Highlights from my media diet for the winter of 2024-25

Ordered from most to least serious, which naturally places most non-fiction at the front and fiction at the back.

Boaz Barak with Six Thoughts on AI Safety

Not sure what to think of these six thoughts, but I also appreciate that a moderately main-stream, respected AI/ML academic is placing his two cents on the topic.

A blog post from the Alignment Research Centre on a No-coincidence principle.

A computational formalism of what makes a coincidence (it can be efficiently explained). Appreciate the update on what ARC are up to, and I also really like the flavour of this kind of research.

Inference Magazine with “AGI is an engineering issue”

Well laid out explaining what the big labs are focusing on at the moment, and how they see their capability path forwards, which has a bone-chilling lack of serious notkilleveryoneism.

Lesswrong User tandem with a short comic about EAs and rats

Potentially one of the best “community observation” posts ever, describing both vibe (meta-level) and topics (object-level) of the EA/rat community.

LessWrong user L Rudolf L with a review of Planecrash

Making planecrash accessible to the masses. Amen. Conditioning on the fact that you are reading this, you might be particularly suited to reading planecrash, but it is hard to give a blanket recommendations, due to its weird mix of decision theory, BDSM, lectures on logic and probability, reliance on fantasy tropes, and more.

Iris van Rooij on “Is the brain (just) a computer?”

Succinct, clear, and educational, this blog post by dr.van Rooij sketches the difference between asking “is the brain a computer” and “can we explain human cognitive capacities computationally?”.

A Tom Leinster talk on Entropy and Diversity

An excellent talk starting with a relatively pure maths concept - how do you define the magnitude of finite sets - and leading into how this concept applies also to diversity and” effective number of [unique elements]. Was reminiscent also of different Information Criteria and similar in Bayesian Model Comparison, which proposes to measure “effective” model parameters.

Abel Jansma on Complex Systems and Quantitative Mereology

Seeing this complexity-theoretic, math-heavy approach to “how do we carve up systems” is very cool, and it seems also to connect to other stuff I’ve been thinking of recently, category theory. If your category is mereologies (of a system), then you want some kind of functor to map between those mereologies; but on a little bit further thought it’s not actually clear that this is doable, as you might compress information, meaning that composition and symmetry doesn’t follow.

Tracing Woodgrains on Wikipedia editor David Gerard

I quite liked this deep dive, showing the intersection of what I now consider my community

Good conversations have lots of doorknobs, by Adam Mastroianni

Adam takes creates a dichotomy of conversationalists - givers and takers - and showcases what each of them gets right and wrong in a conversation, by comparing it to practices in group improv comedy. Great insight. In the past I’ve also liked his How to De-Bog Yourself.

What goes Without Saying, by Sarah Constantin

Some short observations on extended lesswrong/rat culture. There is a difference between bering seen as serious, and actually being Serious, recognising when you are doing what, and that it is your job (you in particular!) to make sure you are being Serious.

Wind and Truth

Wind and Truth was good, maybe even great, but not peak. I think Sanderson is a decent writer, a great world-builder and plotter, but that some of the moments did not hit as hard as I had expected (and hoped). At some times it felt a bit too Marvel-ish.

Pantheon

Pantheon is an animated science fiction series on Netflix. It was quite good. However, I think its first season doesn’t end “satisfyingly”, and I am abstaining judgment of it for when it ends. What is out, is good sci-fi, but if you’re the kind of person who needs to watch a finished story, this isn’t for you (yet).