The company introduced an artificial intelligence that supports more than 400 languages ​​as the first step of this project. This model was described by Google as "the widest language coverage currently seen in a speech model". Language and artificial intelligence have been at the center of the US tech giant's work for some time. Recent advances in machine learning have also led to such significant developments in multifunctional large language models and the emergence of new products.

Language models; It can be used for many different things in search engines, such as creating language-based content and translation. Google's new initiative of 1000 languages, on the other hand, is intended to be used for different functions with a single system that has a wide knowledge about world languages ​​rather than a single function.

Speaking to The Verge, Zoubin Ghahramani from Google's artificial intelligence division also states that the company believes that such a large model can provide various artificial intelligence functions for languages ​​that are not common enough in online environments. Ghahramani makes the following statements regarding this:

“A model exposed to and trained on many different languages ​​allows us to achieve better performance in the under-represented languages ​​(low-resource language) we're talking about. The way to reach 1000 languages ​​is not to create 1000 different models. Languages ​​are like organisms; They evolved from each other and therefore show similarity. When we incorporate data from a new language into our model, we can translate from a common language to low-resource languages ​​and make significant improvements.”

The company adds that it will fund hard-to-access data such as written text and audio recordings required for low-resource languages. It should be added that Google has no direct plans for where it will implement the functionality of this model. For now, the tech giant expects it to be used in various Google products such as Google Translate and YouTube subtitles. Ghahramani adds that such wide language models can perform many different tasks such as translating codes for robots into commands, solving mathematical problems, and translating from language to language, so they have great potential.

A similar project of this kind also came from another technology giant, Meta. The company, led by Mark Zuckerberg, announced earlier this year that it is working on an artificial intelligence project, described as a 'universal speech translator', that can break the speech barrier both online and physically.