The French LightOn launches the V2 of its open source LLM Alfred

In a recent interview, the CEO of OpenAI shared his ambitions in terms of fundraising and development. His ambition: to make AI as intelligent as humans. The giant, thanks to a personalized GPT function, is also tackling the creation of chatbots dedicated to specific tasks.

OpenAI thus encroaches on the field of open source models, such as Mistral 7B, the 1st generative AI model of Mistral AI, or those of LightOn, another hexagonal actor. The resistance is organized and LightOn announces a new version of his LLM Alfred.

Less contextual hallucinations

The French unveiled today, November 17 Alfred-40B-1023, the latest iteration of its open source language model. LightOn’s positioning consists in developing solutions dedicated to companies and public services.

The new model was trained on Amazon SageMaker. Alfred-40B-1023 will soon be available on HuggingFace and AWS Jumpstart, which he believes will facilitate its integration with other work environments.

For the managing director and co-founder of LightOn, Laurent Daudet, the continuous improvement of the ongoing LLMs “contributes to increasing the adoption of generative AI by strengthening the efficiency and the quality of the tasks that can be carried out.”

In this context, Alfred would therefore make his share of improvements. LightOn insists in particular on the reduction of contextual hallucinations, the main ailments suffered by generalist models such as those of OpenAI.

Taking care of the use case of documentary research

The French publisher also claims for its technology a better self-knowledge. More concretely, this means that the model, when it does not have a satisfactory answer, is programmed not to respond.

“I don’t know,“ Alfred V2 may thus indicate, ”which improves its transparency and reliability” for its designers. LightOn also takes care of a very fashionable use case in companies: research in documentary databases.

This feature called “Chat with docs” has been improved. The open source LLM is specially trained for these tasks consisting for a user to ask a question in natural language on the basis of internal documents.

“The model thus simplifies the interaction with the documents and optimizes the search for information”” says LightOn. Finally, the editor highlights an expanded context (8192 tokens), giving the possibility to the model to understand and generate longer and more complex contents.

More functions on Paradigm

Always to respond to the concerns of corporate clients, LightOn would also attach more importance to data security.

“When it is deployed in a client’s infrastructure (on premise or in a private cloud), all data interactions are confined to this environment, thus guaranteeing the confidentiality and security of the data,” he details.

But LightOn is a double approach in the field of generative AI: autonomous open source models and also a complete platform named Paradigm. By opting for the Alfred v2 version integrated into Paradigm, users have a “more advanced version”.

The publisher cites “improvements and additional features, ensuring that they always benefit from the most efficient capabilities.”Through its platform, LightOn also intends to simplify the use of generative AI.

“We have also focused on creating intuitive interfaces and workflows, so that everyone, regardless of their technical expertise, can harness the power of generative AI. Features such as “Task Factory” thus allow non-experts to integrate AI into their daily tasks in a transparent way,” says Laurent Daudet.