You misunderstand, I only mean that it’s disconcerting that there may be some reason that cis-women do not find the hobby/group appealing

I know ghost has a container deployment and uses activity pub

I do wonder how many within the man/woman responses are trans, too.

Idk if that survey was mainly advertised on lemmy, but i know that at least one instance that did a survey had maybe 2% woman respondents, but more than two thirds of those were transfem.

Either way, a little disconcerting. I’m not sure what to make of that or what (if anything) to do about it

crawling other people’s content for the purpose of making money off it (in a way that does not benefit the content creator).

You’re describing capitalism there, bud

It just occurred to me that convincing someone of leaving a social media site is a lot like convincing someone to leave a big city.

They have friends there and have grown accustomed to the vibrant and diverse activities, but realistically nothing they do or have there can’t be replicated in a smaller town, a smaller media site.

They’re liable to put up with a lot of shit to stay with their community, but eventually people get pushed out and find greener pastures and a quiet space for themselves elsewhere. At least, that’s what I attribute to what I perceive to be a higher average age on the fediverse.

I’m too old to find the constant stimulation and activity attractive anymore, and I much prefer the freedom to move around and be choosy about my media choices.

I have a Samsung smart TV that is not connected to any networks, and every few days it will display a ‘detecting device’ loading screen when switching to my input that fails after 30 seconds or until I cancel it (canceling does not seem to impact its functioning)

I have no evidence but I strongly suspect this to be related to attempting to record and send device data to a remote server.

Someone didn’t read the parent comment.

Facebook, Instagram, Twitter, Reddit…

Uhhh, because these were bombs - bombs that were remotely and indiscriminately detonated. Some of the people were driving, some standing next to children or on busses full of people. There are reports of children who died because they were standing next to a target at head-level with the pager.There’s no guarantee they were even being carried by “Hezbollah’s guys”.

I don’t even know why anyone would assume otherwise. This was a loosely targeted terror attack

Likely because the bulk of those wounded by this attack were not Hezbollah

I don’t even know how you’d reasonably expect to only injure your targets in an attack as widespread and remote as this one. Seems blatantly indiscriminate at best.

I’m honestly surprised peertube has lasted as long as it has as it is

If you put up any guards at all against data tracking, they get pretty bad pretty quick. They get skewed toward the one or two datapoints that you didn’t shore up, so they think “huh, this user must really like phone games because they played doodlejump in 2016 and still has it installed on their phone”. Or at least I think. My wife gets ads that are far more on-the-nose than I do, but she doesn’t lock down her tracking data at all.

But I don’t even like them trying to match me to ads, I don’t want to incentivize their data collection practices.

I mean this genuinely: I would rather YouTube die than be subject to their overlong and hyper targeted ads.

If the ads were untargeted I’d feel less adamant, but as it is now I would sooner give up YouTube entirely.

“I have a girlfriend but she lives in Canada and doesn’t have any socials”

Lots of good suggestions here

I’m a bit surprised by your budget. For something just running plex and next cloud, you shouldn’t need a 6 or even 3k system. I run my server on found parts, adding up to just $600-$700 dollars including (used) SAS drives. It runs probably a dozen docker containers, a dns server, and homeassistant. I don’t even remember what cpu I have because it was such a small consideration when I was finding parts.

I’d recommend keeping g your synology as a simple Nas (maybe next cloud too, depending on how you’re using it) and then get a second box with whatever you need for plex. Unless you’re transcoding multiple 4k videos at once, your cpu/GPU really don’t need much power. I don’t even have a dedicated GPU in mine, but I’m basically unable to do live 4k transcodes (this is fine for me)

If i’m understanding the last graph right, it’s showing the total number of active monthly users per instance’s top communities, filtered by the overall top 100 communities?

So if an instance has activity spread out over many niche communities, that activity isn’t represented on this graph?

I would think having a diversity of smaller communities is more in-line with the spirit of the fediverse, I’m not sure of the value in slicing the data in this way.

It wouldn’t be surprising to me if they’ve had this implemented for awhile.

There’s still some question about why their 3.5 model had an apparent sudden drop-off in quality about a year ago, and among the plausible explanations for it could be that they were fucking with their weights in order to watermark the outputs in exactly the way you’re mentioning. They were also fighting against prompt-injection methods and censor disapproved uses at the time, so who the fuck knows.

You could feed it through a different, smaller model that could even be self-hosted. It isn’t difficult to make a model that rephrases an input in another style.

Not to mention that it would be extremely difficult to implement an effective watermark on text below a certain size

There are hundreds of thousands of pixels in an image where you can hide a watermark, but in a text output of a paragraph or less there are only a couple hundred characters.

How precise is the watermark? Is it a specific sequence of characters? Is it a sequence of words? A number of characters in a row? Non-print characters?

How precise the watermark is will determine how easy it is to get around. I imagine some of the most important uses to detect would be twitter/social media influence bots where the output length is only 140 characters or less. I find it hard to imagine a watermark on output of that size being effective or reliable.

They were good for about a paragraph, maybe less.

As soon as they reached the attention limit they started talking gibberish.