This is only the beginning. At the moment, mobile devices must run sub-SOTA models, such as a MobileNetV3[a] for visual recognition, to provide an acceptable us

This is only the beginning. At the moment, mobile devices must run sub-SOTA models, such as a MobileNetV3[a] for visual recognition, to provide an acceptable user experience. As mobile devices continue to get more powerful and start incorporating dedicated AI hardware for multiply-add-accumulate ops, I expect to see devices that can run today's SOTA models locally within a couple of years. Exciting stuff! [a] https://ai.googleblog.com/2019/11/introducing-next-generation-on-device.html https://ai.googleblog.com/2019/11/introducing-next-generatio...