Google Is Introducing Its Own Version of Apple's Private AI Cloud Compute

Google has unveiled Private AI Compute, a cloud platform designed to deliver advanced AI capabilities while preserving user privacy. As The Verge notes, the feature is "virtually identical to Apple's Private Cloud Compute." From the report: Many Google products run AI features like translation, audio summaries, and chatbot assistants, on-device, meaning data doesn't leave your phone, Chromebook, or whatever it is you're using. This isn't sustainable, Google says, as advancing AI tools need more reasoning and computational power than devices can supply. The compromise is to ship more difficult AI requests to a cloud platform, called Private AI Compute, which it describes as a "secure, fortified space" offering the same degree of security you'd expect from on-device processing. Sensitive data is available "only to you and no one else, not even Google."

Surprise!!

[registrations_suck]

Another AI product I don't want, don't need and didn't ask for.

The hits just keep coming!

Re:

[registrations_suck]

I don't. My working days are over.

That has its pros and cons. In my case, the cons are rather significant. I'd rather be working.

Trust

[Ender_Wiggin]

And just why should I trust this?

Re:

[h33t l4x0r]

Trust Google not spying you mean? Someone smarter than either of us (that we both trust) will need to audit the software I imagine.

Re:

[sabbede]

Why wouldn't you? It's not like they have ever mislead the public as to how their AI tools interact with user data. Except for when that just happened, but I'm sure it was a fluke.

So what they're saying is

[ddtmm]

We value your business so much, we're willing to spend billions on data centers and you get to use them free, and we're not going to harvest your data at all, no strings attached. Just the kind of goodwill we're devoted to.

Re:

[ZiggyZiggyZig]

Black Mage, are you Google's new PR Manager now?

Apple is actually ahead on

[Tablizer]

something in AI? That's surprising, they've been slow to jump on the LLM bandwagon. Maybe dragging their feet will look like the right move after the bubble pops.

Re:

[h33t l4x0r]

You realize that things don't disappear when bubbles pop, right? They just temporarily lose value. Dot coms are still here, housing is still here...

Re:

[Tablizer]

Often the direction changes after a bubble. If you built a fancy little e-store for your niche store in 1999, likely it would be rendered nearly useless in the face of Amazon a decade later.

Re:

[allo]

Amazon is not killing small stores because of a bubble, but because of being a large company with questionable business practices and too many users who like large websites. One also has to say that they are not bad with customers like allowing for uncomplicated returns.

Re:

[h33t l4x0r]

Worse case scenario here though is we have unused data centers. We'll grow into them eventually.

Like bitcoin? 3DTV?

[Somervillain]

Most would consider bitcoin a scam and 3DTV a dud. They're both still around, but didn't live up to the hype. What LLM companys promised and what they delivered were faaaaar different to the point they should be investigated for fraud. Housing and dot-com companies delivered value, just not profitably.

Pets.com and Kozmo.com offered legitimate, valuable services...just not profitably. They didn't lie about what they offered. Jensen Huang, Sam Altman, and all the others are outright lying about LLM capa

So, this is unsustainable?

[Cley Faye]

The thing that seems to be the most useful, small tasks over a limited subset of data, the thing that runs fine on device (and if the trend of providing the appropriate hardware keep going, will continue running fine locally), the thing that provides immediate, almost tangible benefit, with in most case easily and quickly verifiable output, while also not needing the computational power of a city block, THAT is what google deems "unsustainable" ?

And obviously, what seems to be sustainable to google is centr

Never trust a computer you don't own

[paramour]

'Never trust a computer you can't throw out a window.' - Steve Wozniak

corollary: Never trust anything connected to a computer you can't throw out a window.

Private?

[Ol Olsoc]

If it is not on your computer. If it is not air gapped. If it is on the cloud, or on the internet - it is not private.

Sounds more like a big honeypot to me. People - never put anything on the internet you wouldn't tell or show your mother. Kryste, how long are people going to buy into this canard?

I don't think so :o

[Mirnotoriety]

> Google has unveiled Private AI Compute, a cloud platform designed to deliver advanced AI capabilities while preserving user privacy

"only to you and no one else"

[Gilmoure]

[doctor evil skeptical face]

RIGGGGGHT.

That secure feeling.

[Ostracus]

If they're using the enclaves built into Intel and AMD, there may be side-channel issues to deal with. ARM is closer to what Apple is trying with their enclave.

Re:

[swillden]

If they're using the enclaves built into Intel and AMD, there may be side-channel issues to deal with. ARM is closer to what Apple is trying with their enclave.

ARM's TrustZone is definitely more secure than the alternatives on Intel/AMD, but TrustZone is also subject to side-channel attacks. To a first approximation, it's impossible to run two workloads on the same CPU and keep them perfectly isolated from one another.

However, I don't think any of these secure enclave concepts are relevant in this case. The way you'd build a private AI cloud is not to run it in enclaves (which are essentially just security-focused VMs) on CPUs that are running other tasks, the

Keep it local

[BcNexus]

Many Google products run AI features on-device. This isn't sustainable, Google says, as advancing AI tools need more reasoning and computational power than devices can supply.

I say that's BS and it's just a matter of will, greed, and control. Number one, models can be quantized, made in lower FP precision, and can even work in integer form instead of FP. Number two, models with many more parameters often don't produce results that are commensurate with the difference in the number of parameters between them and models with fewer parameters. But more importantly, I'm tired of the hegemony, gatekeeping and control of tech companies keeping AI service locked up in the cloud. After

It's cheaper to not privacy rape you with AI

[Somervillain]

If you think of it from Google's perspective, they want to provide you LLM-based services that adapt to your actual data....storing that on servers and mining it is costly. It is cheaper to do as much on device as possible. If it can't be done on device? It is tangibly cheaper to throw away the data. Sorry, not everything you do can be monetized and Google has so much experience privacy raping, they can be selective about what they want.

Also, they know so much about you from their various other offeri