EI Portal (PIE) : What is your definition of artificial intelligence? Is it possible to have a simple and “vulgarized” presentation of this concept?
Juliette Mattioli (JM) : AI consists in giving cognitive capacities to an artificial system. These different cognitive abilities correspond to perception, learning, abstraction, reasoning, decision-making as well as communication and action.
We find two paradigms which clash but which will hybridize, in my opinion. On the one hand, data-driven AI, in vogue via learning, computer vision (artificial intelligence technique for analyzing images) adapted to the world of perception and prediction. On the other hand, the more symbolic paradigm of AI, which is based on knowledge. It covers techniques of combinatorial optimization, planning, multi-criteria decision but also knowledge engineering with ontologies. These two paradigms are often opposed when I think they are complementary: solving a complex problem requires the combination of the two going towards hybrid AI.
PIE : How do you define “Machine Learning”, “Deep Learning” and what is their difference?
JM : Machine learning, known in French as “Automatic Learning”, includes technologies such as Deep Learning or reinforcement learning. Overall, learning is about learning from examples, much like a child learns to recognize an apple by seeing examples of apple varieties.
Today, there are many technologies at the base of neural networks. It is thanks to linear algebra that we can infer these models. Deep Learning is a special network with many layers, which makes it possible to do better than the neural networks of the 90s. For example, this technology makes it possible to recognize a face and it is in particular this application which has earned it its fame. . However, to work well, this technology requires many, many examples.
There are also increasingly complex technologies, which also require working from examples, less static than images, such as time series. We can cite the CNN (Convolutional Neural Network).
PIE : What are the main activities in which AI is used today? For which sectors is AI strategic and why?
JM : AI can help solve problems in all areas. It is notably present in all the verticals of Thalès. For example, in aeronautics: it helps air traffic controllers to better estimate delays during landings. In the railway field, it makes it possible to detect obstacles. The Earth can also be monitored from space to identify the impact of climate change, recognize targets or plan defense operations.
AI is also used in engineering such as to design test batteries automatically or to capitalize and manage business knowledge through knowledge engineering. Similarly, for manufacturing, AI makes it possible to develop Industry 4.0 by optimizing the supply chain. In the automotive field, AI applies to autonomous vehicles, in the field of health and agriculture, not a field escapes AI
However, it will not be able to solve all the problems. You have to be able to show its added value compared to solutions based on more “classic” technologies.
In short, many application areas at Thales (and elsewhere) use and will continue to use AI, whether it is the basis of machine learning, symbolic AI or hybrid AI.
PIE : AI has been a popular topic of discussion for several years. What’s new today? Has anything changed since its popularity began?
JM : AI was already popular more than thirty years ago, for example, AI beating Kasparov in 1996. Since then, several things have changed: computing power in the first place. It allows you to do much more complex things and much faster. The availability of data has made it possible to design new approaches such as deep learning. Internet and Open Data have also contributed to the renaissance of AI From these innovations comes the direct explosion of database-based AI approaches, notably thanks to the work of LeCun, Bengio and Hinton.
In my opinion, years and years of algorithms have enabled the capitalization of libraries (scikit-learn). The arrival of the semantic web has also democratized scientific work (google scholar, etc.) allowing the breaking of the scientific silo. All this leads for example to reducing the barriers between health professionals with AI scientists
As a result, the exchange of knowledge is much more fluid. Finally, a lot more money is injected, particularly through the GAFAMs and of course through the video game industry, making AI known to a wider audience.
PIE : Today, how is France positioned on this subject? What are our strengths and weaknesses?
JM : In 2018, France positioned itself on AI for humanity (i.e. ethical AI) in particular via the mission of Cédric Villani. This gave a great boost for academic research.
At that time, major French manufacturers such as Renault, Michelin, Safran and Thalès were not very concerned by this strategy. Some then signed the AI manifesto in 2019, which today brings together 15 French manufacturers. Indeed, trusted AI was their concern. This impulse wave was heard by the government. The latter subsequently launched the Grand National Challenge of Trusted AI or the Grand Challenge of AI in Health. Today, France’s strength lies in being the opinion leader on trusted AI, that is, AI for mission-critical systems.
France’s main weaknesses in this area are based on the fact that the various professionals in the field do not really know how to sell themselves, even if they have real know-how. The direct consequence: the French are not very visible internationally and if they are, it remains individual successes around a person, a lab, or an entity, but not a French ecosystem recognized as a whole. Secondly, we do not have the same resources as in China or the USA, whether in terms of salary or in research teams. Finally, the teaching of AI is not necessarily well structured and remains unequal between the different training courses available. To sum up, France must bet on training in AI, visibility and “to let people know”. However, it is important to stress that the “know-how” is there, because we have very brilliant scientists in the sector.
PIE: We often hear about brain drain, what do you think?
JM : Whether on the side of GAFAM or BATX, both recognize the French formation as well as its talents, mentioned above, because the majority were installed in France. Their research center (FAIR for facebook, Huawei), one of the main sources of this brain drain happens to be the problem of salary attractiveness. There is a multiplication factor of almost two between the salary of a CIFRE PhD student and that of a Facebook PhD student.
The aforementioned IA manifesto is also interested in the problem of attractiveness: how to keep one’s talents in French industries. A working group is also studying the subject and an event was organized in November 2021, where all the engineering schools in France were invited to discover the French industries which have great problems in AI and to remedy this leak.
PIE: How can France take advantage of this technological upheaval?
JM: France must continue to innovate, to make itself even more visible, to file patents and open source around AI to develop this culture of intellectual property while keeping the specificity of France, its specialization in Trusted AI, Ethics for Critical Systems, Healthcare and B2B.
Interview by Yacine Ioualitene for the club AEGE Data Intelligence
Second part on May 06
We would like to give thanks to the writer of this post for this amazing material
[CONVERSATION] Juliette Mattioli: Artificial Intelligence, a new growth lever for economic intelligence? [1/2]
You can find our social media profiles here and other pages on related topics here.https://www.ai-magazine.com/related-pages/