Lancement du « Commun numérique des sciences en français » : 1,25 m. de documents pour l'IA, dont notamment des thèses et des articles, publiés dans la période 2007-2026 dans les bases OpenAlex, HAL et ThesesFR. tradso.hypotheses.org/931 @operaseu.bsky.social #Pleias @chaire-dcsf.bsky.social
Merci à #Pleias, #FranceDiplomatie, #WikimediaFrance et mes collègues de la direction des Archives, Bibliothèques et Collections muséales de #SorbonneUniversite pour cette journée.
Quelques informations supplémentaires sur la journée du 21 mars à Jussieu, dont la composition de la table-ronde dans ce communiqué de presse. www.sorbonne-universite.fr/actualites/s...
#Wikipedia #Pleias #TeamESR
> Today, we are announcing Amazon, Meta, Microsoft, Mistral AI, and Perplexity for the first time as they join our roster of partners, which includes Google, Ecosia, Nomic, Pleias, ProRata, and Reef Media. All these organizations utilize Wikimedia Enterprise to integrate human-governed knowledge […]
Really happy to see a new #copyleft -based #LLM , and this one seems to be more general-purpose than former attempts such as #PleIAs. The #Comma model is trained with #CommonPile, a new training pile with 8 TB of public domain and copyleft data. huggingface.co/papers/2506.052…
Really happy to see a new #copyleft -based #LLM , and this one seems to be more general-purpose than former attempts such as #PleIAs. The #Comma model is trained with #CommonPile, a new training pile with 8 TB of public domain and copyleft data. huggingface.co/papers/2506.052…
Ah, if only self-hosted AIs trained on copyleft data (like #PleIAs ) were the industry standard. I mean, the MCP could be used to make such models compatible with your private environment out of the box...
#IA Schéma des étapes de construction d'un modèle LLM Open Source (type #Pleias) parlezmoidia.fr/content/Z416... #AI
Je pensais (par expérience) que ce genre d'initiative était réservée aux grandes entreprises, @dorialexander.bsky.social, Alessandro Doria (son frère) et Pierre-Carl Langlais me prouvent le contraire avec #pleias
Good news: since a company's training requires one, I finally found a locally-hosted #LLM, #PleIAs, trained solely with freely redistributable data.
Bad news: it's so new, it hasn't been integrated with #LocalAI yet and I'm still tweaking YAML files around.
Good news: since a company's training requires one, I finally found a locally-hosted #LLM, #PleIAs, trained solely with freely redistributable data.
Bad news: it's so new, it hasn't been integrated with #LocalAI yet and I'm still tweaking YAML files around.
Ah, and I was about to download #PleIAs myself to test it. The AGPL share-alike restriction I don't mind, the problem is the non-commercial-licensed data would taint the license of the output. Any plans to filter the #CommonCorpus even further to prevent these issues? @dorialexander.bsky.social
Le français #Pleias lance des #LLM entraînés sur des données autorisées
👉Ses LLM reposent sur une approche utilisant exclusivement des données ouvertes conformes à la législation européenne
👉« Ils disaient que c'était impossible »
www.ictjournal.ch/news/2024-12...
Pour une #IAgénérative qui respecte le droit d'auteur
👉 #Pleias, une toute jeune start-up propose un corpus de textes tombés dans le domaine public pour entraîner des grands modèles de langage #LLM
www.lesechos.fr/idees-debats...