wizardbeard

Querying the LLM is not where the dangerous energy costs have ever been. It's the cost of training the model in the first place.

I won't call your point a strawman, but you're ignoring the actual parts of LLMs that have high resource costs in order to push a narrative that doesn't reflect the full picture. These discussions need to include the initial costs to gather the dataset and most importantly for training the model.

Sure, post-training energy costs aren't worth worrying about, but I don't think people who are aware of how LLMs work were worried about that part.

It's also ignoring the absurd fucking AI datacenters that are being built with more methane turbines than they were approved for, and without any of the legally required pollution capture technology on the stacks. At least one of these datacenters is already measurably causing illness in the surrounding area.

These aren't abstract environmental damages by energy use that could potentially come from green power sources, these aren't "fraction of a toast" energy costs only caused by people running queries either.

I feel like the original Google AdWords, where they were only text based ads in their own boxes separate from the page content and clearly labeled as ads, weren't too egregious.

But there were already obnoxious ads around by then.

What was it meant to be an alternate ending to?

Someone made a streamer-bait frustration game where you played as Sisyphus pushing a boulder up a hill with increasingly difficult obstacles. Was very popular a few months back. Kind of like Only Up but with boulder pushing.

For those missing the fairly obscure reference

So, word of Maximum Mike (Mike Pondsmith, creator of the Cyberpunk TTRPG the game takes its setting from) is that beyond gameplay reasons, V is effectively immune to cyberpsychosis because of Johnny. Johnny splits the load of the implants with V, and most importantly, Johnny pre-death is already a somewhat functioning cyberpsycho. In cannon Johnny has slaughtered through stealth jobs and says the arm made him do it, which is how he refers to his breakdowns.

I love that they brought Mike in as a radio host in-game.

Do you really look around at the current consumer tech landscape and think the mods available to anyone but the 1% would be useful?

We already have people with artificial eyes that got shut off when the company that made them went out of business. Similar issues with other implantable medical devices.

Was a full, but short, novel that I think was summer reading: The Chocolate Wars.

Not traumatizing as much as just a shit message. Don't quietly try to opt out of what the public wants, don't rock the boat, or you'll be executed publicly as a spectacle while your peers cheer.

Kid doesn't want to participate in his high school chocolate selling fundraiser, bunch of other things happen in between, and then his classmates organize a rigged boxing match between him and the biggest school bully where they all cheer while the bully beats the main character to death. And it just hard cuts, ends there.

What a garbage book.

Against. As Lena has indicated, this does not require spinning up a full instance and admin account, but just to spin up a copy of LemVotes, which is open source. Easier than that, I've also read that votes are available without admin rights through queries to the Lemmy API. Even easier, the votes are also already public through the *bins. Just make an account on them.

I understand the use of having a small hurdle to dissuade people, I regularly build them into my scripts at work so people can't accidentally break shit with them. But my point is, removing our instance from LemVotes does not raise that hurdle to any significant degree.

This is a core limitation of ActivityPub. Votes must be sent with username attached for federation to work properly. The data is already out there. Any ActivityPub system that doesn't make them public is just doing so on the front end. It's set dressing, not actual voting privacy.

I don't like that it works this way, but I've chosen to accept it as the cost to be part of the Fediverse, to be uncensorable.

If you want privacy, the path is the same it always has been: rotate accounts regularly.

As far as I'm aware, the only true workaround is in piefed (I think it's piefed at least) where a hidden account with a randomized name is created with your real account, and the hidden one's name is attached to your votes instead of the real account. So it would require your own instance admin to see the link in vote and identity. Or basic levels of observation skills to connect the person posting negative replies is the random username also downvoting.

I also don't like the idea of even being able to opt out. It creates an entirely false sense of security and privacy, and could be seen as a signal that our instance doesn't intend to participate in the wider fediverse transparently and in good faith.

Those are simple. SMH: System Managed Heraldry, SMDH: Shaving my dog here!

This has been my argument for a while. If you're doing boilerplate once in a while, it's a good way to keep even the boring part of your skills sharp.

If you're doing it regularly, just make a fucking template you can copy paste, or set it up in your IDE's code snippet functionality.