The analyst doesn’t think attempts to create sovereign clouds will succeed. He mentioned past French attempts to create sovereign clouds named “Andromeda”, “Numergy,” and “Gaia-X”, which he says went nowhere - but did produce some nice white papers.

The argument I would make here is that - even though France’s government always had an autonomous philosophy - the idea that relations between the US and France&Europe would become so sour wasn’t seen as realistic until very recent history.

So, any attempts to create an alternative when the current solutions were working just fine - and there weren’t any worries about US becoming antagonistic to European businesses & governments - would obviously be fighting an uphill battle.

But now the environment and context changed. The idea of needing an alternative is no longer seen as pointless or ludicrous. Several governments now want to have data sovereignty, unlike before. The new attempts will - I hope & expect - be more serious and committed than previous ones.

I’m not claiming that these new projects for data sovereignty are going to work out - I don’t know, I can’t predict the future - but I am saying that the context changed, so past attempts need to be seen through a different lens.

Yeah, it’s what I said. Right now it’s defaulting to the cloud

For clarity sake, that’s not what’s happening here. (Don’t misunderstand this comment as defending google, I could write a book about how much they suck)

The model downloaded is a LLM called Gemini Nano, and it’s used for things like “help me write”, checking if an incoming message is scam, summaries, etc.

Don’t worry about it itself being a spyware. It’s not; but for argument sake, if we were to assume that it was: they already know a lot about you through their usual apps and services, and get a lot more info out of you through them. This LLM would hardly move that needle.

The actual issue is that they download it for everyone, even if their devices don’t match the minimum requirements. And without consent. And to enable it, you need to go through several menus, as the default behaviour is to use the cloud (this could change eventually, my understanding is that in this update they’re just laying the foundation)

But, it’s Google that we’re talking about. Last year they were sentenced to pay a fine for spying on users despite them having their tracking settings off. And it wasn’t the first time iirc. This kind of behaviour is par for the course with them

I use arch btw

The idea sounds reasonable to me.

Of course, between idea and execution a lot can change. But as long they take some sane design decisions (opt-in, transparent, sandboxed, give the user freedom to use their own API or local models, etc), I’m fine with it

The Juicero was one clear red flag of this

https://en.wikipedia.org/wiki/Juicero

I imagine that the article mentions it

I could’ve told you that for free, no need for a study