I also write some aphorisms: https://www.linkedin.com/pulse/aphorismes-et-petites-histoires-jean-rohmer/My current technical research is to explore how to develop large language models WITHOUT NEURAL NETWORKS, but with a pure symbolic approach. Intuition that we need not scalar products to understand a text, that machine learning does not imply linear algebra and floating point numbers. I develop an interactive environment that, for instance, helps me to read and understand La Comédie Humaine de Balzac. Sort-of "Reading Machine", while LLMs look more as a "Writing Machines". I call it "Litteratus Calculus". My main contribution to AI is to have initiated the "Logic and Data" domain (Datalog) with the so-called "Alexander Method": how to make efficient and parallel inferences (abduction) on large databases. I also created and managed the teams (#200) which realized and delivered some of the largest operational Expert Systems. We also were instrumental in the industrial deployment worldwide of Constraints Programming (planning, scheduling). Full details in https://mediaserver.lecnam.net/permalink/v125f59482d6e1ta1vyg/I desperately try WHY (not HOW) LLM work so well i, many cases. Wery little on this WHY is publicly available. Eager to share ...My current reflexion as a citizen is about the hypothesis (pros and cons) of a law to separate human world and digital world. See https://www.veillemag.com/Faut-il-une-loi-de-separation-du-numerique-et-de-l-humain-pour-lutter-contre-le-rechauffement-semantique_a4847.html