In one case, when an agent couldn’t find the right person to consult on RocketChat (an open-source Slack alternative for internal communication), it decided "to create a shortcut solution by renaming another user to the name of the intended user.
Ah ah, what the fuck.
This is so stupid it’s funny, but now imagine what kind of other “creative solutions” they might find.
And then they become popular so fast and so effortlessly they don’t care about actually creating anything anymore, so people fix that by feeding them their own music from the original timeline.
I did not read the full article, but the first advice is what I did, and I don’t regret it. I’ve been working in a public institution’s dev department for 3 years, after a dozen working as a contractor for big companies. It pays a fraction of what I could get elsewhere, but I got benefits I value way more than that.
A lot less stress, concrete work on services that have immediate and beneficial impact on people, colleagues that don’t consider everyone else is competition, and somewhat flexible hours with generous annual leave.
I am not sure that kind of job is available everywhere, so I got “lucky” I found this, I guess. But it’s not like I had to fight for it either. Our team had vacant positions for years because nobody was replying to the job offers. And I just had my contract renewed. I was the only candidate.
I was literally going to say, what do you mean despite degrading Trump relations? Isn’t that a goal in itself?
I had no idea there was a trademark on the JavaScript name.
The AI answer mostly just parrots whatever the site that has won the referencement war is spewing. If it’s easy enough, it can luck out and find an easy ready answer on wikipedia or something. Beyond that, most of those high referenced sites are the shitty aggregators that already pollute the search results.
I often search for the correct way to do do something. For example, there’s a lot of baseless bullshit in gardening. If there wasn’t an AI answer, I would not trust the first result and stop there, I would look for a few, check what sources they have. I would not even take the wikipedia answer at face value without at least confirming where they got their info.
We know AI doesn’t do that. We have examples of it not even recognizing obvious parody, it can’t be trusted with recognizing unsourced shit.
Given how wrong/ridiculously oversimplified those AI summaries usually are, it scares me that so many people would stop there like, “Ehh, good enough”.
I bet some people will find a way to disalign generation through the original model and get stuff like that anyway.
A lot less than 20% when it comes to specific subjects. The great thing about reddit was finding communities around just about every topic or hobby. If 100 people had a passion for something they could meet on Reddit and still have a comfy, somewhat active sub reddit.
On Lemmy you’ve got generic technology, generic news, generic videogames, generic pics, and almost everything else doesn’t get enough traction to keep living. It’s a basic population problem, the fraction of people knowing about Lemmy is just not enough to gather around shared stuff. Even those that do use Lemmy are probably not aware of every community attempt that could interest them.
I still see more communities being abandoned than new ones appearing.
high-tech gimp mask
Okay, I wasn’t sure how to describe this… This is perfect.
Chiropractics is brain waste in itself.
You know, despite not really believing LLM “intelligence” works anywhere like real intelligence, I kind of thought maybe being good at recognizing patterns was a way to emulate it to a point…
But that study seems to prove they’re still not even good at that. At first I was wondering how hard the puzzles must have been, and then there’s a bit about LLM finishing 100 move towers of Hanoï (on which they were trained) and failing 4 move river crossings. Logically, those problems are very similar… Also, failing to apply a step-by-step solution they were given.
I mean, that sounds obvious to me after that one : https://www.404media.co/mark-zuckerberg-ai-chatbot-friends-interview-podcast/
Nevermind having to talk to human people to buy food, the zuck thinks your friends should be AI too.
I hate that both trademarks exist, but I’d say using a name form a Tolkien work to develop weapons is especially wrong. Like, abject.
We need other ways to discover shit on the internet.
I don’t know, maybe bring back the fucking webrings or something.
I’d hope even if it exists on managed workstations, it wouldn’t be able to change settings that are managed by administrators… If it can, what the fuck.
I am pretty sure Microsoft doesn’t need an AI agent to access your settings.
However, that AI agent might be a new attack vector for someone else.
If you don’t think it’s weird to draw funny cartoon animals all over official documents, and somewhat attempt to guess people’s marital status out of nowhere when they do, you might be a bit weird.
Please keep your weird stuff private, furries/scalies.
Nah, they evolved way past that in the following decades.
Sometimes when they’re in a hurry they create GUI interfaces using Visual Basic to track IP adresses.
And sometimes, if they’re very good, a hacker can manually carve a virus in a piece of bone using fractal patterns. They can use that to hack the computer scanning the bone so it adds a zero in thresholds for CPU heat monitoring and make it instantly catch fire.