While I'm a huge supporter of #linguistic diversity, I'm a little cold on Meta's Language Technology Partner Program. It aims to get language communities to speech training #data, for free, with the promise Meta will release trained models "open source".
https://about.fb.com/news/2025/02/announcing-language-technology-partner-program/
https://about.fb.com/news/2025/02/announcing-language-technology-partner-program/
Comments
1. It concentrates power in the hands of Meta - the data, when provided, is not being made openly available, which means that only Meta is able to benefit from the data.
Promising to "open source" models does not respect data sovereignty.
Will future generations have to pay to learn their ancestral language?
The bigger piece here is enabling people all over the world to speak with each other ...
as long as each interaction is mediated by Meta.