pizza_is_yum

Our king is a benevolent liter.

I have to use this.

That 80% is important. We need non-techies, because they remind us that there's more to life than just computers.

Could have left out the first part of that headline... I was confused about why this was in my feed for a second 😅

The real world is where things that matter happen. Life, love, nature. The Web is distracting and loud, but it's a big, flimsy illusion. So I don't think there is any chance that Metaverse ideas will take off.

Regarding LLMs: good news! You can already run them at home. Check out KoboldAI. LLMs will become smaller as time goes on, too. There's lots of room for improvement in that field.

If, on Mastodon, you make a DM between you and someone else, then you @ a third user in that DM, then the third user becomes able to see that DM. At least that's what I heard a few weeks ago.

This is because DMs are, regrettably, just normal posts with default visibility of two people. If you @ more people, then they gain visibility into the DM.

This guys is one of my favorite Science tubers. Him, and Joe Scott (Answers With Joe)

Cool. Btw, the authors tested their own 2 adversaries. The 1st failed to breach the defense, and the 2nd was deemed "impractical" because of how slow it took to train.

I appreciate their positive outlook, but I'm not so sure. They say they are well-defended because their equations are non-differentiable. That's true, but reinforcement learning (RL) can get around that. Also, I'm curious if attention-based adversaries would fare any better. Seems like those can do magic, given enough training time.

Great work though. I love this "explainable" and "generalizable" approach they've taken. It's awesome to see research in the ML space that doesn't just throw a black box at the problem and call it a day. We need more like this.

Agreed 100% on the account proliferation and type asymmetry points. The way things stand, right now, the user's choice of account provider will determine what actions they can take on the fediverse as a whole. It is a wholly unfortunate state of things.

An interesting exception would be Owncast's "Fediverse auth" option for stream chatting. That sends a One-Time code to your mastodon inbox for authentication.

As @jackalope@lemmy.ml suggested, Solid would be a shoo-in for your "User Data" server. If, that is, Solid could shake off some of its sheer conceptual gravity. People say the fediverse has a geek problem, i.e. only geeks use it. Well, I think Solid has a worse version of that problem. It is only approachable by the deepest loremasters of geekdom. They are also still vague on its actual operation. What's more, they are still deliberating what their actual security model will look like.

Which makes me sad, because the Solid sounds exactly like what we architecturally need.

EDIT (3:25 am EDT): Just wanted to add on here, I really think that "linked data" and SPARQL were bad, possibly self-defeating decisions for the Solid project. I sorta see their motivation–they want that sweet, sweet flexibility. But I think this approach is not a good solution.

EDIT again: added links

For about half a year, I was using a dumbphone. I only switched back because it started having bugs a few months ago. Losing MMS messages, failure to show the name associated with a number, and some other stuff. It's a shame. If you're curious, it's the Sunbeam F1.

While I'm waiting for an update to fix it, I've switched back to my Samsung smartphone. Sure, it's heavy and distracting, but at least I can reliably receive calls and messages. Which is the point of a phone to begin with :/