Google To Create A Worldwide AI Language Model For 1,000 Languages

Google is planning to develop an advanced AI language model which will support 1,000 common languages worldwide.

Google has disclosed an ambitious new endeavor to create a single language artificial intelligence model that supports 1,000 of the most widely spoken languages in the world. 

It demonstrated an AI training algorithm on more than 400 languages as a first step toward achieving this aim, according to The Verge, which calls it “the broadest language scope seen in language models to date.”

Although language and AI have likely always been at the core of Google’s products, recent developments in machine learning, such as the creation of potent, feature-rich “large language models” (LLMs), have given these fields a new level of importance.

While brushing off concerns about the systems’ usability, Google has already started integrating these language models into services like Google Search. 

Language models have a variety of flaws, including the inability to analyze language with human empathy and a vulnerability to detrimental cultural prejudices like racism and xenophobia.

According to Zubin Ghahramani, Google’s vice president of AI research, creating a model of this size will make it simpler to bring different AI functionalities to languages that are underreported in virtual platforms and AI training datasets (also known as “low-resource languages”), he said in an interview with The Verge.

Such a technique has been proven beneficial in other studies, and the size of Google’s proposed model may provide important improvements over earlier research. 

Such expansive initiatives have come to be expected from technology firms vying for control of AI development. 

The present effort by Meta, the parent company of Facebook, to develop a “universal language translator,” is a similar initiative.

Access to data, meanwhile, is a problem when learning several languages, and Google says it will operate on the 1,000-language model by funding data collecting for low-resource languages, particularly audio recordings and written texts.

The business only anticipates the model to be extensively used across a variety of Google services, from Google Translate to YouTube subtitles, and does not directly plan to apply the capabilities of the model.