To enhance AI-driven language detection and translation, especially for under-represented languages, Meta has started a project in collaboration with UNESCO.
The program called the Language Technology Partner Program, is looking for partners who can provide written material, audio recordings, and translated sentences in several languages. The ultimate objective is to create open-source AI technologies that support language variety and improve speech detection and translation.
The International Decade of Indigenous Languages, a global initiative by UNESCO to preserve and revitalise endangered languages, is in line with this. The government of Nunavut, Canada, where many people speak Inuktut, a group of Inuit languages, has worked with Meta on this project. These languages will be incorporated into open-source machine translation and speech recognition models by Meta’s AI research teams in collaboration with partners.
According to Meta: “Our efforts are especially focused on underserved languages, in support of UNESCO’s work.”
Developing AI systems that can effectively communicate across linguistic and cultural divides is another goal of the company, it noted.
Aside from the Language Technology Partner Program, Meta has introduced an open-source machine translation benchmark to evaluate AI translation models. This benchmark is accessible through the AI development platform Hugging Face and was created in collaboration with linguists.
BrandSpur technology and information news desk report that translations can be made by developers and researchers to improve AI accuracy. The most recent initiative from Meta expands on earlier attempts at AI-driven language support.
With the launch of the No Language Left Behind (NLLB) project in 2022, the business developed an open-source machine translation model that greatly enhanced translations for numerous lesser-known languages. The Massively Multilingual Speech (MMS) project, started by Meta more recently, extended AI-driven audio transcription to more than 1,100 languages. With the addition of zero-shot speech recognition, this system can now transcribe languages it has never encountered.
However, to allow creators to dub their movies in many languages, Meta has also experimented with AI-powered language solutions for social media, such as automatic voice translation for Instagram Reels.
Technical workshops conducted by Meta’s research teams will be available to partners who sign up for the Language Technology Partner Program. Participants in these courses will learn how to create their own language technologies by utilising Meta’s open-source AI models. Advanced Machine Intelligence (AMI), or AI systems with human-like reasoning capabilities, is the focus of the company’s Fundamental AI Research (FAIR) team. The goal of this research is to create AI capable of carrying out intricate cognitive activities, such as producing accurate translations.
Despite these developments, Meta has received criticism for how it handles content that is not in English.
According to research, only 29% of COVID misinformation in English was reported, whereas nearly 70% of misinformation in Italian and Spanish was. Internal records that were stolen again revealed that posts in Arabic were routinely mistakenly labelled as hate speech. Meta has promised to enhance its translation and moderation tools in response to these concerns. The organisation aims to create more accurate and culturally inclusive AI systems by working with linguists and language preservation specialists.
Experts stress the ongoing significance of human monitoring despite the tremendous growth of AI translation. Human translators are still crucial in domains like international relations, law, and medicine because they can capture context, subtleties, and cultural importance that AI is still unable to fully mimic.
With its most recent project, Meta aims to increase AI’s accessibility and inclusivity on a worldwide scale. The company wants to close language gaps and increase digital accessibility for marginalised areas by making language models open-source and encouraging contributions from other linguistic communities. As AI-driven translation becomes more widespread, it may become more difficult to guarantee accuracy, equity, and ethical application.
