Yes! The way those physics models are created is so cool. The article somewhat explains it, but it’s mostly a fluff-piece for things unrelated to genAI. More in-depth:

The physically accurate simulation is great but slow. So we can create a neural network (there’s a huge variety in shapes), and give it an example of physics, and tell it to make a guess as to what it’ll look like in, say, 1ms. We make it improve at this billions of times, and eventually it becomes “good enough” in most cases. By doing those 1ms steps in a loop, we get a full simulation. Because we chose the shape of it, we can pick a shape that’s quite fast to compute, and now we have a less-accurate but faster simulation.

The really cool thing is that sometimes, these models are better than the more expensive physics simulation, probably because real physics is logical and logical things are easier to learn.

We’ve done things like this for ages. One way we can improve them is by giving them multiple time steps. Unfortunately they kinda suck at seeing connections over time, so this is expensive. Luckily, transformers were invented! This is a neural network shape that is really good at seeing connections over one dimension, like time, while still being pretty cheap and really easy to do run in parallel (which is how you can go fast nowadays).

With a bunch of extra wiring, transformers also become GPT, i.e. text-based AIs. That’s why they suddenly got way better; they went from being able to see connections with words maybe 3-4 steps back, to recently a literal million. This is basically the only relationship with “AI” this has.

It’s anarchy for the hierarchies, not for the components of the hierarchies

So… databases? Especially in data centers? Still a nice boost in that case

If you haven’t already, try out KDE’s Krita. Incredible piece of software, much better for drawing imo

Ngl that link puts me slightly off. It reads exactly like what people booted for very good reasons say

The following paragraph shows how so-called cancel culture was used weaponising […]

And in the email, Mozilla talks about him violating their “inclusivity” policy… we also don’t know what was reported, only the reasons stated.

Not saying that it wasn’t unjust, just that we only have 1 perspective and it’s written in a way that raises some red flags.

Do advertisers maybe require a bit of JS to be run to validate a click? I can’t imagine they’re happy to lose money to completely invalid clicks…

Yeah, did:web exists, but I still called it centralized because it still relies on did:plc pretty much everywhere (though honestly domain name handles might actually be did:web, not sure). Didn’t know about that dual setup by Bluesky though!

I did notice the @handle.invalid! Thanks!

My understanding was that activitypub was basically a rough formalization of existing protocols, designed to be as flexible as possible. More a template than a real protocol. Unfortunately mastodon’s popularity basically made a bunch of things de-facto obligatory but not well documented, and there’s still a bunch of ways to do… anything.

That link doesn’t work for me, but I ended up finding a post by them that seems to correspond. Good to know, thanks! Seems like it’s realistic but expensive still (150$/mo?), and it’s not gonna get cheaper… I hope they figure out a way to make them less centralized.

I believe that’s your handle, not your identity. Your handle resolves to your identity, but your identity isn’t directly tied to it, in case you lose the domain.

The aggregator is called the Relay, and I haven’t even found anything suggesting one could realistically selfhost it. Then you need to handle the massive stream of data coming through it with AppViews, which are tough to handle too (there are a few but not many iirc).

That said, I am also impressed with the thought behind ATProtocol. It seems much more robust and defined than ActivityPub.

Bluesky’s federation model is actually quite interesting, they go for a very portable approach vs activitypub’s instance-basis. Unfortunately, there’s still a massive centralization point (the main relay, the only thing that can really handle the firehose), and identity is also centralized, albeit has mechanisms to be decentralized.

Yeah that’s fair. I don’t quite know why I read that the way I did, but I read the “choosing” as “lives there and isn’t actively attempting to move”.

… do you just expect everybody who lives there to pack up and leave? Even though their entire lives might be there and moving costs a ton?

Wow, I didn’t know that! That’s absolutely wild

… The root of the thread you’re replying’s main body is stuff JSO has actually achieved.

I’m assuming this is referring to JSO.

https://en.m.wikipedia.org/wiki/Just_Stop_Oil

Beginning on 1 April, they carried out England-wide blockades of ten critical oil facilities, intending to cut off the supply of petrol to South East England.[33][34][35]

On 26 August, the group blocked seven petrol stations in Central London and vandalised fuel pumps. Forty-three people around London were arrested on suspicion of criminal damage.

On 20 June, the protestors spray painted private jets at a private airfield at Stansted Airport. The group had been targeting a jet belonging to singer Taylor Swift, but could not locate it.[140]

Yes, a lot of their protests are “awareness” stuff (basically none of which do actual damage. Unlike oil, actually!). No, it’s not just that. The UK isn’t an active warzone so bombing stuff is slightly more difficult to justify.

I’m not an OS dev, I have no idea how stuff this low-level works.

I’d suggest some kind of “press this key to view debug information” text (or make it documented but not visible, to avoid people just pressing whatever button is written on the screen)