Yeah, something in the algorithm that prevents one magazine or one instance dominating would be the best approach.

I like that kbin and Lemmy aren’t going to have an entire team of people doing continuous A/B testing to find what drives the most engagement.

Much less chance of this ending up as a dopamine slot machine.

Hi from Artemis 🫡

It’s pretty great.

Do we trust them not to steal credentials?

To get even more pedantic…

It’s defined on how far light will travel in a vacuum in the time it takes caesium-133 to do a certain number of transitions between hyperfine ground states.

It’s cool how almost all units of measure are defined on caesium

cafelat robot lever espresso machine

• $320 USD
• makes espresso as good as a +$1k machine
• doesn’t need descaling
• looks cool
• takes up hardly any bench space
• simple, well engineered, will outlive me

I appreciate the point you are making, but here’s a solid example of how it gets confusing.

ANNOUNCEMENT: defederating effective immediately from lemmy.world and sh.itjust.works

Who is defederating?

It’s the defacto term for how we fit a statistical model to data, unrelated to any copyright concepts. I’m pretty sure we called it “training” back in 1997 when I was doing neural networks at uni, and it’s probably been used well before then too.

Neural nets are based on the concept of Hebbian learning (from the 1930s), because they are trying to mimic how a biological neural network learns.

This concept of training/learning has persisted because it’s a good analogy of what we are trying to do with these statistical models, even if they aren’t strictly neural networks.

You missed an important part of the headline “is feared to have…”

It says she caught a “cholera like” infection

It annoys me how none of the news articles mention spez’s lying about the Apollo Dev trying to blackmail Reddit.

That’s the singular thing that drove me away.