Meta Platforms, the parent company of (the company formerly known as) Facebook, on Tuesday disclosed advanced artificial intelligence model Which is characterized by the ability to efficiently translate and transcribe speech in many languages. This technology has the potential to facilitate communication between languages in real time, something sci-fi fans will certainly be excited to hear about.
According to the company’s official blog, the newly introduced “SeamlessM4T” AI model combines technologies to facilitate text-to-speech translation across nearly 100 languages and can perform full speech-to-speech translation for 35 languages.
“[SeamlessM4T model’s audio training data was derived from] Raw audio originating from a publicly available repository of crawled web data.”
meta engineer
Meta CEO Mark Zuckerberg envisions these tools playing a critical role in enhancing interactions between users from different parts of the world within the Metaverse and asserts that embracing an open AI ecosystem is a strategic advantage for Meta. The company will gain more by harnessing the collective efforts of the community to craft user-centric tools for its social platforms, rather than following a model that charges for access to these resources.
In an effort to encourage widespread use, Meta has made the SeamlessM4T model publicly available for non-commercial purposes, continuing the company’s trend of releasing mostly free AI models throughout this year.
Artificial intelligence has raised new legal questions
However, Meta, like other players in the industry, is not immune from legal inquiries regarding the source data used to train its models. in Noteworthy case of Julycomedian Sarah Silverman and two other authors have filed copyright infringement lawsuits against both Meta and OpenAI, alleging unauthorized use of their books as training materials.
As described in a research paper by researchers at Meta, the voice training data for the SeamlessM4T model was derived from a wide range of “raw audio originating from a publicly available repository of crawled web data”. Details of this repository are still undisclosed.
Regarding the textual data, the paper notes that it was obtained from datasets generated in the previous year, extracting content from Wikipedia and related online sources.
The latest Meta release of the SeamlessM4T AI model is an important step towards achieving seamless interlanguage communication. However, the company, like its competitors, will need to navigate the complexities of using training data within it The legal framework If she hopes to achieve long-term success.