Understanding AI to Better Deploy It – Coruscant at Stéphane Mallat’s Conference

Artificial intelligence is fundamentally transforming industrial systems.
As its adoption becomes more widespread, a key question arises: what scientific foundations does it truly rest upon?
Coruscant’s participation in Stéphane Mallat’s conference, organised by INSA Toulouse, provides valuable insight into this question.

As part of the partnership between WILLING and INSA Toulouse, the Coruscant team attended Stéphane Mallat’s conference.
Entitled “From philosophy to mathematics: intelligence that is not so artificial”, the event brought together researchers, engineers and students around a central challenge: understanding the deep foundations of artificial intelligence, beyond its immediate applications.

This article highlights the key takeaways from the conference and what they reveal about the links between fundamental science, artificial intelligence and industrial applications.

Stéphane Mallat: a leading authority on the fundamentals of artificial intelligence

Professor at the Collège de France and holder of the Chair in Data Science, Stéphane Mallat is one of the key researchers who have helped shape the scientific understanding of artificial intelligence.

At the intersection of applied mathematics, signal processing and machine learning, his work has shed light on the theoretical functionning of neural networks, highlighting their ability to exploit data structure beyond mere statistical logic.
It thus provides valuable insights for understanding what today’s artificial intelligence models actually entail.

Former professor at École Polytechnique and internationally recognised researcher, he embodies an approach that bridges fundamental science and practical applications, an essential dimension in building reliable and controlled AI systems.

Recipient of numerous distinctions, including the CNRS Gold Medal (2025), he offers crucial light on the still largely unknown mechanisms of artificial intelligence.

Understanding AI to better transform it

One of the key contributions of the conference is to move beyond a purely statistical view of neural networks.
Far from being limited to numerical optimisation, these models are capable of leveraging the underlying structure of data, revealing a form of knowledge that extends beyond parameter optimisation. This is particularly critical in industrial environments, where robustness, interpretability and model control are essential.

This perspective highlights a major challenge for industrial players: deploying AI models is not enough, understanding their mechanisms is crucial.

It is precisely within this framework that Coruscant’s approach is framed:
to develop explainable, reliable and scientifically grounded artificial intelligence by integrating advanced mathematical models at the heart of its applications.
The links established between artificial intelligence, philosophy of knowledge, neurophysiology and statistical physics finally open up a broader perspective: AI does not merely produce results, it offers a new framework for understanding reality.

Strengthening the bridge between research and industry

Coruscant’s participation in this conference reflects its positioning: bridging cutting-edge academic research and industrial applications.

At a time when Europe is seeking to strengthen its technological sovereignty in artificial intelligence, the connection between fundamental research and industrial deployment is becoming increasingly critical to transform scientific advances into robust, concrete solutions. As highlighted by Stéphane Mallat’s conference, the power of AI alone is not sufficient—its understanding is essential.
Understanding artificial intelligence is no longer a scientific luxury: it is a prerequisite for its deployment.

In this context, Coruscant supports organisations in structuring AI initiatives that are explainable, robust and aligned with both business and regulatory requirements.

Looking to structure or secure your AI initiatives?