Thursday - 9:45am - 10:30am
Halle 6 - Ouest

From ChatGPT to Devana or how to go from "fast-think" to fighting fakenews from GPT4

Marvin Sant



In 2019, Marvin meets Barbara Delacroix. She tells him about her startup. Its name is Scriptor Artis.
If Marvin has never taken Latin, preferring big data since his first lines of code, the expression will quickly speak to him. It is about teaching others to transmit the right information.
The idea is ambitious. But Marvin is convinced that it's time for him to bring a bit of lucidity in a world overwhelmed by information. He left Webedia, abandoning the Gafa clouds to join Nantes and its pool of startups.
Today, he is the technical director and partner of the deeptech company Scriptor Artis, which specializes in the design of cognitive technologies to simplify and optimize the transmission of knowledge.
In 2020, convinced (by Barbara) that the NPL AI would become a great ally, he eagerly awaits its access to GPT3. The file is solid. The project is on the rails.
The concept is simple: Devana must be able to assist its users in understanding and producing knowledge, in direct connection with current events. Devana must be an ethical AI that increases our critical minds. For this, Marvin started working with researchers to understand how qualified knowledge is produced and how he could transmit it to his AI, with prompts and algorithms. At the end of 2021, Devana knows its V1. 2023, the technological revolution is here. Devana runs on GPT4, its algorithms start to be formidable. This is only the beginning, Marvin's team is already working on a sovereign AI, more ethical, and the first results are amazing.