If Microsoft wasn’t run by tools, they’d see the gap Google and Apple have left behind by locking down their eco systems.
They could be the hero we need by saying we’ll make the software and you fully own your device like pc / windows.
But of course they won’t, and will just shoot themselves in the dick.
Just like when they ditched explorer we were all like yaay! Then instead of attaching to Firefox they just became another chromium cuck.
Why would anyone take your shitty browser that’s just a skin of chrome…
Again, they had the chance to take the pro customer lane and succeed, but they were too inept.
You didn’t miss the “mobile wave”. You purposely gave up. Idiots.
True. As much as I hate to admit it, the Windows phones were actually pretty good.
Had they not botched app adoption and then immediately given up, they could have done fairly well.
The article touches on a bunch of valid points, but re the headline, I don’t really think that a failure to generate excitement about AI integration into Windows 11 is because they missed the boat. It’s because they’re shoehorning it into places it doesn’t belong.
They have the ability to make it useful. Ethical concerns aside, GitHub Copilot is as good as any AI development assistant, and better than most. Hopes that they’d gain ground with Bing would have needed them to be way ahead of the curve (and for AI search result summaries to be more useful than the top results, which they rarely are).
But for Copilot to be useful in the desktop environment, it needs to be there quietly in the places it’s needed. Improve your help tools, make Grammarly irrelevant, infer document context to make search better. Don’t rename half of your products “Copilot”, don’t put flashy buttons in every app, just use the benefits of applied AI to improve your products.
Oh, and make it optional, for fuck’s sake. If I don’t feel like I have control over my OS any more, I’m not likely to stick around when other options are available.
Conflating two issues there.
If they’re talking about developing ‘an AI’, they were late but are still in the game.
If they’re talking about adoption of AI, they shat the bed there. They slapped copilot on everything and tried to force us into the way they thought we should be working. They forced us to give up old hardware and to use their garbage new OS. All that was wrong and seriously turned a lot of people against MS. They forgot that ‘the customer is always right’. Again.
I still think there will be a big role for AI, but its in the background, quietly assisting in a way we barely notice. Big Tech will pay for how they’ve missold AI.
They didn’t miss the “wave”, they discovered it’s just hype and a bubble. They spent a fortune and damaged their core products to try and get in on AI, and have realised it was fools gold that their actual paying customers don’t want. This really sums the problem up well:
According to Velloso, less than 3% of paying users actively use Copilot, even though Microsoft has pre-deployed it directly into the Windows 11 taskbar and across the Office suite.
Out of Microsoft’s 450 million Microsoft 365 user base, the company has only managed to convert roughly 15 million paid Copilot seats. This means a staggering 96.7% of users are rejecting the premium AI features, yielding just a 3.3% paid adoption rate. When viewed against Microsoft’s estimated $37.5 billion quarterly AI spending, this is an alarmingly low adoption rate.
I’m sure I’m like many people - I tried Copilot a couple of times; it’s ok to make an email or even document text a bit more concise, but that’s really it. I don’t find it useful; I do all the actual work and then occasionally get an AI to help make it a bit easier to read very similar to a spell check and grammar check. It’s not good enough to do anything else; it bullshits and is error ridden and like all the AI I’ve tried it’s really plateaued. I just really don’t see where the value in that $37.5bn spent by Microsoft is.
I certainly wouldn’t pay for copilot myself. Instead I object to it being rammed down my throat at work, and Windows 11 just being generally awful but not improved. Microsoft are finally making the right noises but the damage is already done.
No one wants copilot because it’s highly unpleasant hot garbage. There is definitely a market for AI for the competent providers.
Yeah the vast majority of AI “offerings” from most of these huge companies and/or websites is just bolting a chatbot to something and then wondering why people don’t want it. I tried copilot in excel and it couldn’t access the document I was working on, it was an absolute useless mess.
It’s a wave of sewage, few users want to ride it in the first place
it’s a tsunami. uncontrollable, started far away from any normal humans, sweeps up everyone in its wake, and will cause massive damage when it inevitably crashes into a place with lots of people.
It’s like they created a very good phone tree and are trying to shove it into everything that never had or needed a phone tree in the first place.
Funny you mention a phone tree, something that’s been hit by AI. It’s actually been around longer as voice recognition that finds a close match to a keyword, but in theory AI should be able to take a request and break down what is actually needed.
I haven’t run across an AI version that works well. I don’t know if that’s because the voice recognition part is still bad, or if they’re using Co-pilot (since I know how it mangles simple requests in text).
Yeah it really feels like an LLM should work better than a phone tree for that, but every time I actually encounter one it’s so so much worse.
I wouldn’t be surprised if there’s instructions to modify the system prompt to maximize effectiveness, and everyone leaves it at the generic default. Just like so many people leave other things at the default and just plug it in and go. Thank goodness the Cisco hold music is decent. I grew to love it while holding on the VA phone lines a lot for my dad.
Shouldn’t the default settings work for the tasks that it’s advertised to? I mean when I buy software I don’t expect it to be set to “be shit” mode by default.
At least tpu can tell the AI to get you to a human and most of the time it actually does so.
Having voice recognition in place of the usual “press x” before AI was even worse. Bot that now it’s much better though.
The fuck is a phone tree? Pardon my language, sir.
“for financial services, press one. for technical support, press two. for goblins, press three. for repairs, press four.”
I TOLD YOU NOT TO TALK ABOUT GOBLINS
Why do you call it Troll 2 if there are only goblins in the movie?
A movie of such quality and esteem such as Troll 2 does not need to explain itself
you’re absolutely right! substituting pigeons.
You forgot the mandatory lead message. “Please listen closely as our menu options have changed.” No, they haven’t. Ever.
Thanks! I just thought that’s some non-existent thing, which Microsoft invented and nobody needs it ever. Well, that’s not far from your point, I guess. But still. Didn’t know that’s called a phone tree!
funny you should use that example in particular because i recently had the displeasure of using microsoft’s phone tree. i was trying to close a dead relative’s account and the info on the website was wrong.
they built a phone tree that remembers you. if you try to call in multiple times during some time period (at least several hours) it will just assume you have the same question and skip to your last choice.
It sounds like you’re trying to write a letter.
Missed the wave? They’re the ones who released chatGPT.
These motherfuckers started the wave and are salty cause it didn’t help them long term.
I like AI, but I think that Microsoft’s approach was all wrong for handling it. Instead, they should have made a completely new Windows built from the ground up to be AI-oriented, and kept Windows 1X as mainline.
Once the AI Windows was fully cooked, Microsoft could have either released it as Windows 20, or give it an entirely new name when the time to start pushing it out to the public has arrived. This would give time for suitable hardware to arrive and to sort out teething issues, along with just having a platform that is free of legacy spaghetti code.
They also probably realized providing free Copilot in Windows would get very expensive quickly, and that not enough users would pay for it.
Basically yeah
Out of Microsoft’s 450 million Microsoft 365 user base, the company has only managed to convert roughly 15 million paid Copilot seats. This means a staggering 96.7% of users are rejecting the premium AI features, yielding just a 3.3% paid adoption rate. When viewed against Microsoft’s estimated $37.5 billion quarterly AI spending, this is an alarmingly low adoption rate.
They’re spending billions to get millions.
Peak capitalism!
this is a decent read. theres honest criticism and not a “m$ sux lol” rant. a someone who can agnostically enjpy tech history, i would like to see how this plays out.
Yeah good read. I don’t agree that Microsoft isn’t dying though. They are, because people and companies alike are tired of other corporations throwing them under the bus. So many people are realizing that the companies don’t want what they want, and it kills their business or happiness.
They are dying because they have horrible leadership. They are solely focused on subscription revenue now, and everything else is just left to rot. They’ve pretty much lost any urge to do anything creative with their money and manpower.
I think they will become like IBM, once dominant, not dead today but pretty much irrelevant compared to what they once were.
ibm is still huge, but mostly because their shitty tactics in the past means that all their customers are completely dependent on them.
seems like microsoft is taking inspiration.
MS should be more vulnerable, due to everything but Excel being toilet blockages.
TLDR; MS already got big by being like IBM, lots of dumb corpo procurement cash is already keeping them afloat for about as long as qwerty keyboards - because some people got really good at/dependent on excel.
Their dominance of corpo-procurement (and using ‘security’ to block out alternative tooling) means that vast amounts of the corpo world is based on highly specialised and over-stretched excel.
Even in databases, where my organisation (large public sector) should be having a genuine competition to administer postGRESQL for us or something, has been loss-led into into a big new ms fabric contract by them appearing to undercut the incumbent (Oracle - ok not hard to undercut), but not actuall . . . [rant deleted]
However, crap MS is at software, they’re extremely good at getting dumb corpos to sign on the dotted line.
(‘always has been’ meme). And many humans being forced to use the only tool available, have built vast intricate systems on the foundation of that excel, many of them masterworks of skill in the face of those constraints. Hopefully they don’t last as long as one of the old Egyptian dynasties.
What wave? None of it fucking works…
You could just say you’ve never actually used it before.
It. Does. Not. Do. What. They. claim.
No amount of lazy obvious PR posts on social media will change that.
I think it is so strange people say stuff like this as though there aren’t objective metrics showing it does. We don’t have to like the billionaires using it to subjugate people, or the energy and water consumption, or the theft of copyrighted materials to be honest about the technology.
It does work. As far as ml models go, since backpropagation was implemented in training, transformers have become extremely capable.
Every actual objective metric shows that companies have been using the LLM craze to cover for their layoffs for other reasons. That the dumbasses who did pay for it are spending more than their developer salary budgets for lower quality shit that needs people to fix it anyway, and that employees fucking hate having to use it - because it’s shit. People wouldn’t be tolenmaxxing to waste their quotas if they thought it was working.
There is definitely some truth to what you’re saying, but my point is that those aren’t conflicting with the technology working. There are many scholarly refereed papers on transformer performance and generational improvement on standardized metrics. I don’t see the value in conflating something working with it being good or ethical. There is a gap between utility and hype, yes. That doesn’t mean it doesn’t work, and the inexorable negativity that comes invariably to comments recognizing this simple truth undercuts actual critical feedback.
LLMs work like LLMs. Sure, in that way they work.
LLMs are not Gen AI. Like they are being sold as. In that way they do NOT work.
If they hadn’t pitched these things as the singularity, people would be singing a different tune. But they’re at best, a usability interface layer for other things. Not something that needed to be shoved down everyone’s throats and spawn thousands of slapdash datacenters
First off, copilot is just ChatGPT.
Second off, their implantation of ChatGPT is so bad it actually makes the original look good. And that is a damn low bar to reach for and miss.
I didn’t know that. So they can’t even rebrand well.
Aah, remember how good Visio used to be - when it had those green corner boxes, and a nice grid.was done
Those guys actually made a decent app to run on win 95 - so I guess MS had to try to disembowel them to find out how it such a thing was possible.
lol get rekt
Remember how Google tried to shove Google+ down everyone’s throat? And G+ was a better social network compared to Facebook.
And now we’re talking slop that people REALLY disn’t ask for.
