<trp-post-container data-trp-post-id='26829'>L’IA view by the father from Mercator

Julien Levy is the author of the best-selling MercatorHe is also Director of the HEC Digital Centre and of the AXA Digital Strategy and Big Data Chair.

Adwise : In a recent speech at the Adetem Factory, you said that one of the major challenges of our time was to "produce and exploit digital data on everything, everywhere, all the time, by any means": is this a real need, or just an obsession?

Julien Levy : One doesn't exclude the other!

In fact, for Adetem, I have traced this imperative back to a much longer movement that began in Europe in the early 17th century, that of the quantification of the world by science. Deciphering the world by quantifying it through the magnitudes of physics and the language of mathematics has been the great business of science since the Renaissance. As Galileo wrote in Il Sagiattore in 1623, "The world is written in mathematical language". The great physicist Max Planck deduced: "What is real is what can be measured".

As measurements are nothing more than data, the digital revolution is a natural part of this movement, as it gives us an extraordinary capacity to collect, process and exploit this data.

Today, we digitise the news (all the press and its archives are online), human knowledge and the arts (books, painting, music, photography, cinema, etc.), social relations (social networks, user-generated content, etc.), our geographical environment (cartography, geolocation, etc.), our physical environment (visual recognition, probes and sensors), all our online behaviour, our bodies and the way our brains work via portable objects, logistics, the various production and distribution operations thanks to the use of mobile devices, etc.), our physical environment (visual recognition, probes and sensors), all online behaviour, our bodies and the way our brains work through wearable objects, logistics, the various production and distribution operations thanks to the Internet of Industrial Objects, etc.

We know that IBM has published a study establishing that 90% of the world's data has been produced in the last two years: the movement is exponential. That's why I said that today's imperative is to produce and exploit data on everything, everywhere, all the time, by all kinds of means.

One of the consequences is that the old divide between "digital" and "non-digital" businesses no longer makes sense. All businesses are digital because data permeates all human activities, particularly economic ones.

Adwise : We produce, we process, and yet, faced with these mountains of data, the mess is immense ...


Julien Levy : The paradox is that we produce much more data than we process. To take a very simple example: a web user's browsing on a site produces data, which is what enables the site to display pages and carry out actions. But the exploitation of this data beyond this function is often non-existent or very limited. At best, the data will be used for statistical analysis of the site's performance and to improve its ergonomics or traffic management.

The raison d'être of big data, as we know, is to exploit not aggregated data but masses of data to produce more knowledge and intelligence. But there's a long way to go between the promise and the reality. The reality for businesses is, as you say, that data is wasted rather than over-exploited ...

In fact, since data is produced for very limited functional purposes, it is a considerable task to identify the sources of data within the company, collect them, format them and make better use of them. We often get lost along the way before we can make the most of it.

As data scientists like to say: "We spend 90% of our time collecting data and the remaining 10% ... complaining about it! That's why, beyond the utopia of a "data lake which would collect all the data produced, we more often start from specific objectives, which leads us to identify the data we need and then see how to collect it. Conversely, we may start with a limited set of data and ask ourselves what we can learn from it.

The challenge of wasting data is not that we have too much data - that's what big data and artificial intelligence are for - but that the way the company is organised is not designed to make full use of the wealth it produces. In other words, it's an organisational issue.

Everyone is talking about the artificial intelligence you mentioned, without necessarily knowing ? how it works: isn't that also part of the reason why it fascinates us? And yet, it's just algorithms ...

What I find most fascinating about artificial intelligence is that the machine produces intelligent responses without being intelligent. It's a paradox that we find hard to understand. Incidentally, the Turing test is based on the idea that there is artificial intelligence when the human subject, in an exchange, is unable to discern whether the other party is a human being or a machine.

So according to this test, intelligence is inferred from our perception of the machine with which we are interacting, not from the machine itself, if you follow me. In other words, the machine's intelligence is defined by our perception of it, not by any inherent quality.

Because the machine only does statistical processing. For example, visual recognition. An experienced dermatologist can distinguish a malignant melanoma from a mole. The machine is just as good, and soon to be better. The inference is that the machine is intelligent because it gives an intelligent response. But the machine doesn't have the slightest 'idea' of what a melanoma is, and it can't 'see' anything else (for that you'd need a pupil and an optic nerve). What it does process are pixel values. It learns to algorithmise these pixels to give the correct answers of "melanoma" and "mole", based on thousands of images that human beings have previously labelled.

Through a process of trial and error, the algorithm becomes sufficiently refined to predict the right answer: this is the principle of machine learning. The machine gives an intelligent answer, but nothing resembling thought in the sense that we understand it.

Adwise : If we look at the Gartner curve, where does AI stand?

Julien Levy : Ah, having had a bit of experience, I've seen fashions come and go. You know, I'm a sceptic, or at least a critic by profession. Every day I see the gap between the promises of gurus or solution vendors and the reality of business practices. On Gartner's hype curve, we could place AI on the rising curve at approaching the "exaggerated expectations. And yet I don't think we overestimate its importance.

Firstly, because AI is part of a long history, that of the quantification of the world, which has become the digitisation of the world. The principle resulting from this fundamental trend is, as we have said, that it is a question of It is now possible to "produce and use digital data on everything, everywhere, all the time, by any means". Artificial intelligence needs masses of data to enable learning, and these masses of data need automatic processing, particularly AI, to produce meaning and be exploited.

Secondly, because the technology is still in its infancy. While the history of AI failures is long - it has even been called the winter of AI - the artificial neural network approach (the heart of machine learning) emerged only later, and its snowball effect is very recent. When I was writing the latest Netexplo study of global digital innovation trends, I noticed that in the space of a few months, what had been identified as being at the cutting edge of the issues had ceased to be so: the cycle of invention is quite staggering. This is an area where investment is massive and progress is considerable and rapid.

Of course, technological invention does not necessarily translate into innovation. Innovation requires use, i.e. social appropriation. Use involves considerations that are not just technical, and the process is much longer. But AI is already penetrating familiar objects: Siri, a search engine and an autonomous test car are just a few examples. But this is only the beginning of a movement: just as the spread of the autonomous car will turn the transport sector upside down, AI will no doubt profoundly affect marketing, as the Internet has done in the past. It appeared in the latest edition of Mercator and the next few years will see it take on increasing importance.

The best way to avoid making mistakes in forecasting ... is not to make any. Without saying what tomorrow will bring, I can say what I see emerging today. On the one hand, there's marketing that's very much geared towards innovation and creativity, enriched by design thinking. The aim is to infuse innovation into the major aspects of marketing policy, and the expectation is that marketers will be creators of sorts, 'imaginers' if you will.

The other type of marketing is very much geared towards quantity, data and data science. It's all about making better use of the under-exploited resource that is data, as we have seen. The more AI develops, the more sophisticated data science expertise will become, because the basic expertise will be processed by the machine.

In both cases, data and AI will affect marketing. The more automatic processing there is, the greater the need to make better use of data. Conversely, the greater the need for creativity - a field that will resist AI for a long time to come. But we shouldn't pit these two profiles and these two areas of expertise against each other, because a large part of the work of creative people in marketing consists of imagining how technology can be used, which is why these creative people are also responsible for the development of new products and services. "digital savvy.

It's a twofold challenge for the schools that train the talent... but also for the companies. When I look at the consumer companies that made me dream when I was a marketing student at HEC, well, they seem to me to be completely out of touch today. I recently looked at the placements of students in the Digital major at HEC over the last five years: I didn't see a single student who had joined these companies, not one. If I was in charge of recruitment in these companies, I'd start to worry about that!

(*) Have digital skills.

They we have fact confidence. Discover our achievements