… tests from earlier this year found that AI agents failed to complete tasks up to 70% of the time, making them almost entirely redundant as a workforce replacement tool. At best, they're a way for skilled employees to be more productive and save time on low-level tasks, but those tasks were already being handed off to lower-level employees. Having an AI do it and fail half the time isn't exactly a winning alternative.
post
Copilots numbers likely that high merely because it rides along with Office365... I tried using it a few times, and was completely useless. Even failed at sorting a spreadsheet with a few parameters
I wrote a page long documentation on a project. I asked Copilot to "format it to look nice but do not change a word"... it told me how to make some headings bold (would not do it itself) and what not... that's the "assist" I got
I couldn't even copy/paste the format since it's reply window does not apply and the text it provided was interlaced with its own stupid comments letting me know bold headings are more visible than regular font
I get better results just bouncing ideas off my cats
That was my experience. Wife had work telling her to use it, she asked me for help. I tried to get it to do things and all it would do is suggest stuff that we both knew perfectly well how to do with shortcuts. As for anything complex like have a chat and generate a document: fuck no. Might as well go to chat-jippity and copypasta it's result and format it yourself. Utter waste of time. I don't see why it's there, I can't find a use-case.
I work with sensitive data... so I often grab a real message, gut all the PHI and refill with some fake data.
I had a project where I needed to do a lot of these, so I got DeepSeek to give me a list of superhero "real names", DOBs, gender and a few other fake things so I could automate filling these messages with fake data. This is the most success I have had with AI and even then it messed up (minimally) thinking for some reason that Wanda was a guy hahahaha
i think MS is just putting AI into everything, just for the sake of it, they dont really care if its useful or not at this point. they just need to buy time to soften the blow when the bubble bursts.
We put it in everything, and bet all of our money on it to "soften the blow"
MS is running into a real problem where its two major product lines, Windows and Office, don't have any major improvements that justify an upgrade. It is an existential crisis to the company's profitability.
Now, MS has been able to make Office into a subscription, but it can't do that with Windows.
Did you try asking copilot tips on how to use it properly? Edit* this was a joke. Ai is garbageware that nobody wants, just like ads and capitalism for capital sake. Now untwist your panties.
Yes, I did. Everything I tried on copilot wanted me to upload corporate data to the cloud. (Yeah, NO). It told me it could help with my email…if I uploaded them individually. (Still bad practices here and breaking corp policy).
I expect LLMs should be really good pattern driven activity, but I’ve yet to figure out how to make this useful.
I tried to use a local LLM to summarize outline and discuss my *.md notes for annual review. It sucked at it if it didn’t completely crash the model. It couldn’t even provide a unique list of all tags in the files. (It took me about 30min manually). I thought that it would be good at that. I would have been better off spending to learn find | grep commands or spent time learning python.
I’m still searching, but maybe one day I’ll figure out a use for these.
My employer is all in on Microsoft, copilot is terrible, it can't even find a word in a document. Cntl+F find's it no problem.
Now when I have a tech issue that I need an answer for, the bing AI generally gets me a detailed answer on the first try. But it's my understanding that bings AI is just a reskined ChatGPT.
HP: if your non original printer cartridge fails 7 out of 10 times, is it really a savings?
Microsoft: we want 80% of our work to be handled by these ais that fail 70% of the time.
Also fuck HP.
And Microsoft.
What's with all the shills in that comment section? Yeesh.
There's a lot of money and a lot of careers riding on this bullshit somehow becoming successful.
A sentence for the ages.
That's like people trying to save the Titanic by bailing water with shot glasses.
There's a lot of money and a lot of careers riding spent on this bullshit somehow becoming successful.
Both can be true
All of them have exactly 1 post too.
I hope all the money thrown at this "AI" (misnomer, IMHO - it's really just extremely overwrought pattern matching) causes at least some significant humbling (if not outright downfall) of some tech giants. I haven't programmed in a couple decades, and yet even I could tell they weren't gonna get to AGI offa this crap - I can't believe how badly some of these supposed techies fell for their own hype.
When discussing it, I often call it "simulated intelligence", because at the end of the day that's what neural networks are.
Edit: only to non-technical people, as simulations are a different thing.
In science fiction I've often seen the term VI (Virtual Intelligence) to refer to machines that look intelligent, and could probably pass a Turing test, but aren't really intelligent (normally VI coexists with actual AI, often used as interfaces, where it would be a waste, or too risky, to use a proper AI).
LLMs look a bit like that, though they're probably too unreliable to use as an interface for anything important.
IMHO it's real intelligence, but not artificial. LLMs have been fed virtually everything created by the actual intelligence: humans. Like I said, all they do is execute what is effectively pattern matching (on serious steroids) to distill what humans have created into something more bite-sized.
It's pattern matching, but it's not matching intelligently. An intelligence should be able to optimize itself to the task at hand, even before self-improvement. LLMs can't select relevant data to operate on, nor can it handle executive functions.
LLMs are cool, and I think humans have something similar to process information with, but that's just one part of a larger system, and it's not the intelligent part.
Ask an LLM how many 'r's are in the word 'strawberry' and tell me it has any actual intelligence behind its output.
the correct term is Stochastic Parrot... that is what LLM do. It sound even more cool that AI imho
No its not. They havent been this way for years
There are several dozens of these studies
Doesn't matter. There is no cognition. Just word salads mixed and matched with no possibility of receiving "I don't know" for a answer.
Still stochastic. Even now they still can't reliably do repeat tasks
So they no more use probability to choose next word? I wonder how they do it now
That's was a remarkably uninsightful way to approach that topic. Please link to more of these "studies", that one was way too short.
The virgin cited study vs the Chad Ad Hominem
Did you read the study? It's hilarious. They're using LLMs to "grade" the number of observed "skills" based on the output of LLMs. They're using a stochastic parrot to evaluate another stochastic parrot, and concluding that there is some kind of emergent "skill" going on. Sheeeesh. It'd assume the authors of the paper are just having a laugh. But, one thing is for sure, the AI stupidity train keeps chugging along.
It's AI in the same way the computer controlled opponents in any video game since the 80s is AI. When you realise that is the metric by which something is judged to be AI or not, calling LLM AI seems a lot more valid much also much less impressive.
Congrats to all the insiders are Microsoft who were able to make money off a hype cycle. They'll fall for the next one too.
They could have saved so much time, money, and reputation if they had just listened when absolutely everyone told them to fuck right off with that "AI" bullshit. No one wants it, no one needs it, no one likes it. On the other hand the healthy bump in linux market share is a nice trend I hope to see continue.
They could have even just put AI code/options in, and left it opt-in, so that people that wanted to use it could and those of us who don't wouldn't have to constantly fight it. Instead they forced it on all of us, and even had the gall to raise the price of office claiming that is because AI is in there. They create the perfect one-two punch for us all to hate it, even those that may have been just neutral before
My favourite part about the price of Office going up is that there is an option to NOT have AI and pay the old price, but they didn’t tell anyone about that, and just jacked up the price.
In Australia they got caught and had to send out an email to give everyone the option to switch back and receive a refund.
In what feels like their last petty act of revenge, that refund was made in seperate ~$5AUD ($3.30USD) payments for every month that they had lied to their customers. Like a disgruntled guy paying his parking ticket in pennies. I received a dozen identical miniscule payments in one day, and could feel the unsaid “fuck you” from MS. The sentiment is mutual.
I've been tasked with getting more copilot automation integrated at my job. After a bunch of research on security, governance, app Lifecycle, and the complexities to get that setup properly, I brought pages of notes to the manager in charge. I expressed how expensive it is, how the product is awful out of the box, and how we don't have any defined use cases. She asked me if it's so complicated and expensive, how are other companies doing it. I told her I don't think there are many companies that have really gone in on agentic. I think 99% of use cases is people using copilot as a search engine. I'm going to be sharing this article even if it's filled with bias.
The problem is that it's a truly shitty implementation because it can't actually do anything. It will tell you how to do things, but it won't actually do a task you would think it's suited for.
It kind of reminds me of some kinds of Christians that believe people will convert if they just tell people about Jesus. As if we've never heard of Jesus and the Bible before. We know. We just don't believe.
At least this time around they scaled back instead of doubling down on marketing.
Progress???
I want a local only AI where my research about breast size doesn't end up in the wrong hands.
When copilot was added to my work computer and we got a big email announcement I wanted to see what it would say about some pointed questions about our industry, but in the end I gave up because it failed to even work enough to see what weird shit it would say
If they’d just advertised it as a handy little feature I might have used it, but the way they’re constantly ramming it down your throat makes me so suspicious that I just don’t want to touch it.
I mean just think back to the voice assistant craze 10-15 years ago (SOMEHOW!) Apple had Siri, which actually solved a problem. Smart phone-sized touch screens are the genital cancer of computer UI, so what if we add the ability to do common tasks such as web searches, media control, clock and calendar operations etc. by voice command? Google followed suit.
Then came Alexa. Amazon wanted in, but they didn't want to help you use the device you owned, they wanted to facilitate using their services. Meaning you had to either install the software on some other device, or they had to make and distribute new devices. Commercials featuring Alexa, often starring a pre-homicide Alec Baldwin, usually demonstrated making product and media purchases on a speakerphone-like device. I wonder if anyone has ever actually done that? Because almost all the time I want to examine the listing to make sure I'm getting the thing I want. Like, "Alexa, buy a copy of Army of Darkness." Okay, did I just "buy" it on Prime Video? Am I about to get a copy of a DVD or Blu-Ray in the mail in two days? Did it just install the 2011 tower defense game on my Kindle Fire?
Then came Bixby, because Samsung are under the impression they need to make software for the phones they make. It does mean I can ask my cousin's refrigerator to beatbox, and it will.
Then came Cortana. Microsoft, as usual, arrived late to the game doing a mediocre job of what everyone else was already doing for no reason other than to do what everyone else was doing. Naming their voice assistant after the video game character made it more difficult to google, most of what it did was yap during the Windows install and onboarding process, and then nobody used it because the entire reason for a voice assistant to exist - making up for the shit UI that is a tiny touch screen - doesn't exist on PC with a large screen, mouse and full keyboard.
Windows 10 was kind of fascinating for that; Cortana was one of several technologies they tried to go all in on, releasing a big update that added apps and frameworks right into Windows, only for them to unceremoniously rip it all back out later.
And now we get to the AI era and it's happening again. I imagine Copilot will be the first to go because nobody likes Microsoft's built-in shit.
A big part of the problem with AI is that for your every day use cases, voice assistants already filled that niche
Most people just need some very simple language parsing that can automate some tasks like, "set a timer for 5 minutes". We had that years ago and how exactly modern AI differentiates itself from that is largely unclear to most people. And when it is clear, it's apparent that it's a lying machine that lies because that's fundamentally just how the tech works. And it boils the oceans
And yet, with Android Auto, Google stuck the dumbest version of assistant to handle voice queries.
(While driving to hardware store) "Hey Google, what size screws does a standard VESA mount require?"
"I'm sorry, I don't understand"
It's the ONE place i need a working assistant/AI the most. Yet it's useless with natural language queries 99% of the time.
I do have to wonder how much especially Google is gimping ordinary search in order to make their LLMs seem more useful. Just using the internet has gotten so much worse in the last few years.
A big part of the problem with AI is that for your every day use cases, voice assistants already filled that niche
Actually, the problem with AI is that we have a metric fuck ton of tools that already do a better job and take way, way, way less resources. Instead these companies are pretending their AI is approaching AGI and want it to be the solution to everything.
Window’s real problem is not the quality of their model or implementation, don’t get me wrong they’re bad but not really any worse in regards to usability or usefulness.
The real issue is that Google can implement it in a way where it’s difficult to turn off or avoid using. Microsoft can’t really do that, no matter how hard they try, people can choose to just ignore the features, Google can just make them appear and run without user’s choosing to engage with them.
You say this, but apparently LG has started to install copilot on their TVs as a system feature, so you can’t even remove it.
Sure, but people can just… not hit the button if they don’t want to. They are tools that Microsoft is pushing users to engage with, not systems that automatically modify the service whether users want it or not.
Google has set it up in a way where the “demand” is generated whether or not users engage with it. Microsoft’s stuff needs users to actively engage with it for “demand” to be created. Google has stuff like the summaries running by default, or the weird smeary filters on YouTube videos or auto dubbing foreign languages.
Basically, Google owns all data you access with it's services and generate in a process of using them, while the CoPilot vampire usually needs to be invited first. OneDrive autouploading your files and leaving shortcuts was one desperate attempt at it, Recall was another, but they are comparatively disadvantaged with you having your own local storage/OS and not their tightly controlled cloud from the beginning.
Google can just make them appear and run without user’s choosing to engage with them.
That is what the start menu is for.





all 55 comments