTranscript
00:05Well, gentlemen, we have one glass of water for the three of us.
00:08We'll just pass it around during the session if you get thirsty.
00:12What a pleasure.
00:13What a pleasure to be on stage with two icons of French business.
00:17My hope today is that we'll get into the ways that AI is helpful,
00:22the ways that AI is not helpful to your businesses,
00:25and some of the geopolitical questions that came to the fore.
00:28So why don't we start, and Rudolf, why don't we start with you.
00:31What is the most interesting way you've implemented AI,
00:35and what is the most value you've gotten from it?
00:39Before coming on stage, you were asking me about football,
00:42so now I see that we're talking about something different.
00:45I was trying to get them to fight.
00:46They support very different teams for very different...
00:49They did support France for Senegal, but they have some...
00:52They almost got in a fight backstage, I've got to say.
00:54I'm a Marseille fan, and Patrick is a Paris fan,
00:58so that was a bit difficult.
01:00I'm a fan for the champions, you know.
01:02Quite easy.
01:03Too easy.
01:05Okay, please go.
01:08There's this thing, it's called artificial intelligence.
01:11It's almost as important as football.
01:12In what ways has it been most valuable to your company?
01:17I think AI has become critical in the way one company is operating.
01:24We at CMA CGM, we are in shipping, in logistics,
01:28and also recently in media,
01:30and our objective is to try to implement AI
01:35in the different businesses we operate.
01:38We have developed a couple of initiatives in shipping,
01:41but I'm sure that we will be able to detail a little bit more.
01:44But again, the intention is to give access to AI
01:49to all our associates within the CMA CGM group.
01:53Maybe let me ask it this way.
01:54So, to me, one of the most interesting things about AI
01:58is that it's amazing, right?
02:01You use the models, they're incredible, right?
02:03They can do things that tech has never done before.
02:06So, on the one hand, it's amazing.
02:08It gets better all the time.
02:09On the other hand, if you look at productivity
02:12and you look at how much companies are actually getting from it,
02:15it's very little, right?
02:17And so, my question, maybe ask it this way.
02:20Do you think your productivity has gone up more than 5% because of AI?
02:26I think it's the only criteria,
02:29if the only criteria to implement AI is productivity,
02:33do something else.
02:34I think productivity is definitely part of what AI can bring,
02:40but it's not the only thing.
02:42I think AI brings much more than just productivity.
02:45I think it gives all the elements for experts or associates
02:51within the company to take decisions.
02:54This is what is interesting about AI.
02:56And then we can talk about productivity,
02:59but it's not the only thing that AI can bring.
03:03Patrick?
03:04I think 5% today.
03:06Mine, sir, I will be honest, is not yet, clearly.
03:09I think you can see it as different.
03:12Productivity, by the way, when people think productivity,
03:15they think reducing stuff, which is not already the objective.
03:19I think there are two ways to be interested in productivity.
03:22It's on one side to get the most of our assets,
03:25to increase the revenues because we have a better use of our assets.
03:29You know, we are an energy producer.
03:31If I can make 1% more running my utilization rate of my refineries,
03:37it represents an increase, I would say, of my numerator.
03:41So it's good.
03:42And I think, honestly, we are more working on this slide,
03:45on this side, generating more revenues by having better understanding
03:49and using all these datas.
03:51And on the other side, it's also shrinking costs.
03:55But in the way, for example, we are using some good example
03:59where we try to make predictive maintenance to try to anticipate failures,
04:04which will reduce the costs that we can have.
04:07So it's on both ways.
04:08Reducing costs.
04:08It's not a matter of people.
04:10By the way, why don't we have yet 5%?
04:13Just because before to speak of AI, you need to have the data foundations.
04:17And I think that people, and the message is that,
04:20and today what we do in the company,
04:21we have engaged in, I would say, large programs
04:25to deploy at a worldwide basis, 25, 26, 27,
04:31the data foundations because we are developing some big platforms
04:35in different businesses in order to have access to all these datas
04:39on a worldwide basis.
04:40And it's not given.
04:40You know, physical datas are somewhere, but then you need to link them.
04:43Then you can build a model in AI.
04:45In parallel, of course, we are already developed some AI case.
04:49We have a digital factory, which is working since 2020,
04:52which is working to develop case.
04:54But I would say it's still at the early stage for me
04:57of what we could deliver tomorrow.
04:59What is the digital factory?
05:01In Paris.
05:02No, no, but what does it do?
05:04What does it do?
05:05It's a bottom-up approach, in fact.
05:07We have generated in the company,
05:09we have some different business case locally on different data,
05:14and they came, and we are working on, I would say,
05:17a very lean way, agile way, in order to develop some specific app.
05:23We have developed hundreds of apps.
05:25Then one of the challenges for a company like us is to scale up,
05:29moving from that.
05:30So with AI and all these models which are coming to the market today,
05:35the idea is not only to make a bottom approach,
05:37but also a top-down approach to identify the critical areas
05:40where we can apply AI.
05:42For example, for us, you know,
05:44it's a company where we are buying a huge amount of data,
05:47seismic data for geoscience, reservoir modeling,
05:51and we really think that here, with a genetic AI,
05:56developing some foundation model,
05:58we could be much more efficient,
05:59and probably we could even have a better way to develop our fields.
06:04So there is there.
06:06AI could help us to model all this modeling
06:09and to extract more, I would say, oil and gas from the subsurface.
06:14Let me ask you,
06:15I'm going to give you two theories about AI,
06:18and I want you to tell me which one is more correct for your company.
06:21So theory A is that the reason there hasn't been large economic gains
06:26because of AI is because there's something of a J-curve, right?
06:29It's hard to learn and implement,
06:30but then eventually you do, and you become extremely productive.
06:34So we're kind of there.
06:35And when that happens, we'll hire more people.
06:39Because if you have engineers who can do four times as much,
06:42why not hire more engineers?
06:43If you have scientists who are four times as productive,
06:47hire more scientists.
06:48The other theory,
06:50you can tell me which one of these is more true,
06:52is AI is becoming more like us.
06:54It can do more of what we do.
06:56It can kind of act more like a person.
06:59And so eventually, we just kind of move people out.
07:01We won't need to have humans in customer service.
07:04You don't have AI in customer service.
07:05You won't need to have humans writing code
07:07because the AI can do it better.
07:08So ultimately, there'll be fewer people working for the company.
07:11I'm curious, which of these two?
07:13I think it depends where in the company.
07:16When I have some customer service,
07:17where, for example, we are selling electricity to customers,
07:20clearly we think the level one, level two,
07:22could be replaced by smart, I would say, agents,
07:26which could do the job rather than having some call centers
07:29everywhere in the world.
07:30It could be very efficient to manage these customers
07:33when it's becoming complex.
07:35But when you speak about engineering,
07:37I'm more looking to have augmented engineers.
07:39I think augmented in terms of giving them,
07:42as what I've said, capacity to decide in a better way
07:47because they would have access to more data
07:49or being able to be augmented, I would say, more efficient.
07:53But it could also, for them, be more efficient
07:55in the way to analyze many scenarios,
07:58for example, developments.
07:59You know, today we are, for example, in renewables,
08:03we are designing some solar plants, wind plants onshore.
08:09And with this type of tools,
08:12instead of analyzing five different scenarios
08:15and design to optimize a plant,
08:17you can very quickly go through 20,000 different designs.
08:22Maybe it's not so valuable,
08:23but so that's a way to, I would say,
08:26to have a better decision-making process.
08:28And that's also another way.
08:30So on this side, I don't need to diminish my people.
08:33I need them to give them the tools.
08:35And one of the objectives for me
08:37is that all these engineers of TotalEnergies
08:39should be quickly understanding what the tool can give them
08:43and being able to use them,
08:45not to create the foundation model,
08:47not everybody will not become an AI, I would say, engineer,
08:51but for sure it will become part of their daily life.
08:54So, yes, there is a J-curve.
08:57And I think I'm absolutely convinced
08:59that we need to embark quickly now on these technologies
09:02because these technologies will help us to jump in the 21st century,
09:07like, in fact, Internet came 30 years ago, you know.
09:11And so I don't know how we were working before without Internet,
09:14you know, without all that, all these tools.
09:16Now we have a new frontier when we need to embark,
09:19and positively, I'm completely positive about it.
09:23I'm sure it will help each of our people to have a better way to work, in fact.
09:30To add on what Patrick is saying,
09:33I think we are at the beginning of something that is really big with AI,
09:37and it would be too soon to say this is the outcome of AI.
09:41We are learning every day.
09:43I think that the critical thing about AI is adoption.
09:47Not all people within companies want to switch to AI,
09:51so you need to train them, you need to explain to them
09:54why is it so good for them to use AI.
09:56And if you start by telling them you are going to lose your job,
10:00but use AI, it's going to be very difficult.
10:03I think that AI brings quite a lot of value to the way we do business.
10:08For example, today, as a shipping company,
10:10we have 11 ships that are stuck in the Strait of Hormuz.
10:14And we are using AI to collect all kinds of information from our company
10:18to make sure that when we push the button,
10:21and I understand it was pushed last night,
10:23we are going to be in a position to exit the ships,
10:27and AI can help us get as much information to take the right decision.
10:33This is the way I would see AI.
10:35Wait.
10:35For the ships in the Strait of Hormuz,
10:38you are analyzing, using AI, sort of taking geosatellite data,
10:44taking information about communications.
10:45Like, what is the information you're aggregating
10:47to figure out how your ships can get out?
10:50We have, within CMA, CGM, millions of information
10:54that we are collecting from customers,
10:57from our experts about fuel consumption,
11:00about number of containers on board,
11:02about distances from where the ship is to where it should go.
11:07All these information will allow us, once it is analyzed accurately,
11:12it will allow us to take the right decision.
11:14And this is where we see AI playing an important role.
11:18Doing it on our own might be challenging.
11:20So the way we look at it is we team up with big names like Amazon,
11:26with Google, and with Mistral,
11:29that help us scale up and move faster.
11:33Just another example of what we can do today, thanks to AI.
11:38It's about sustainability and the methane emissions,
11:42which methane is an agent which has a big eating power.
11:45We have deployed last year 13,000 sensors
11:49across all our operations around the world.
11:52Now we are collecting all this data in a special center
11:56that we have integrated recently called Methane Live,
11:59where we analyze all the data.
12:02And thanks to taking this data with all the data on the equipment,
12:07we are able not only to detect the methane leaks,
12:12methane, you cannot see it,
12:14but we detect the leaks,
12:16and then we can intervene and repair the equipment leaking.
12:22And we do that on a worldwide scale.
12:24And that, honestly, it's again acquiring the data,
12:27putting that together with all the information and equipment,
12:30and being much more efficient.
12:32So we have a world center which is able then to react with our teams on the ground.
12:37And to tell them you have a leak, you need to intervene.
12:39Since the beginning of the year, we have solved 35 leaks,
12:42which were significant.
12:44Why is that AI?
12:45I mean, couldn't you just...
12:46Because you have some models which will analyze all these types of leaks,
12:49and then we can replicate what we observe on one field to other fields.
12:53So it's a question to be more efficient on all that.
12:55Let me ask you about American AI models.
12:58On Friday, the United States Defense Department...
13:02Sorry, the United States Commerce Department said that Anthropic had to change their latest model,
13:07their best model,
13:08so that no foreigner could work on it.
13:10So no French people, right?
13:12Only Americans and Anthropic said,
13:14wait, wait, we can't do this.
13:15All of our engineers are foreigners.
13:18And so they shut down that model.
13:20They shut down Fable.
13:22Had you been using Anthropic at that point,
13:25you would have been in a pickle on Friday night.
13:28So my question is this.
13:30Will you ever trust an American AI company again?
13:35You know, in this world...
13:37By the way, I trust yesterday I spent my evening in front of President Trump, you know,
13:40so I will not tell him that I don't trust.
13:44But no, I think just for joke.
13:47No, I think honestly, in this world,
13:49we observe that geopolitics is going in many areas,
13:53which is maybe not the best.
13:55You know, we are doing 30, 40 years.
13:57We build a unified global world.
13:59I'm sure it will come back.
14:00It's a matter of today we have some events, wars around the planet.
14:04I think we are solving them one by one,
14:07and then we'll come back to a more peaceful world.
14:09Having said that, the answer to that for me is I need to diversify.
14:13It's difficult, honestly.
14:14It's clear that frontier labs like Anthropic or others
14:18are really in advance on some cutting edge, on some AI technology,
14:23so we need to be able to work with them.
14:25When we observe that, that means that I also need to work with MISRAL AI,
14:29which is good for my little, my friends, Arthur Mensch.
14:33But we need to diversify of source of technologies.
14:36We cannot rely on only one.
14:38It's true that for European companies, it is a challenge today,
14:41because in fact you have two big, I would say, ecosystem of AI.
14:46One is China.
14:47The other one is the US.
14:50It's part of it is linked also to the data, to the cloud, so it's not new.
14:54So the question for all of us, European companies,
14:56is also to be able to encourage the development of European solutions.
15:01And I think, for me, Europe is looking to what could be the next vision for Europe.
15:07I'm sure innovation, digital, AI is an incredible journey for all the young people,
15:13and we must embark that.
15:14So it's an alternative.
15:15Well, let me make it more specific then.
15:17When you're analyzing the methane, and you're taking your data,
15:20and you're feeding it into an AI model, are you feeding it into Anthropic?
15:23Are you feeding it into DeepSeq?
15:24Are you feeding it into MISRAL?
15:27Sorry, for what?
15:29The data from your methane analysis.
15:30Ah, my methane today, it's with MISRAL.
15:33MISRAL.
15:34Are you feeding any data into Cloud and Anthropic?
15:38We have, on data, we have two different types of data.
15:41All the geoscience data, which cost me a lot, are not.
15:44We want to have a private cloud, and we are developing it.
15:47What is surface data, like methane emissions, it's not so strategic, I would say.
15:52So I have no problem to share that with others.
15:54By the way, my objective is that all the technology we are developing on methane
15:59is to share it with the other oil and gas companies.
16:01So I will not, it's really a matter of what is really strategic in terms of value for the company
16:07against what, in fact, even if we are in advance on this type of topic,
16:12at the end, it's not so, I would say, the question of sovereignty is not so strong on it.
16:17So it really depends on the topic on which we work.
16:19But one of the consequences, you know, for me, is that today we are thinking seriously to invest,
16:24to develop a foundation model internally on the geoscience data.
16:30Clearly, my choice will be to go more with Mistral AI than with Entropic,
16:36even if we might also think to develop a dual source in order not to be stuck,
16:41because these technologies are moving very quickly.
16:43And so the question for us is to find a partner which will be ahead of the race,
16:50and not suddenly, we've seen recently Entropic taking over OpenAI, which was a big star.
16:56So we need, this is a moving technology.
16:58So for us, the question is to be able also to continue to keep, to take the right partners,
17:04which will develop their competence and will maintain us, I would say, in the race.
17:10Rudolf?
17:11What I will add to what Patrick said is I was lucky enough to be seated next to Patrick last
17:17night.
17:18So I also did not mention anything about the subject to you-know-who.
17:23But you should have, right?
17:26I mean, you might have been kicked out of dinner, but isn't it important?
17:31No, it was not the topic of the dinner.
17:33What I will tell you is we at CMA-CGM, 25% of our revenue is in the U.S.
17:40So we are not going to leave the U.S. because of what is happening with Entropic.
17:45What we are saying is, and I share the views of Patrick, is we cannot put all our eggs in
17:51one basket.
17:52We need to work, we do not work as such with Entropic.
17:56We work with Google, with OpenAI, and a few other American AI tech companies.
18:01But we also work quite a lot with Mistral.
18:04And I think we need to diversify the way we do business on AI.
18:08Not only because of geopolitics, but because every tech company brings something different.
18:14Whether it is French, American, or eventually Chinese.
18:18And I think it's up to us to make sure that we have the possibility of working with everybody.
18:26But at the same time, if we are going to start restricting access to AI models, then maybe this is
18:34not good.
18:35But what would you do, your CTO comes to you on Monday and says, I need a thousand more seats
18:42on Claude Code.
18:43It's the best.
18:45Makes our engineers work faster than Mistral's coding product.
18:49I like it more than Codex.
18:51I like it more than all the alternatives.
18:52I don't trust DeepSeq.
18:53We need it.
18:54You then get that request.
18:56And you know it will make you more productive.
18:58It will help get your ships out of the Strait of Hormuz.
19:00Right?
19:02In the back of your mind, are you going to think, maybe I shouldn't do it because the Defense Department
19:07is going to shut it down.
19:08The U.S. is going to ban French nationals from working on it.
19:11How do you think through that decision?
19:13I think that when it comes to specific applications that we are developing that are strategic for CME, CGM,
19:21we will definitely start looking at a trusted cloud.
19:25And there are initiatives that are being developed in France, and we're talking to Thales about it in partnership with
19:32Google or with AWS.
19:34So we are looking at alternatives, definitely.
19:37Because, again, as I have said, we don't need to put all our eggs in one basket.
19:41But at the same time, what we will also tell our CTO is who is the cheapest, who is the
19:48best.
19:48And not only bring the fact that it is entropic and with what we have said about restrictions.
19:54I think that it's also a matter of cost and a matter of does it apply to what I want.
20:00Patrick?
20:01No, I think, by the way, all that is quite new.
20:04I mean, let's engage with different governments to explain that business is cross-borders, you know.
20:08And I'm more optimistic than that.
20:12You're very optimistic.
20:13The world is coming back together again.
20:15At the end, you will see the business people in the U.S., I know them very well.
20:19They want to develop the business, not only in the U.S., you know.
20:21They want to expand around the world, including on the European market.
20:26So I think we will find ways about it.
20:28But, again, for me, it's not only models.
20:31It's also a matter of data, you know.
20:33And the data today, which you have some dominant clouds, it's also the Data Act is causing some concern.
20:39And we try to find ways to take care of that.
20:43So I think my view is that we need to engage also at business level with the different authorities, including
20:49U.S. authorities, to find a way.
20:51At the end of the day, Europe and the U.S. are allies, you know.
20:55So we'll come back.
20:55I mean, I'm sure we'll find ways.
20:57Okay, can we stay on this for a minute?
20:58Because I would like to think this is true, right?
21:01And we have a business, too, the Atlantic, and we try to expand internationally.
21:04And I would like to think that the U.S. and Europe work together, and China and companies will all
21:11be international, and they'll compete fairly, and a company with the best product and the best people and the best
21:16strategy will win.
21:18But we're not going in that direction.
21:20We're going in the other direction and have been for a while.
21:23Politics are becoming more about sovereign identity.
21:26It doesn't seem like the U.S. and China are going in a good direction.
21:29The U.S. and Europe, we almost took over Greenland six months ago.
21:34So explain where your optimism comes from, because I like it, but I'm not persuaded.
21:44It would say too bad for you, I guess.
21:47No, no, no, no.
21:48I think there is a strategic dimension between China and the Western world.
21:55That's very clear to me.
21:56I'm not sure we should divide the Western world between ourselves.
21:59So today you have this trend.
22:01I think I'm more optimistic.
22:03It's a matter of a few years to put all that in the right order.
22:06And on the European side, by the way, I would say we are not also the way we are trying
22:11to regulate AI much too quickly, from my point of view.
22:15I can tell you much too quickly.
22:17Because today we develop a feeling in Europe that AI might be dangerous, but it's a risk.
22:22I think on the contrary, AI, we must be optimistic about that.
22:26Again, it's a question for all of us to have a better way to work, to enhance even a way
22:32of life.
22:32I just spent three days in China, visiting a lot of companies, looking to robotics, AI.
22:39There, everybody is fine, fan of AI.
22:42There's no, no, no, nobody speaks about protection of data, blah, blah, blah, job discussion.
22:49No, it's not the right way to speak about it.
22:52We will find, everybody will find its place in our new world.
22:57It's just a matter, we will have more capacities.
23:00And instead of spending time of crushing data in a painful way, we'll have quicker access to decisions.
23:07So I think my view is that, yes, I'm definitely optimistic.
23:10But you know, it's Churchill who said, a pessimist is somebody which in each opportunity see a difficulty.
23:17An optimist is somebody which in each opportunity, in each difficulty see an opportunity.
23:23So I'm definitely on the second camp.
23:25And this will come.
23:26And again, let's embrace it.
23:28And that's, for me, today, it's a race to modernity.
23:31And it's an opportunity to make a jump in the 21st century, like the Chinese are doing.
23:36So we should not fight between us, but more looking at it and develop it in a cooperative way.
23:44That's just a matter.
23:45And you will see that business is very important.
23:47The U.S. business people are able to convince even their political leaders.
23:51What I will add is also, I guess, that it's not a question of being pessimistic or optimistic.
23:57It's a question of being pragmatic.
23:59And it's up to us to take the right decision that applies to our different businesses.
24:06Let me ask you a question about AI development.
24:09So clearly, they're incredible French companies.
24:12You run both of them, right?
24:13You can make biggest companies in the world in France.
24:17But most of the AI companies are in the United States.
24:20And, in fact, the vast preponderance are in the United States.
24:22Why is that?
24:26That's a very interesting question.
24:27So we must change this phrase.
24:29Probably, I think it's – we have plenty of French engineers, by the way, working for AI companies in the
24:36U.S.
24:37So they should keep them here.
24:38Oh, Jan LeCun, right?
24:39Yeah, yeah.
24:40In fact, I think it's a question of risk appetite for capital risk.
24:46Clearly, there is a risk appetite.
24:49Directing equity, capital more to new ventures, to startups, to allow this ecosystem to, I would say, expand.
24:57I think it's a major – it's not only a question of financial reform, it's also a question of mindset.
25:04You know, our society in Europe are quite old, in fact.
25:08AI is a question of future, of young people.
25:10So we must give the power to young people, in fact.
25:12And not thinking only about taking care of our elder people, even if it's true that we have more and
25:18more people which are also taking in our society is important.
25:22So, really, it's a question of finding a way to innovate.
25:26Innovation should be, for me, one of the main drivers of the world, European.
25:31If we want to rebuild Europe, it's about innovation.
25:34And we have the young people.
25:35We have the talents.
25:35It's a question of giving them the capacity to spend some time, to start up.
25:41All these startups that we speak in the U.S.
25:43maybe are there, but are not financially positive.
25:46A lot of them are still – I would say.
25:49But they find the people who support them, the equity, the capital.
25:53So that's something that we need to absolutely to change here and to support.
25:58That's why, on our side, we try to support some startups at our level in order to give them the
26:03chance to develop their market.
26:04So that's this thing.
26:06We spend too much – you know, today, a lot of European savings are, in fact, financing the U.S.
26:11market.
26:12I don't know.
26:12Let's keep that in Europe and let's finance the innovation in Europe.
26:16That's what we must do.
26:16You do have this strange view in Europe that companies should make profits.
26:21Let's wrap this up.
26:23One last final thought, Rudolf.
26:24What I would say is what is very good is that companies like Total and CMA-CGM are partnering together
26:31on shipping.
26:34We have set up a joint venture to operate bunker barges, and this is very positive.
26:40Maybe the next step that we need to look at, Patrick, is how can we use AI to do a
26:46better job when it comes to the partnership that we are establishing together.
26:51I promise you we will not talk money because every time we talk money with Patrick, we end up in
26:55a fight.
26:56So we will focus on how to develop AI.
26:58It was a fight until I proposed to Rudolf to make a GV together.
27:01So then I solved it.
27:03Don't forget I was the one to propose a GV.
27:06Oh, good.
27:08That's not true.
27:12I'll let you each renegotiate one term of the deal right here on stage as we wrap up.
27:17But AI, and by the way, there is something also in Europe.
27:19We need to invest in data centers quickly.
27:22We have also in France, it's very well positioned.
27:24We have a very cheap and decarbonized energy in France, a lot of nuclear energy.
27:29We have abundance of low-carbon energy, nuclear and renewable.
27:32So it's an opportunity to build the infrastructure.
27:35Building the infrastructure, we also attract, I think, more and more talents to build on that.
27:40So I think let's not do it in the U.S.
27:43We, Total Energies, as we are an energy producer, we invest, we produce electricity.
27:48So today I've signed some large contracts in Texas with Google and others in order to develop data centers.
27:54But now it's time to develop all that as well in Europe.
27:58And we are negotiating today a large contract in Spain with some of them and in other places.
28:03So I think that's also part of the ecosystem we need to build, to direct part of the equity on
28:08building this infrastructure.
28:10And I think this is also something on which we need to accelerate.
28:13Well, you already have one very important sovereign resource, which is a fabulous, gigantic tech conference.
28:17Thank you very much.
28:18Wonderful time on stage.
28:19Our time is up.
Comments