Artificial intelligence (AI) is now used in all aspects of daily life from translation systems, controlling cameras, identifying tumors and monitoring human movements. But the AI has a problem that it is trained using only the world's main languages. The long-term impact is that it will have trouble helping those who don't use these major languages.
Google has announced the 1000 Languages Initiative with the hope of developing an AI model that can understand 1000 languages. On paper 1000 languages is a large number but in human civilization there are now 7000 languages.
The model will be trained to support 1000 languages so that in the future it can not only be a universal translator but can also be used to recognize images, videos and voice commands in various languages.
So far Google has succeeded in producing a Universal Speech Model (USM) that supports 400 languages. It is the backbone of Google Translate's translation system now. This Google effort is similar to the Universal Translator project by Meta which has recently succeeded in translating the Hokkien language using only audio without a written system.