Meta Platforms, the parent company of Facebook, introduced an AI model named SeamlessM4T. It is capable of translating and transcribing speech in multiple languages, potentially enabling real-time communication across language barriers, according to an Aug. 22 report on Reuters.
"CEO Mark Zuckerberg has said he envisions such tools facilitating interactions between users from around the globe in the metaverse, the set of interconnected virtual worlds on which he is betting the company's future," Reuters said.
The model supports translations between text and speech in nearly 100 languages and full speech-to-speech translation for 35 languages. This development aligns with Meta's vision for fostering interactions in the metaverse, and the model is being made available to the public for non-commercial use, Reuters reported.
The publicly available model, offered by Meta for non-commercial use, marks a part of the prominent social media corporation's strategy to introduce an array of AI models this year, many of which are free. One significant release, the Llama language model, poses a substantial challenge to proprietary models from Microsoft-backed OpenAI and Google owned by Alphabet, Reuters said.
Zuckerberg highlights an open AI ecosystem serving Meta's interests, as the company benefits more from engaging the public to contribute to the creation of user-oriented tools for its social platforms rather than charging for model access, according to Reuters.
Nevertheless, like its counterparts in the industry, Meta grapples with legal questions surrounding the training data employed to develop its AI models. In July, comedian Sarah Silverman and two other authors filed copyright infringement lawsuits against both Meta and OpenAI, alleging the unauthorized utilization of their books as training data, Reuters reported.
For the SeamlessM4T model, Meta researchers disclosed in a research paper that they harnessed audio training data from a collection of "raw audio" sourced from a publicly accessible repository of web data, without specifying the exact source, Reuters said.
