Because it’s proprietary software. They have an open source model (based off it) called Gemma but Gemini Nano is super locked down. There aren’t even public APIs for 3rd party developers to use it through the OS yet.
None of those are cutting edge AI models that could be ripped open and examined if people had access to the files. It’s not just an app or something, there are internal trade secrets at risk.
Because Google is a monopolistic piece of shit and they try to lock you in to their shitty, privacy-invasive ecosystem. In my opinion it’s like a hundred times worse than Apple. Only Google hardware (phones and tablets) are worth buying, but only for the strong hardware security features, definitely not for the stupid proprietary software they come with by default.
Not to mention that even with the stock OS you can disable most if not all sniffing components. Of course Google still has root but they don’t put their typically don’t embed obfuscated stuff because they don’t need to.
Note that I’m not arguing that Google isn’t privacy invasive “because you can turn it all off.” The user shouldn’t have to go through this trouble to recover their privacy and the user experience definitively degrades if you do.
Definitely high. It’s pretty funny how people manufactured this privacy perception and projected it over Apple. Apple was happy to capitalize on it and keep it going. Reminds me of the security perception over BlackBerry. Hot air either way.
I mean Apple is certainly more private than stock Android of any flavor, but Google at least gives you the option to take a bunch of that privacy back.
Pixel Camera (previously known as Google Camera) can take full advantage of the available cameras and image processing hardware as it can on the stock OS and does not require GSF or sandboxed Google Play on GrapheneOS. Direct TPU and GXP access by Google apps including Pixel Camera is controlled by a toggle added by GrapheneOS and doesn’t provide them with any additional access to data. The toggle exists for attack surface reduction. Every app can use the TPU and GXP via standard APIs including the Android Neural Networks API and Camera2 API regardless.
TPUs and GXP are what enable apps to do on device ais with whatever model they choose to bring.
Lol. That’s the hardware. Of course it has access to the device hardware. You still need software. All of Google’s local AI features use Gemini Nano, which absolutely 100% I guarantee you will not ship with GrapheneOS.
Did you even look at that? Your own link disproves your claim. It’s just a general AI model that powers a variety of tasks, and is integrated into apps.
Does it receive all 16GB of RAM or is the 3GB still locked away?
Is Gemini Nano supported?
There isn’t a chance in hell you’re getting Gemini Nano outside of the stock ROM.
Why not?
Because it’s proprietary software. They have an open source model (based off it) called Gemma but Gemini Nano is super locked down. There aren’t even public APIs for 3rd party developers to use it through the OS yet.
An enormous portion of Google’s products are proprietary and yet still made freely available to anyone.
None of those are cutting edge AI models that could be ripped open and examined if people had access to the files. It’s not just an app or something, there are internal trade secrets at risk.
No one is asking for access to the source code.
That is basically what you get if you have direct access to the LLM file(s).
Then thousands of people already have that on their Pixel 9s and 8s.
Because Google is a monopolistic piece of shit and they try to lock you in to their shitty, privacy-invasive ecosystem. In my opinion it’s like a hundred times worse than Apple. Only Google hardware (phones and tablets) are worth buying, but only for the strong hardware security features, definitely not for the stupid proprietary software they come with by default.
LOL are you high?
Yea try to run an aftermarket OS on an iPhone and then we’ll talk
Not to mention that even with the stock OS you can disable most if not all sniffing components. Of course Google still has root but they don’t put their typically don’t embed obfuscated stuff because they don’t need to.
Note that I’m not arguing that Google isn’t privacy invasive “because you can turn it all off.” The user shouldn’t have to go through this trouble to recover their privacy and the user experience definitively degrades if you do.
Definitely high. It’s pretty funny how people manufactured this privacy perception and projected it over Apple. Apple was happy to capitalize on it and keep it going. Reminds me of the security perception over BlackBerry. Hot air either way.
I mean Apple is certainly more private than stock Android of any flavor, but Google at least gives you the option to take a bunch of that privacy back.
Who knows. We do know that all of the pixel photo features work assuming you install the pixel photo app and give it NPU permissions.
The exciting bit is that we know you can deny internet access and all the picture AI stuff still works.
Presumably the GOS devs. And anyone who owns a P9x.
Yeah, that’s why I ask. Some of this stuff works already. The on-device AI in Gboard is incredible. As is the spam detection in GMessages.
Source? There’s no way any of the offline stuff works.
https://grapheneos.org/usage#pixel-camera
So this says absolutely nothing about on device AI…
It does?
TPUs and GXP are what enable apps to do on device ais with whatever model they choose to bring.
Lol. That’s the hardware. Of course it has access to the device hardware. You still need software. All of Google’s local AI features use Gemini Nano, which absolutely 100% I guarantee you will not ship with GrapheneOS.
You’re wrong. They already ship with GOS.
Prove it. Show me Gemini Nano in GOS. I’ll wait…
Fuck Gemini. Why would you want that piece of shit on your phone?
Having recorded and summarized phone calls is a pretty damn useful feature…
Why would you not?
privacy
That’s why people here are discussing Gemini nano, which can run locally on your phone and not in the cloud (like the “normal” gemini model)
Exactly.
You might be able to get ollama running
Ollama doesn’t do anything that Gemini Nano does.
Ollama isn’t a model. It is a software that allows you to run llms and query them in layer 7
I didn’t say it was a model. I said it doesn’t even do what Gemini Nano does. Gemini Nano is not an LLM.
Gemini Nano is an LLM.
Did you even look at that? Your own link disproves your claim. It’s just a general AI model that powers a variety of tasks, and is integrated into apps.
What else would it be except an llm? What do you think model means?
…what do you think LLM means?
When are you going to admit you have no idea what you are talking about?
An LLM literally is a “general AI model that powers a variety of tasks”.
When are you going to admit you have no idea what you are talking about?
An LLM literally is not a “general AI model”, it’s a Large Language Model, as in it processes language.