Pixel 8 handing bulk of AI work to Google’s servers as Tensor G3 not capable enough apparently،
In a world where so much is fake and pretentious, Google’s new Pixel 8 and 8 Pro give you the tools you need to stand out. While your friend with an Apple iPhone or Samsung Galaxy might be able to edit their photos, Google’s latest phones take it a step further by letting you tinker with things like capturing a group photo that doesn’t represent accurately the real moment. This may be impressive from a technological point of view, but it seems a little scary. But this is not the question.
Google’s claims about generative AI built into the Pixel 8 seem very impressive
This year was all about AI. But as cool as it sounds, this AI thing requires a lot of resources. For example, some experts estimate that it costs
$100,000 per day to run ChatGPT.
You may be wondering why ChatGPT is so expensive to operate. Well, that’s because it requires a lot of computing power. On top of that, they have to pay researchers and developers. And of course, there’s also the cost of data storage.
But this article is not about ChatGPT. It’s Google’s incredible Tensor G3, which the company says “brings the latest in AI to the world.”
Pixel 8 And
Pixel 8 Pro.
As Google notes in
a blog post, most AI innovations rely on processing power only available in data centers. But the
G3 tensor is different, depending on the company. It brings AI to the Pixel 8 and 8 Pro. The company acknowledges, however, that on-device generative AI is very complex.
But
Google somehow managed to bring it to the
Pixel 8, including the text-to-speech model used in Google data centers. Some of the features that have benefited from this are currently only available in the
United States.
Machine learning algorithms also improve cameras, allowing phones to do things like take a series of photos, select the best photo for each subject individually, and create a group photo (Best Shot). Google says it uses in-device algorithms to achieve this.
Another AI feature is Magic Editor, which lets you resize and reposition topics with “a few clicks.” There is also the Audio Magic Eraser feature which removes distracting sounds such as background noise from your video.
Most AI magic apparently happens away from your phone
Google may have overestimated the capabilities of G3 tensor, according to YouTuber Mrwhosetheboss. He noticed that every AI-based feature requires an internet connection. This is because it seems that the Pixel 8Google’s chip isn’t powerful (or smart) enough to perform all of the aforementioned tasks and must send them to Google’s servers to complete.
In the case of Magic Editor, an image must first be saved to your Google account before being processed.
This is not the picture painted by Google and not everyone would agree with sending their images to Google for processing. And even if you’re willing to accept that, Mrwhosetheboss says the process isn’t as seamless as Google has portrayed it to be, and clearing items and moving items takes a long time. Phones sometimes have trouble identifying objects and in some cases you get an error saying that Magic Editor cannot display the desired results because it might violate its terms.
Even with all the lags and delays, it’s obvious that
Google put a lot of work into these editing features, but most of the work seems to happen off-device. Mrwhosetheboss likened the experience to using an AI website on your phone.
The slowdowns also make it clear that the
G3 tensor isn’t as powerful as the chips that power other high-end 2023 phones. During our review, we found that the
iPhone 15 Pro Max and Samsung Galaxy S23 Ultra are significantly more powerful than the
Pixel 8 Proat least when it comes to benchmark scores.
Mrwhosetheboss compared it to the iPhone 12 and says tasks that haven’t been optimized seem slower than on a competing device. And while it’s not as fast as competing phones, the Pixel 8 doesn’t seem to be as effective as Google made it out to be and it doesn’t take much to make it heat up and slow down.
Hopefully phones will improve over time.
