300€ is already a lot people won’t be willing to pay for just one part of a computer.

You may. You seem to not have seen the potatoes people are using. ;)

Decades worth of windoofs hate. Finally did the full switch a few months ago. I have been using windoofs for gaming only anyway and professionally I already had plenty of Linux experience.

Linux isn’t great either though and can be a real pain in the ass. Recently had to fix few lines of code in the kernel, because I had no audio. But the fact that one can do it in principle is already a huge win over windoofs. And although some games stubbornly won’t run (stable), a lot do without having to tinker much, thanks to Wine, Proton, and a couple of startup parameters.

Professionally I mainly use Ubuntu, for gaming I did so at first as well but switched to CachyOS recently.

What decisions for example?

You can install things from random websites for Linux too, though.

Ah I see. This is getting ridiculous.

I wonder where you draw the line of when the use of resources for technology is okay and when it isn’t.

Did it make a mistake?

The reasons (summarized using Copilot):

• The platform no longer aligns with Debian’s values, social contract, code of conduct, and diversity statement.
• Concerns over X becoming a place where people they care about don’t feel safe.
• Abuse on the platform happening without consequences.
• Issues with misinformation and lack of moderation.

GPU rendered text interfaces are pretty ubiquitous already. You can find that in IDEs, browsers, apps and GUIs of OSs. Drawing pixels is still a job the GPU excels at. No matter whether it’s just text. So I don’t see a point why we shouldn’t apply that to terminal emulators as well.

This again?

If we’re speaking of transformer models like ChatGPT, BERT or whatever: They don’t have memory at all.

The closest thing that resembles memory is the accepted length of the input sequence combined with the attention mechanism. (If left unmodified though, this will lead to a quadratic increase in computation time the longer that sequence becomes.) And since the attention weights are a learned property, it is in practise probable that earlier tokens of the input sequence get basically ignored the further they lie “in the past”, as they usually do not contribute much to the current context.

“In the past”: Transformers technically “see” the whole input sequence at once. But they are equipped with positional encoding which incorporates spatial and/or temporal ordering into the input sequence (e.g., position of words in a sentence). That way they can model sequential relationships as those found in natural language (sentences), videos, movement trajectories and other kinds of contextually coherent sequences.