Comment by calgoo | Hacker News Reader

Comment by calgoo | Hacker News Reader

Parent Full thread

I wonder if MS will then just make it a online processing version, so it just ships all your data "for free!!!!" to MS servers and then processed remotely for you!

I also wonder how "knowledge transfer" will happen when you get a new machine in the future? What about backups, do they sit in the cloud already? These all sounds like ways that this "local" AI PC will share data with the MS cloud in one form or another. With Apple doing similar things already on iPhones (I know there are differences, but its still analyzing your data etc), I wonder if Linux might actually become a more mainstream OS in the future.

Could also be that people stop using computers as much and just use tablets with docking stations + keyboard & monitors (again, there is work to be done, but its a possibility from a HW level). That would leave us with MacOS & Windows for business desktops (with some Chromebook & Linux sprinkled in there). Education would probably be more Tablet & Chromebook style compute, and gaming is already moving to the cloud (I guess the positive here is that we might finally be able to get rid of AntiCheat software:) ).

> I wonder if MS will then just make it a online processing version, so it just ships all your data "for free!!!!" to MS servers and then processed remotely for you!

> I also wonder how "knowledge transfer" will happen when you get a new machine in the future? What about backups, do they sit in the cloud already?

Honestly, I guess this will become very expensive for Microsoft, and they won't find a good business case what to do with the collected data. So Microsoft is wasting a huge load of money, and additionally their AI spyware causes a huge reputation damage for Microsoft: a lose-lose situation. :-(

> they won't find a good business case what to do with the collected data.

ads.

they've been clear about windows becoming more ads oriented.

guess they are testing the water to try and get a competitive edge over google in a few years with this new data trove.

Other use cases for on-device NPU silicon have been demonstrated, e.g. radio sensing of room geometry and human motion, breathing, gestures.

Thanks for the clarification of other kinds of spying that this enables.

Nevetheless I still have difficulties seeing how this is supposed to make money for Microsoft.

> how this is supposed to make money

No idea about this specific feature, but billions are being spent to train models in the cloud (including Azure), with the expectation that some models will be used for on-device inference. It remains to be seen whether those investments will return dividends. In the meantime, cloud and GPU vendors are making money on model training.

This seems likely to me.

Considering the global market share of Windows, there would be a need to roll out such a grandiose service slowly, too.

Reply on news.ycombinator.com