Microsoft admits AI agents can hallucinate and fall for attacks, but they’re still coming to Windows 11
Microsoft admits AI agents can hallucinate and fall for attacks, but they’re still coming to Windows 11
Remember those YouTube videos of scammers who push fake antivirus programs on old people? That's your computer now. Don't even need the old person in front of it anymore, because an AI version of that is now pre-installed on your system.
I'm sorry but that's being a little disingenuous. AI on a computer in no way, and I am not defending it whatsoever cuz I've gotten rid of all of that shit off of all of my devices, could do something like fall for a scammer going through and setting something up like this. It's still requires interaction with somebody opening an email clicking on something downloading something etc. There's always the human element that's going to have issues as well.
It's a joke, dude. But let's play the reality game. An email assistant that opens Word documents attached to emails and activates Macros without my explicit consent? Completely possible.
"Hello, yes? I'm calling from Microsoft, and your computer has been downloading some bad things from the Internet yes?"
(a call from a week ago)
"This product is dangerous, unsafe, and broken, and god damnit, you're gonna use it whether you like it or not" - Dave Microsoft.
"Look, I know we're heading straight for the ravine with the engine red-lining, but surely you don't want me to pump the brakes now and waste all that effort?"
"...I do in fact want that."
Not sure what this is a quote from, but I read it as Micheal Scott and Jim.
Those of us not externalizing our thinking to an LLM can still be creative all on our own. Feel free to quote me - I'd be honored.
Sunk cost fallacy is that you?
🐧🐧🐧
🐧
They're still coming to your Windows 11. You think the Microsoft execs are using it? Hell no. They're either using Macs, or they're disabling all this dumb shit they're putting in Windows 11, probably with controls they don't even give you access to.
(Disclaimer of bias: Happy Mac user over here, reveling in the fact that Apple Intelligence is a complete and utter failure... I see it as a feature, not a bug. I love that Apple is so far behind on AI and it makes their platform more valuable to me. That said, they are partnering with Gemini for the next version of Siri "sometime next year" (they've been saying "new Siri next year" for years though) so macOS is not a safe space from AI. Just presently the one with the least developed one.)
Microsoft has had a long history of a company culture of "eating their own dog food", forcing themselves to use what they force on users, so they're making the underlings use it at least.
As far as executives go, at almost any sizable company they hardly ever spend time at their computer. Cell phone or tablet, and have their assistant or direct reports do anything more complicated than half paying attention to a meeting.
Lastly, the controls to turn it off will be available to all of us as long as you have a Pro or Enterprise SKU (Windows license/install). They aren't going to fuck over their business customers with the unwanted slop they force on the proles.
Protips:
Well that explains a lot.
Do you think our users will like this experimental broken... Even with the security, privacy, and...
You're absolutely right...
Happy Linux user over here. Free open source AI models are becoming much more powerful, and things like "Apple Intelligence" and "Co-Pilot" will be looked back on like Netscape.
Getting a free older computer from my work soon because it's too old to "upgrade" to Windows 11 so I'll be turning it into a Linux machine. Pretty dang psyched mostly for all the free software!
Gotta be honest, though, a locally hosted 70B model with basic RAG functionality isn't exactly playing in the same league as the market leaders, which can be bigger by two to three orders of magnitude. And a model that size is already around the limit of what a beefy gaming PC can do with reasonable performance. We're unlikely to ever beat the big players on quality with local models.
What might happen is that the market collapses, the big players all go bankrupt, further LLM development ceases, and locally hosted Qwen3-80B will be the pinnacle of available text generation for the next thirty years.
You think the Microsoft execs are using it?
I actually think it is very possible that they are using it. Because I also think that the execs have no clue how any of this works and use AI extensively to make their decisions. At least that seems to explain a lot of the stuff going on at Microsoft and elsewhere.
Microsoft execs are 100% using AI.
You would probably be surprised by how many "computer illiterate" executives work at Microsoft, and they are perfectly happy for their underlings to use their own broken ass systems. Most executives do all their work from their phone or a tablet anyways.
I love that Apple is so far behind on AI and it makes their platform more valuable to me.
Same here! There's dozens of us.
AI agents are literally the only ones coming to windows 11.
The rest of us aren't even in the mood for a quickie with Microsoft anymore.
AI, AI über alles, über alles in der Welt...
There are things you can do better with AI, but not fucking everything and not anything important