School
by School Editorial Staff
Creation of the Natural Language Processing (Nlp) research group of Sapienza University of Rome, within Fair (Future Artificial Intelligence Research) and the Leonardo supercomputer of Cineca
1′ reading
They are called Minerva and they are the new generative Artificial Intelligence models designed by a group of Italian researchers to train from scratch in our language with 500 billion words, the future Large Language Models like ChatGpt but entirely in Italian. They were created by the Natural Language Processing (Nlp) research group of the Sapienza University of Rome, within Fair – Future Artificial Intelligence Research, and the Leonardo supercomputer of Cineca. «The distinctive feature of the Minerva models – said Roberto Navigli, at the helm of Sapienza Nlp – is the fact that they were built and trained from scratch using open access texts, unlike the existing Italian models which are based on the adaptation of such as LLaMA and Mistral, whose training data are still unknown.”
Model
Each Minerva model has been trained on a vast set of online and documented Italian and English sources, for a total of over 500 billion words, the equivalent of over 5 million novels. Models that are now available to the scientific community of Fair, the project led by the National Research Council that implements the national strategy on AI, and in the future also to the public. A project that aims to provide transparent foundations for artificial intelligence systems that are applicable in different fields, from natural language understanding to text generation.
View on ilsole24ore.com
