So LLM-based AI is apparently such a dead end as far as non-spam and non-party trick use cases are concerned that they are straight up rolling out anti-features that nobody asked or wanted just to convince shareholders that ground breaking stuff is still going on, and somewhat justify the ocean of money they are diverting that way.
At least it’s only supposed to work on PCs that incorporate so-called neural processor units, which if I understand correctly is going to be its own thing under a Windows PC branding.
Now they take one with an NPU and rebrand it as Copilot+™️®️ PCs
They have market research that says initial sales are going to be soft and they panic because early soft sales create a bad vibe
So, without doing any of the usual build up of exciting the tech press, without hyping trade show buzz, they rush an unfinished, insecure, unwanted product to market in the hope it will be the killer app at last for high-battery life ARM on Windows.
They use lot of AI hype language to capitalize off the hype cycle, even though besides the OCR it seems to be pretty limited in its relationship to anything machine learning at all.
This might be the case, but at least the first I heard about the Copilot+ launch was that it was finally a “Macbook Air” killer - which I suspect would already be a strong selling point (at least if MSFT solved the backwards compatibility issue). Yet right after they announced the Recall stuff, and at least from what I have read it was received very negatively. So now they have the story that if you want the latest fast, efficient windows machine, you need to aloow it to spy on your screen. Not the best marketing imo.
OK, I didn’t know that (mslty because I don’t follow PC news, it’s aggressively boring). FWIW the only tech podcast I do follow (all mac people) did single out this release as “this time MSFT proabbly got it right” - but they’re mostly interested in Apple getting some competition.
I’ve seen that take somewhere too, and personally I don’t think it holds water. I’m a mac/*nix user and the hardware is only part of that. not fucking using windows is a hell of a massive part of my choices. and there’s basically nothing that they can do there to make a real dent
there’s definitely a lot of people out there that have web-heavy/-only workflows and on paper that group could move over, but in reality fucking windows is still fucking windows and the related problems that have plagued it[0] for years won’t just magically evaporate because of switching to a new arch
nevermind all the other crazy shit they’ve been pushing lately
[0] - think stuff like cruft buildup, spy-/track-/push-ware, etc etc
Yeah, I’m a windows user and agree with you completely. People choose operating system, not battery life.
I would love if they solved the problems that made windows on ARM not ready for prime time, even though I’m enough of a power user it will probably never be for me. But this is not the way.
Part of this is still trying to make a combination full featured windows laptop that’s also a Chromebook equivalent that’s also a tablet that’s also a dessert topping, when those should be separate devices with different ecosystems. UWP Metro apps were tablet-first when they first launched, sucking on desktop. The tablet pushing in Windows 10 initially broke accessibility. 2-in-1 Surfaces are way too heavy to be good tablets, because they’re still full featured PCs.
I do not want to mix this duck sauce with that chocolate bunny.
@gnomicutterance
I kind of wonder if part of this isn’t literally ‘so we can siphon up all the things you type for more training data’, and literally everything else is just to walk around that.
So just a straight leverage to get everything everyone types!
Besides the ocr there appears to be all sorts of image-to-text metadata recorded, the nadella demo had the journalist supposedly doing a search and getting results with terms that were neither typed at the time nor appearing in the stored screenshots.
Also, I thought they might be doing something image-to-text-to-image-again related (which - I read somewhere - was what bing copilot did when you asked it to edit an image) to save space, instead of storing eleventy billion multimonitor screenshots forever.
So LLM-based AI is apparently such a dead end as far as non-spam and non-party trick use cases are concerned that they are straight up rolling out anti-features that nobody asked or wanted just to convince shareholders that ground breaking stuff is still going on, and somewhat justify the ocean of money they are diverting that way.
At least it’s only supposed to work on PCs that incorporate so-called neural processor units, which if I understand correctly is going to be its own thing under a Windows PC branding.
Since this isn’t really even related to GenAI at all, in our house the theory is:
This might be the case, but at least the first I heard about the Copilot+ launch was that it was finally a “Macbook Air” killer - which I suspect would already be a strong selling point (at least if MSFT solved the backwards compatibility issue). Yet right after they announced the Recall stuff, and at least from what I have read it was received very negatively. So now they have the story that if you want the latest fast, efficient windows machine, you need to aloow it to spy on your screen. Not the best marketing imo.
Yeah but they’ve been marketing Windows on ARM as a Macbook Air killer for a few years now. This is more of a rebrand of that effort.
OK, I didn’t know that (mslty because I don’t follow PC news, it’s aggressively boring). FWIW the only tech podcast I do follow (all mac people) did single out this release as “this time MSFT proabbly got it right” - but they’re mostly interested in Apple getting some competition.
@gerikson Heh. Am waiting for the next Apple ads—“It’s a bit like Microsoft’s ARM laptops, only better and there’s no spyware.”
I’ve seen that take somewhere too, and personally I don’t think it holds water. I’m a mac/*nix user and the hardware is only part of that. not fucking using windows is a hell of a massive part of my choices. and there’s basically nothing that they can do there to make a real dent
there’s definitely a lot of people out there that have web-heavy/-only workflows and on paper that group could move over, but in reality fucking windows is still fucking windows and the related problems that have plagued it[0] for years won’t just magically evaporate because of switching to a new arch
nevermind all the other crazy shit they’ve been pushing lately
[0] - think stuff like cruft buildup, spy-/track-/push-ware, etc etc
Yeah, I’m a windows user and agree with you completely. People choose operating system, not battery life.
I would love if they solved the problems that made windows on ARM not ready for prime time, even though I’m enough of a power user it will probably never be for me. But this is not the way.
Part of this is still trying to make a combination full featured windows laptop that’s also a Chromebook equivalent that’s also a tablet that’s also a dessert topping, when those should be separate devices with different ecosystems. UWP Metro apps were tablet-first when they first launched, sucking on desktop. The tablet pushing in Windows 10 initially broke accessibility. 2-in-1 Surfaces are way too heavy to be good tablets, because they’re still full featured PCs.
I do not want to mix this duck sauce with that chocolate bunny.
@gnomicutterance
I kind of wonder if part of this isn’t literally ‘so we can siphon up all the things you type for more training data’, and literally everything else is just to walk around that.
So just a straight leverage to get everything everyone types!
It’s sure an awful way to do it…
Besides the ocr there appears to be all sorts of image-to-text metadata recorded, the nadella demo had the journalist supposedly doing a search and getting results with terms that were neither typed at the time nor appearing in the stored screenshots.
Also, I thought they might be doing something image-to-text-to-image-again related (which - I read somewhere - was what bing copilot did when you asked it to edit an image) to save space, instead of storing eleventy billion multimonitor screenshots forever.
edit - in the demo the results included screens.
The NPU requirement has already been proven to be circumventable, as well as a number of other problems found: https://awful.systems/comment/3492449
That you can jailbreak recall and run it on non compliant hardware seems to be the least concerning thing in that article, recommended reading.