• Accueil
  • Accueil
  • Accueil
  • Accueil



Accueil >

Séminaire M2 MICAS "Towards 6G and a semantic communication between intelligent machines : Where do we go ?"

Dans le cadre du module "Advanced Wireless Communications" du M2 MICAS de l’IP Paris, nous organisons le séminaire :

"Towards 6G and a semantic communication between intelligent machines : Where do we go ?"

présenté par Pr. Jean-Claude Belfiore, Huawei France.

Le séminaire aura lieu le Mercredi 10 Février de 13h30 à 15h en visio sur le lien :

Résumé :

In 5G systems, AI is currently helping the network but the intelligence of all communicating devices is essentially human. The communication channels convey information (videos, voice, text …) that is made for the human brain. 6G networks will make the era of communications between intelligent machines happen. The nature of the carried information will be very different of what it has been in former generations. This information will be interpreted, not by human brains but by neural networks, Bayesian machines …

It means that, beyond the notion of information defined by Claude Shannon in 1948, we need another notion which is no more independent of the meaning. A first and brilliant contribution to this notion of semantic information was introduced in 1952 by Carnap and Bar Hillel. It was based on the existence of a language that was carrying the semantics. We claim that this notion of language is of paramount importance. The language of Carnap and Bar Hillel is very simple, Boolean, and based on Set theory. It does not cope with what happens in deep neural networks, for instance.

I will try to generalize Carnap and Bar Hillel contribution to neural networks and to less restrictive languages, more rooted in the real world. This generalization needs the use of mathematical tools developed by Alexander Grothendieck from the 60’s to the 80’s, namely Toposes and stacks.