The companies will do anything to make having the ai capacities built locally an acceptable thing via “cool” features like this or apples, and I think it is because once the devices can do the processing locally, it allows them to stop processing it themselves on their servers. This will also allow them to use microphone and why not camera data as additional data points without having to send and process the actual microphone data. The only local software is open source AI implementations that are being used by FOSS applications, with no network access.

To me this isn’t about shareholder value and buzzwords, this is just the excuse to shove it into the OS. It’s a more long term game they are playing: the one of reducing their costs and improving the value/accuracy of the data that they get, since it will be pre-digested 100% locally in the background, which is not limited by network latency and bandwidth.