@MrWhosetheboss video reveals Google's Pixel 8 Pro Tensor G3 off-loads all generative AI tasks to the cloud
https://lemmy.world/post/7145579

@MrWhosetheboss video reveals Google's Pixel 8 Pro Tensor G3 off-loads all generative AI tasks to the cloud - Lemmy.World
So no more pixel phones for me.
So “no more pixel phones for you” because it needs to upload shit that literally cant be done on a phone? LOL. If you cared abkut privacy to the level youre pretending you do, that ONLY leaves Pixels as a choice, and Graphene.
Not a privacy conscious person alive buys a Pixel, and then runs it stock.
I think you have no clue what the g3 is capable of.
Magic eraser for example can be used entirely on device. THAT’S what even Google admits.
So now tell me why it needs to upload the pictures…just for computing? 🤡
Well, the Magic Eraser is just an algorithmic fill. It’s not an actual generative ai. So it doesn’t require so much compute power. If Pixel 8 had the new generative Magic Eraser on device, it would be either really bad or take minutes to process.
It is generative ai via ml. What are you talking about…
It’s obvious that you don’t understand this topic at all. So it would be better to end this discussion. But I will try to explain.
Old Magic Eraser: definitely not generative ai, because it doesn’t “generate” anything - just completes the missing part with the other parts of the image. So it’s not that resource heavy and can be processed on device. It doesn’t require complex diffusion generative models.
New Magic Eraser: uses actual generative ai, because it generates whole new objects that replace the part you want to erase. And that requires a lot of processing, because it needs large image generation models that currently can’t be run on the device.
Although, Pixel 8 could run some small image generation models, they would be far worse than the large models that Google runs on their servers. In the future, the technology will get better and more lightweight. So in a few years, it will be possible to run a good image gen models on device. But that’s not the current focus of the industry.