Only two of your examples (parental controls and ___location sharing) require any kind of network, and those could be done with a private VPN running at home.
The design of cloud-based services is purely for convenience and collection. Sometimes if the collection can be controlled, the convenience is worth it, but every beneficial algorithm could be run locally.
The funny thing is that the rise of cloud computing coincides with the rise of really powerful cheap personal computation, you'd think it would be the opposite.
What "the cloud" offers that personal servers don't is:
- Reliability. The cloud is available with little or no downtime, to 5-8 nines (5 minutes to 1/3 of a second of downtime per year). Each nine costs roughly 10x the previous one.
- Bandwidth. Residential service may work for your own personal file transfer needs, but if you're sharing to the world, even a modest degree of traffic results in a hug-of-death.
- Security. Ideally, cloud systems are managed and monitored against network attacks, as well as affording physical security practices.
- Updates. This becomes Somebody Else's Problem.
- Ongoing development. Dittos.
It's not that these aren't addressible by individuals, but it's a lot of effort to do so, and at population levels, people are simply unlikely to be able or willing to do so. A small percentage, yes. The vast majority? No.
Raw compute power is a tiny fraction of the concerns involved in service hosting.
Smartphones are a massive leap forward in low-power/energy-efficiency, but my 2013 desktop machine (KGPE-D16) still creams every smartphone ever manufactured on any metric other than power consumption.
I'm kinda tired of hearing nontechnical people congratulate themselves on having a "supercomputer" in their pocket.
Compared to the supercomputers of the 1980s (if I remember right my dad had an old copy of Journal of the American Meteorological Society I found as a kid, with an ad for Cray talking about one gigaflop being state of the art), they're not wrong.
An example here for "run locally" would be Apple Health, which uses end-to-end encryption to sync data between your devices, does all the analytics stuff locally, and has a extensive permissions scheme for voluntary sharing of info with doctors or research programs.
Only two of your examples (parental controls and ___location sharing) require any kind of network, and those could be done with a private VPN running at home.
The design of cloud-based services is purely for convenience and collection. Sometimes if the collection can be controlled, the convenience is worth it, but every beneficial algorithm could be run locally.