November 1, 2023

A Giant Continuous Blob of Semantic Functions (rwblog S6E12)

I have had a bad few weeks, I will not pretend otherwise.

In lieu of original thoughts, here are some interesting links I’ve found lately.

• Robin Sloan (you’re reading his newsletter, right?) has a new book coming out next year, and he built a whole darn companion website. I wish more novels came with all kinds of bonus materials and deleted scenes and companion essays! Where’s my director’s cut of Tomorrow, and Tomorrow, and Tomorrow? (Interestingly I do think this is more common in the comics industry — for instance, in preparation for the upcoming TV show, I just finished a reread of Scott Pilgrim in its full-color, bonus-materials-included rerelease.)
  • Of particular interest is the essay "The widening aperture”, about stories that progressively grow in scope. Sherry talks about this a lot.
• A series of articles about the nature of fame and skill in the TikTok era: an art critic puts up a lightly-critical essay about a TikTok-famous painter, TikTok-famous painter responds by unleashing his followers.
  • I am also famously bad at taking criticism, but at least I’ve never spent a day angrily encouraging bad-faith attacks against a critic who wrote a frankly very measured piece 🤷‍♀️
• I am trying very hard to memorize "Beautiful Mind”, the latest banger from Tom Cardy and Brian David Gilbert. I think you should too.
  • As a bonus, consider also watching Tom Cardy’s "Perception Check”.
• Dan Mall answers common design questions, with short, punchy responses that exhibit exactly no nuance. It is partly a joke but partly not a joke. I wish more FAQs read like this.
• Here’s a rather new (to me) way of thinking about LLM prompting — consider it as a giant continuous space of semantic functions that take text to text, in which case prompting is querying in this space for a specific function.
• I recently read Peter Naur’s “Programming as Theory Building” which was pretty mind-expanding (insert that gif of the science guy doing the whooshy science explosions. You know the one I mean). Basically, it’s arguing that programming just is the process of building a “theory” of how the program should work and there’s basically no way to transfer that knowledge from one brain to another (e.g. in an artifact like the code itself) — a new programmer has to just do the work to build a new theory. Anyway go read this paper it’s like 8 pages and very thought-provoking.

In other news, I ported all my newsletters to my website. That also means you can search them now!