Comments by "Gilad Barlev" (@GSBarlev) on "Solving Google's Keyboard Privacy Issue: Introducing Offline voice to text u0026 keyboard from FUTO" video.

  1. This is super cool, and even cooler that it can run locally. And even cooler that it's not using one of the large, general-purpose transformers but a svelte 32MM-parameter model by default. I'm going to point to this every time someone comes at me that we need a 42B parameter GPU-accelerated LLM for some simple, narrow use-case.
    2