They're talking about us... Radiofrance
LUCIE, the 100% open source LLM developed with the OpenLLM France community, based exclusively on transparent data, as you know, has been the subject of heated debate in recent weeks.
However, it is also enjoying massive and enthusiastic support from members of the Tech community, who understand the ambition and challenges involved in developing and deploying AI technologies.
The media are reporting on these debates, giving LINAGORA the opportunity to clarify the background, the issues and, above all, the future of LUCIE.
A huge THANK YOU to everyone who, since Saturday 25 January, has shown us their support and contributed to this vital communication and education effort. You are incredible!
Selected excerpts:
- [...] This kind of incident with this kind of tool is fairly standard. A few months ago, we did a test together: we asked ChatGPT ‘how many ’n's are there in the word astonishment? It answered two, whereas there are three.
- This kind of error is not specific to a French tool, but rather to the technology of language models. We tend to overestimate their abilities because they write well and speak well. But we forget that they are incapable of reasoning. And if the answer to a question is not in their training data, they will invent one and ‘hallucinate’.