Révolution numérique, IA et régulation : conversation avec Aiman Ezzat, DG, Capgemini - WPC 2024

Alright, hello, thank you for having me. So regarding it being a revolution equivalent to the industrial revolution, yes, we have a digital revolution. All the rules are changing, the rules of competition. What is important for a consumer is changing. The physical product has importance, but what the consumer values more and more, and is therefore willing to pay for, is intelligence, the digital part. We look at company valuations today, you have the famous few companies in the United States that represent more than 17 or 20% of the S&P variation. So value is increasingly being created at the digital level. We look at a car, what is a car and how it is evolving, and what a client will see in a car is fundamentally changing. Software becomes a very important part and a significant cost of the development and competitiveness of the car in the future. So all the rules of the game, in a certain way, are changing, and that's what's changing. That's why we talk about a digital revolution. In the same way, we can consider that a software developer or a data analyst becomes the worker of the digital revolution. That's what is fundamentally changing. So we are shifting from a physical world to a world that converges between a physical world and a digital world, and everything is changing at the level of companies, states, and how things function. It's more than an industrial revolution, not to make history, but the history of what we call the industrial revolution, different revolutions. Isn't it much more than a revolution? I say this because in terms of change on society, on behavior, etc., it seems to me that it's still much more than anything we've known since the so-called industrial revolution.

Yes, absolutely, absolutely. It's a revolution that goes beyond what the industrial revolution brought in terms of the evolution of societies and interactions, notably between people.

Yes, and like the rest of the digital revolution, AI is not new. I think AI became mainstream with ChatGPT, but AI is not new. It has existed for decades. It's based on statistics in terms of thinking. So it's been there for a very long time. Now it's gaining more and more importance. Why? Because computing capabilities, and notably data management, are becoming exponential. And it's thanks to that that we can manage this data much more efficiently, and I would say, exponential quantities of data. Hence the applicability of AI becomes more important. But the concept of AI is not new, including the concept of generative AI. The concept of generative AI is the ability to not only manage data but also text, voice, video, and to manage that. So it's changing. Why do we consider that AI is the new oil, etc.? But it's actually the data, it's behind how we manage to exploit the data to improve decisions, accelerate innovation, increase the efficiency of companies. That's what is changing. Accelerate education, that's what is changing with AI. It's a real revolution. Why? Because the computing capabilities that exist today for data management allow us to do things that were absolutely impossible to conceive even 10 years ago.

Absolutely. So we had what we'll call, using an anglicism, the AI hype cycle at the beginning of last year. People were only talking about AI, and everyone said it was going to completely change everything in the very short term. The reality is that there's a difference between technological adoption and the time it takes in terms of adoption, and the potential of the technology. So the potential of AI and generative AI exists, but the adoption time is much longer. Why? Because you have to change processes, modes of organization, reflexes, how people work. And you can develop all the models you want, but instilling that into a human being and changing the reflexes they have is something else. So it takes much more time in terms of adoption. But that's the reality of any technology.

Yes, so our examples, there was a lot of what accelerated strongly, this concept of hype around generative AI notably, is productivity in terms of customer service. Alright, the famous call centers where everyone could, thanks to more intelligent agents through generative AI, automate call centers much more. And that created this hype of saying, 'I can do this everywhere.' In fact, no, what applied to this call center, where we can achieve productivity quickly, cannot be applied elsewhere. Now, when we look at AI applications, there are many assistants. So assistants in terms of sales, even in a factory, there are assistants for operators that allow them to say what would be the best next decision based on the current context. So you always have to think about the concept that what we call 'human in the loop' is always important. These are assistants, so they don't decide in place of the operator, but they suggest options in terms of decisions. And so many sales and customer service assistants in many places in companies. One of the places where we see a lot of potential advancement is in R&D, notably in pharma. So the ability to reduce development cycles, to anticipate rather than a molecule not having the expected impacts and stopping it earlier, and behind that, to reduce this famous cycle between the discovery of a molecule and the fact that we can put a drug on the market of 10 years, to be able to reduce this cycle, that is quite possible. And on that, there's generative AI and there's an important impact, and we see concrete advances on the subject.

Listen, the false leads are everything people tried to do in terms of productivity. I think the reports, including for the IMF and others, which put out absolutely dramatic reports about 400 million jobs disappearing because of generative AI, etc., were completely alarmist. And in fact, it has a very negative effect because everyone was very scared. And so in companies, when people talked about generative AI, it was 'I don't want it because it means my job won't exist tomorrow.' All of that was much too alarmist in terms of anticipation. Obviously, we see that after 2 years, we haven't seen the anticipated productivity levels at all.

Of course, of course. Yes, yes, yes, yes.

In fact, the idea is to find the right balance between citizen protection and innovation. The principle I have is that we try to regulate the use of technologies, but not the technology itself. However, when we see the divergences in philosophy between the Americans, who went with concepts saying, 'Here's what we recommend doing,' and the Europeans, who regulate through the EU, notably the AI Act, which regulates the development of technology, and that is very dangerous, obviously, because it kills innovation. Secondly, trying to regulate something that is new, that we don't understand, I find that dangerous. I gave you the example, I think if we approached nuclear energy the way we approach artificial intelligence, we would never have nuclear energy. So we are over-regulating something we don't understand and at the source, we are killing innovation.

It's the Europeans, yes. The Chinese don't regulate anything.

The Americans, they give guidelines, they leave it. So California risks falling into the European camp. So I think, plus all of this becomes very disparate across the world. So for a company to function, if it has AI models, well, if they are in Europe or the United States, they don't have the same operating rules at all. So it becomes nightmarish in terms of operation. And the risk we have is that all innovation goes abroad because startups, companies find it too complicated to operate in Europe.