WRONG, this completely ignores the most important issue and so is completely wrong.
The important issue is where is the data stored. And there are far to many advantages to having your data in the cloud: you can access it from whatever device you happen to have, and it isn't lost if you lose the device. This also outsources your backups to the cloud which is probably doing a much better job than you would (maybe no on hacker news, but nearly everyone else) - the cloud has earned a bad reputation for backups, but it is still much better than most people would be.
Once you accept the data is going to be elsewhere it doesn't matter if the compute is elsewhere or not. The data is the important part.
What needs to be the norm is more self-hosting your own data. Companies should not be outsourcing this by default - even where you outsource some of it, you need to watch your contracts and ensure the ownership is yours - not shared. Once your data is yours on your own cloud accessible servers we can start asking can we run our AI models in the same data center as we already have our data in. I don't need my AI model to run on my phone, it can run on the server in my basement which has a lot more power available (my phone has a better GPU but I can't afford the battery power to run AI on my phone)
> What needs to be the norm is more self-hosting your own data
I assumed self-hosted AI would fall under local AI for the purposes of this article. Does the author really need to spell it out?
I want a way to backup my data fully encrypted somewhere and have custody of the keys - but importantly, the data should all be decrypted locally where all my apps can use the data without any network