3 ms·
This I/O was all about On-Device ML. As ML chips get more powerful, a lot of these cool features that were only possible on the cloud will be moving to the devi
by ehsankia 8y ago
This I/O was all about On-Device ML. As ML chips get more powerful, a lot of these cool features that were only possible on the cloud will be moving to the device, lowering latency and working in poor network conditions. Google's use of Federated Learning is interesting too.
- AstralStorm 8y agoYet they will all call home tracking your every action in an aggregate. So Google and co. can plausibly deny having direct data - except they wrote the ML algorithm to be useful for the purposes they want so who knows how recoverable real data is.