🎬 Retour en images sur le #CES
Janvier démarre fort. Rencontres, échanges, signaux clairs : open source, souveraineté et maîtrise des technologies sont désormais au cœur du débat tech.
Le #CES a marqué un point de départ. La suite s’écrit ensemble 🚀
#LINAGORA #Tech
GitHub - linagora/twake-drive: The open-source alternative to Google Drive. #twakedrive #opensource #cloud #storage #linagora
🇫🇷 @mmaudet.bsky.social représente #LINAGORA à la matinée “IA générative & agentique souveraine” organisée par l’@IMA_Innovation et le CSF à la @caissedesdepots .
L’open source, clé d’une IA européenne, éthique et durable.
#IA #SouverainetéNumérique #Innovation
🚀 Twake évolue ! Nouvelle interface, navigation unifiée, interopérabilité native : notre suite collaborative devient plus fluide, plus intelligente… et toujours 100% open source 🇫🇷
Une refonte technologique au service du travail souverain.
#Twake #LINAGORA
screenshot of the website written text: "LUCIE — The truly open source AI built on transparency, trust, and efficiency. Lucie the instruct model is no longer available but we work hard to put back online, if you want to be informed when the model will be available, please sign up" On the right, an AI illustration of a blond woman dressed with France's colors (blue, red, white)
pop up in the website that says: "Information notice Lucie is a generative large language model (LLM), meaning that given an input text or "prompt", it will generate a response related to this input. LLMs have to go through numerous phases of training before they are able to respond appropriately and accurately to diverse prompts, and Lucie is at a very early stage in its training, having received just one round of what is known as instruction fine-tuning. This training, while lightweight, gives Lucie the ability to follow basic instructions, such as answering a question or summarizing a text. Its responses may not be reliable yet, but without this additional training, Lucie, when given a prompt, would only be able to generate strings of words that are semantically related to the prompt. Its responses would not have the correct form. In order to move to the next stages of training, we need to collect large amounts of data that come from interactions with human users. In particular, we need feedback on whether an answer provided by Lucie is good or bad or preferable to another. If we get enough high-quality data of this sort, we can use it to teach Lucie to align its responses with the preferences of human users. ... (not enough espace for alt-text, check website)
LUCIE 🇫🇷 : A French Open-Source Language Model developed by Linagora and the OpenLLM France consortium.
LUCIE is a open-source language model aiming to provide a transparent and sovereign AI solution for Europe. Led by #Linagora and supported by various […]
[Original post on social.vivaldi.net]
Erreurs, absences de modération : Lucie, le chatbot de Linagora, rapidement fermé - Next #intelligenceartificielle #lucie #llms #generativeai #linagora #frenchtech
#LucieIA #IAfrançaise #Linagora #CNRS #innovation #technologie #souveraineté #LucieIA #technologie #France2030 #souveraineté #intelligenceartificielle #DeepSeek révolutionne l’IA et défie #ChatGPT presse83.fr/2025/28/1630...
Lucie hallucine et raconte n'importe quoi sur une question simple d'addition de deux nombres décimaux
l'IA me dit que 0,9+3,45 est égal à 3,95 à la fin d'une explication délirante.
#Lucie #IA #opensource de #Linagora, financée par #France2030 lance son premier modèle public... Y a du boulot. J'aimerais y croire mais le premier test est effarant. C'était peut être pas urgent de le rendre accessible 🙄. Je réessaierai dans un mois ou deux. #Asuivre
#maths #education
