Though I praise the ethical stand point that Apple has taken, doing so will ensure their loss. The data that these companies will provide will be infinitely smaller than the huge mountains of data that companies such as OpenAI and Google have.
Also, having to rely on third parties will only slow them down. There are claims that regulations will be passed which may hinder companies such as Google which is scraping the internet for data, I assure you nothing like that will happen in the states. Currently, the world is in a race and no country can afford to be behind, no government (barring EU, duh) is foolish enough to hinder this fledgling field.
Also, Apple is a hardware company unlike Google which is a true software company, they have more areas to integrate and monetise AI.
So in short Apple should steal it? Apple can't win either way they go? By the way the quality of the data is more important than the quantity of data. Building anything in AI is going to require brains, focus, iteration and discipline.
What good is artificially handicapping yourself, when at this point we can't even say if it's legal or not. Apple is only hurting itself.
Google has decades worth of data and their deepmind division is the leader in AI research. The only reason they were behind OpenAI is because they were caught by surprise and are quickly catching up.
Is that why everything on the Pixel 8 Pro has to phone home to elicit a response because of the deep mind? Phoning home to a tensor (me too) processor that is weaker than the 11 Pro iPhone with a bad modem to boot isn't progress...
Some things do, most things don't. Smartphones aren't yet prepared to host the most resource-hogging features fully, and even then many of the Pixel 8 Pro features can't be done by Apple's iPhone at all, cloud servers or not.
Magic Eraser is done right on the Pixel handset. Best Take works on device, no internet connection is required. Audio Magic Eraser is handled on the device too. Gboard Smart Replies happen directly on the Pixel 8, as will AI summaries in the Google Recorder app, and Zoom Enhance (better than expected!). Call screening is also on-device. The latest Google Nano Generative AI also runs on-device.
Three features require the cloud: AI Wallpaper, Magic Editor, and Video Boost.
The Tensor just sucks in comparison to Apple and Qualcomm SOC'S
Two companies with long histories of chip development, one far longer than the other of course, are better at it than a company who first released their own design Tensor chip less than three years ago? What a shock, right? Well of course it "sucks in comparison".
Going back to Qualcomm would be a huge mistake since it would put them back at Qualcomm's mercy for supporting firmware updates longer than three years.
Apple had to start someplace with their own chip designs, and it wasn't at the top. Give 'em time to walk first.
The actual issue is with the foundry (Samsung's foundry) where the SoCs are manufactured. Qualcomm had the exact same issue for SD 888/888+ SoCs in 2021 and SD Gen 1 in first half of 2022. But then they ditched Samsung's foundry and switched to TSMC's foundry for fabbing their recent flagship SoCs SD 8+ Gen 1, SD 8 Gen 2 and SD 8 Gen 3. Why Google didn't switch the fab to TSMC at least this year (after 2 years of missteps with Samsung's foundry) is a million dollar question that only Google's management can answer.
Google is switching over to TSMC for future Tensor processors in Pixel handsets. As I recall it was originally planned for next year, but will probably be 2025 instead to allow for more tweaking.
EDIT: What is meant by "salvage the reputation of Pixel phones"? It looks to me as if they are getting more respect each year. The latest Pixel 8 Pro has even been crowned Phone of the Year by the same high-profile reviewer who tagged the iPhone Pro 15 as having Camera of the Year. They're also the only phone brand building US market share in a declining market, doubling it in 12 months, so isn't that evidence of being on the right track?
If Google is switching to TSMC only in 2025, I think it will be too late. Many of the Pixel phone users will start moving to other OEMs in the next 2 years. By the time Google get back to TSMC fab, Pixel phones would have started losing marketshare which is already very small. And Pixel phones would become unsustainable for Google to continue further. The sooner they ditch Samsung fab and start using TSMC for Tensor SoCs (or switch to Qualcomm/Mediatek SoCs), the better the chances of survival for Pixel lineup. Else, they would join the likes of LG, HTC, Sony pretty soon.
I don't think you've been keeping up with news, reviews and market reports, which would go towards explaining why your outlook is so very... umm... different. I'd suggest getting out and reading more on sites besides Apple-centric ones. Pixel phones aren't going away.
You are right the Google Pixel Pro's like the Microsoft Surface isn't going anywhere fast, the current Tensor SOC is barely ahead of Apple iPhone 10 Pro in performance and 2025 is too late.
I would also would suggest they (Google) add LiDAR (as implemented) by Apple to their smartphones too, that is if the Tensor SOC can handle it on the device. In addition Google will to need to build a R1 type co-processor chip too by 2025 and the same applies to Samsung, Meta and Qualcomm.....Oh that's right they're going to need wait until Apple releases the Apple Vision Pro before they can copy it.
You seem to be judging the Google Tensor G3 purely on simple benchmarks... better to instead view it in terms of the advanced features it provides (especially the ML features).
LiDAR could be useful, but is far from essential (and a Tensor chip could handle the data on device). What do you think the R1 chip does exactly? My understanding is it is responsible for processing information from the Vision Pro sensors.... why would a Google Pixel phone need to process information from headset sensors?
Also, if Apple makes a deal with the publishers, those publishers might want to see money from competitors like Google and OpenAI to let them keep access.
Though I praise the ethical stand point that Apple has taken, doing so will ensure their loss. The data that these companies will provide will be infinitely smaller than the huge mountains of data that companies such as OpenAI and Google have.
Also, having to rely on third parties will only slow them down. There are claims that regulations will be passed which may hinder companies such as Google which is scraping the internet for data, I assure you nothing like that will happen in the states. Currently, the world is in a race and no country can afford to be behind, no government (barring EU, duh) is foolish enough to hinder this fledgling field.
Also, Apple is a hardware company unlike Google which is a true software company, they have more areas to integrate and monetise AI.
So in short Apple should steal it? Apple can't win either way they go? By the way the quality of the data is more important than the quantity of data. Building anything in AI is going to require brains, focus, iteration and discipline.
What good is artificially handicapping yourself, when at this point we can't even say if it's legal or not. Apple is only hurting itself.
Google has decades worth of data and their deepmind division is the leader in AI research. The only reason they were behind OpenAI is because they were caught by surprise and are quickly catching up.
Is that why everything on the Pixel 8 Pro has to phone home to elicit a response because of the deep mind? Phoning home to a tensor (me too) processor that is weaker than the 11 Pro iPhone with a bad modem to boot isn't progress...
Some things do, most things don't. Smartphones aren't yet prepared to host the most resource-hogging features fully, and even then many of the Pixel 8 Pro features can't be done by Apple's iPhone at all, cloud servers or not.
Magic Eraser is done right on the Pixel handset. Best Take works on device, no internet connection is required. Audio Magic Eraser is handled on the device too. Gboard Smart Replies happen directly on the Pixel 8, as will AI summaries in the Google Recorder app, and Zoom Enhance (better than expected!). Call screening is also on-device. The latest Google Nano Generative AI also runs on-device.
Three features require the cloud: AI Wallpaper, Magic Editor, and Video Boost.
The Tensor just sucks in comparison to Apple and Qualcomm SOC'S
Two companies with long histories of chip development, one far longer than the other of course, are better at it than a company who first released their own design Tensor chip less than three years ago? What a shock, right? Well of course it "sucks in comparison".
Going back to Qualcomm would be a huge mistake since it would put them back at Qualcomm's mercy for supporting firmware updates longer than three years.
Apple had to start someplace with their own chip designs, and it wasn't at the top. Give 'em time to walk first.
The actual issue is with the foundry (Samsung's foundry) where the SoCs are manufactured. Qualcomm had the exact same issue for SD 888/888+ SoCs in 2021 and SD Gen 1 in first half of 2022. But then they ditched Samsung's foundry and switched to TSMC's foundry for fabbing their recent flagship SoCs SD 8+ Gen 1, SD 8 Gen 2 and SD 8 Gen 3. Why Google didn't switch the fab to TSMC at least this year (after 2 years of missteps with Samsung's foundry) is a million dollar question that only Google's management can answer.
Google is switching over to TSMC for future Tensor processors in Pixel handsets. As I recall it was originally planned for next year, but will probably be 2025 instead to allow for more tweaking.
EDIT: What is meant by "salvage the reputation of Pixel phones"? It looks to me as if they are getting more respect each year. The latest Pixel 8 Pro has even been crowned Phone of the Year by the same high-profile reviewer who tagged the iPhone Pro 15 as having Camera of the Year. They're also the only phone brand building US market share in a declining market, doubling it in 12 months, so isn't that evidence of being on the right track?
If Google is switching to TSMC only in 2025, I think it will be too late. Many of the Pixel phone users will start moving to other OEMs in the next 2 years. By the time Google get back to TSMC fab, Pixel phones would have started losing marketshare which is already very small. And Pixel phones would become unsustainable for Google to continue further. The sooner they ditch Samsung fab and start using TSMC for Tensor SoCs (or switch to Qualcomm/Mediatek SoCs), the better the chances of survival for Pixel lineup. Else, they would join the likes of LG, HTC, Sony pretty soon.
I don't think you've been keeping up with news, reviews and market reports, which would go towards explaining why your outlook is so very... umm... different. I'd suggest getting out and reading more on sites besides Apple-centric ones. Pixel phones aren't going away.
You are right the Google Pixel Pro's like the Microsoft Surface isn't going anywhere fast, the current Tensor SOC is barely ahead of Apple iPhone 10 Pro in performance and 2025 is too late.
I would also would suggest they (Google) add LiDAR (as implemented) by Apple to their smartphones too, that is if the Tensor SOC can handle it on the device. In addition Google will to need to build a R1 type co-processor chip too by 2025 and the same applies to Samsung, Meta and Qualcomm.....Oh that's right they're going to need wait until Apple releases the Apple Vision Pro before they can copy it.
You seem to be judging the Google Tensor G3 purely on simple benchmarks... better to instead view it in terms of the advanced features it provides (especially the ML features).
LiDAR could be useful, but is far from essential (and a Tensor chip could handle the data on device). What do you think the R1 chip does exactly? My understanding is it is responsible for processing information from the Vision Pro sensors.... why would a Google Pixel phone need to process information from headset sensors?
Google and Samsung being the me too companies they are in hardware/software will copy everything they can including the R1 chip, phoning home isn't a source of strength on a smartphone Google is doing it because their current processor simply is bad (five years behind bad) at performance/power usage and when paired with a bad modem to boot is the very definition of a not ready MeToo product.
And Google chose unwisely with their AR Core object implementation they choose to go cheap again because processor was not up to it?, since the iPhone 12 and the iPad Pro 4th generation Apple has included LiDAR (probably with an eye forward to the AVP thinking ahead).
Samsung also tried but they came with a half assed solution. Bad software and a bad sensor.
The only question left is whether Samsung will pull out a badly conceived non working mockup right before or right after the Apple Vision Pro event next year.
Google and Samsung being the me too companies they are in hardware/software will copy everything they can including the R1 chip, phoning home isn't a source of strength on a smartphone Google is doing it because their current processor simply is bad (five years behind bad) at performance/power usage and when paired with a bad modem to boot is the very definition of a not ready MeToo product.
And Google chose unwisely with their AR Core object implementation they choose to go cheap again because processor was not up to it?, since the iPhone 12 and the iPad Pro 4th generation Apple has included LiDAR (probably with an eye forward to the AVP thinking ahead).
Samsung also tried but they came with a half assed solution. Bad software and a bad sensor.
The only question left is whether Samsung will pull out a badly conceived non working mockup right before or right after the Apple Vision Pro event next year.
This article is about AI/ML. In what way is Google a me-too company compared to Apple in terms of AI/ML? Consider cloud infrastructure, TPUs, widely-adopted ML Frameworks (which are open-source), Google DeepMind, on-device ML features, Bard / Gemini.
Gatorguy informed you in Post 52 about the few Google Pixel ML features that need the power of Google Cloud Servers rather than being executed on-device. Key point was the vast majority can be done on-device today. The power of Google's software combined with its processors amazingly gives it an advantage over processors you consider five years ahead of it...
I am aware of LiDAR, it's on my M1 iPad Pro, though I never used it. I have no knowledge about Samsung's use of it. We'll see if Google deem it worthwhile to add it to ARCore / future Pixel phones or whether they'll stick to software-focussed solutions alongside their cameras.
So getting back on topic, reports say that the reluctance of publishers to sign deals on content comes down to Apple's insistence on including rights to use publishers articles and other content for any purpose they wish now and in the future. I perfectly understand why they would not want to accept Apple's terms just as much as I understand Apple wanting to get off relatively cheap, grab every right they can in one negotiation.
If the publishers are smart, they'll resist signing any contract that includes signing over all future rights. Things can change a LOT in two or three years..
Comments
And Google chose unwisely with their AR Core object implementation they choose to go cheap again because processor was not up to it?, since the iPhone 12 and the iPad Pro 4th generation Apple has included LiDAR (probably with an eye forward to the AVP thinking ahead).
Samsung also tried but they came with a half assed solution. Bad software and a bad sensor.
https://www.androidcentral.com/ipad-pro-has-lidar-camera-should-tech-come-android
https://www.techradar.com/news/6-ways-android-phones-could-use-the-iphone-12-pros-lidar-scanner-tech Google went cheap.
The only question left is whether Samsung will pull out a badly conceived non working mockup right before or right after the Apple Vision Pro event next year.
If the publishers are smart, they'll resist signing any contract that includes signing over all future rights. Things can change a LOT in two or three years..