Google has announced an ambitious new project to develop a monolingual artificial intelligence (AI) model to support it. “The 1000 most common languages in the world”Verge reports.
As a first step towards this goal, the company offers an AI model trained in more than 400 languages. “The broadest language coverage that can be seen in a language model today”.
Google’s initiative does not focus on any specific functionality, but aims to create a single system with a large amount of information from all the languages of the world.
Zubin Ghahramani, vice president of AI research at Google, said in an interview with The Verge that the company believes building a model of this size will make it easier to implement various AI functions in languages that are underrepresented in the online space. and AI training databases.
“By having a model exposed and trained in many different languages, we achieve much higher performance in our low-resource languages”– says Gahramani. “The way to 1000 languages is not to create 1000 different models. Languages are like organisms, they evolved separately from each other and have some similarities”.