3 ms·

Why don't they?

• BoorishBears 5y ago
  Don't they what, port their models to being on-device? I'm not Google, but just look at why they did it for voice: The hardware caught up, the UX is better when it's done locally, and I think they mentioned better power consumption? Presumably at some point that might happen for image recognition. They already started in Google Lens with porting OCR + translation onto the device