In theory, it’s just an algorithm that will always produce the same output given the exact same inputs. In practice it’s nearly impossible to have fully deterministic outputs because of the limited precision and repeatability we get with floating point numbers on GPUs.

I’ll admit I don’t spend enough time on here to have a well-formed opinion, but aside from the occasional obvious spam post, I don’t see a lot of obvious bot activity… Depending on the instance/community, you’re going to get a lot of very strong and, uh… controversial opinions.

Keep in mind that the (still relatively) niche nature of Lemmy/Piefed (and the fediverse writ large) means that a higher proportion of people around here are going to be far from the center of the political gradient… and the western world is very much experiencing a cultural/political paradigm shift which is driving people to the fringes.

I also think that a lot more people are using the platform than a year or two ago and that has knock-on effects on moderation that we still need to figure out how to deal with.

imo when someone is so lazy that they can’t be arsed to put effort into using genAI properly to produce an image that isn’t the most generic slop imaginable, they’re probably also too lazy to have developed the empathy required to come to actually valid conclusions

Kagi is good too.

I’d like to see similar testing done comparing models where the “misaligned” data is present during training, as opposed to fine-tuning. That would be a much harder thing to pull off, though.

I’m just imagining the human half doing like a naruto run.

Never said they were wrong about the application of the first amendment, because they’re not. They just absolutely ruined what could have been a teaching comment by being a condescending asshole.

Nice wall of self-righteous pseudo-intellectual bullshit you have there. Maybe try being constructive next time instead of letting a stranger on the internet rustle your jimmies.

I’m using podman, and I don’t like the practice of unnecessarily setting UIDs. NET_BIND_SERVICE is exactly the flag it needs to set port 80 and it doesn’t potentially complicate accessing the files for maintenance. Does your system have SELinux? If not, that might be why you don’t need it lol.

Vaulwarden does this and I’m really frustrated that I have to cap_add NET_BIND_SERVICE in my rootless setup just to make my password server run.

Honestly I think proper search is one of the biggest things holding the fediverse back from mainstream adoption. It needs exposure, and it needs to be easy to find information and communities around obscure topics in order to really replace silos like reddit and facebook. I’m glad to see this exists, and I particularly like that it supports kagi.

Edit: Maybe this should have been obvious when seeing that it uses other search providers as a backend, but all it does is pass a list of fedi websites to filter results for. That’s… not a proper search. It’s a potentially useful tool, but it’s not doing any of its own aggregation and more importantly the list of websites is also painfully small. Color me disappointed.

Looks more like a pelvis than a head to me. 🤷

that’s definitely one way to say “luck and happenstance”

i prefer to do mine in the czech republic, personally

Seems to be a common thing with Chinese tech, honestly. It’s kinda frustrating.

Laziness doesn’t exist. Everyone has their own struggles and it’s pretty fucked up to judge someone because they don’t choose to put their limited energy toward overcoming the obstacles that you take you’ve taken for granted having the ability to overcome. We should be encouraging curiosity and exploration, not demonizing people who got left behind and are too scared to try new things.

Being able to upload to Google Play Music was such a great feature. Honestly though, maybe it’s best that they killed it when they did. That way it can live on in memory as the example of what a music streaming service should be like… instead of becoming an enshittified zombie like spotify is today.

Hertz is the worst rental company I’ve had to deal with. Enterprise is the only one I’ll use if I have a choice.

Human posting of AI-generated content is definitely a problem; but ultimately that’s a moderation problem that can be solved, which is quite different from AI-generated content being put forward by the platform itself. There wasn’t necessarily anything stopping people from doing the same thing pre-GPT, it’s just easier and more prevalent now.

Yeah but I get like 5 spam calls a week from “Nicole with loan processing” or “Joshua Martinez from the loan issuance department” and they always leave a voicemail that I have to delete because I tried to get a personal loan 4 months ago.