Gregory's Server@RicoElectrico Private machine translation models feels like something that was being handled by transformers and embeddings before everyone got excited about "GenAI".
And Common Voice is very important for people who need text-to-speech. But speech synthesis is even older than embeddings, with lots of non-GenAI techniques.
Strangely, neither of those are what Mozilla promotes. They've focussed on "what if the lying chatbot was in your browser!" and "let's do 'burn the planet' but ethically".
@RicoElectrico I've worked in the industry. To me, there's an important difference between Machine Learning and Neural Nets and all of the good, solid, practical, applied techniques that people were using for well scoped tasks to support users versus "what if we pretended that the advanced auto-complete based on stealing creative works was INTELLIGENT (while it lies) and then everyone anthropomorphised it and started pretending that the generated slop is actual art" that gets called "GenAI".