On February 14th, Matthieu Lambda, the chief editor of Underscore_ approached us on Twitter. Using Cedille to create a mail bot that may not have given optimal responses...
Nabla, an innovative medical app that provides consultations through chat, fine-tuned Cedille for their own use case. They shared their findings with us through twitter. We were curious to see how and if large fine-tuned models could imitate doctors!
Today we released our German language model, that will be a game changing solution in the world of Machine Learning to German-speaking users and companies.
Some people wonder if Cedille is really open-source. Short answer: yes
The development of Cedille was made possible by the generous support of Google's TRC program, with which we were able to access a cluster of 1,000 cloud-based TPUs (Tensor Processing Units), including one large instance and a few dozen smaller instances.
We asked Cedille to write this article for us, without any editing on our part. While this is an interesting and creative article, some errors were made. We have included notes between paragraphs to point out some of these factual and logical errors.
It took three months to fix bugs and experiment with the model. We carried out several benchmarking tests and found for example that Cedille is better at translating to French and less toxic on average than GPT-3.
After two weeks of launch, we have had over 1000 accounts created and 90,000 generations of content. We took the time to review some of the feedback (positive and negative) since the launch of Cedille.
Our agency launches Cedille today, a new artificial intelligence for text generation that provides a game-changing solution for French-speaking users.