So nice to have Nico alongside Crofty. I love the insight he gives, really adds a lot to the race.

Pop has not received feature updates for years, because the dev team focuses on implementing Cosmic.

Given the overall progress of Linux Desktop environments, this might have led many users to switch away from Pop.

I also recommend Deep Learning by Goodfellow. Long chapter about the mathematical foundations. I found it very insightful as side lecture during my studies. Also, the online version is free.

Ok that was dark.

They also love CPU.

For a dev machine, I wouldn’t go below 16GB RAM and at least a current-generation i5 (or Ryzen equivalent).

Why do you think this using racist language is a non-issue?

Ignornace hardly ever helps anyone.

I agree with your statements. Based on my knowledge of German history as well as Kahnrmann’s “Thinking, Fast and Slow” I too think that language forms reality and is strongly related to tractable actions.

However, regarding PCMR, one could certainly argue that the title is more meant like a meme. This might make it acceptable. Further, the meme-like character is in contrast to my aforementioned examples, which came from science and software engineering.

So my point is, while one should treat language seriously, the notion of what is acceptable and what not strongly depends on context. In the case of PCMR, the presence of many younger kids should make us wary of our language, because they might not understand what is meant as a joke and what not.

Soo…what’s the hash?

Fair enough, thanks 🙏

To give a bit of context: there have been similar discussions in the computer science community in recent years. Two examples are the master-slave paradigm in networking and the default git branch which used to be the “master” branch.

I see similar arguments here than in the aforementioned discussions, ranging from “non-issue”, over “language forms reality”, to “respect marginalized groups”.

I’m not familiar with the history of the sub, could you elaborate a bit more please?

I had the pleasure of conducting research into self-supervised learning (SSL) for computer vision.

What stood out to me was the simplicity of the SSL algorithms combined with the astonishing performance of the self-supervisedly trained models after supervised fine-tuning.

Also the fact that SSL works across tasks and domains, e.g., text generation, image generation, semantic segmentation…