Here is my annual roundup of some of the most interesting things I came across in 2023.

Cory Doctorow on TikTok’s Enshittification

Cory wrote this piece at the beginning of the year, which applied the previously coined term “enshittification” to TikTok. The term refers to the observed trajectory of platform technology products that starts great, then increasingly becoming worse and worse as they attempt to extract value more and more egregiously. There is definitely plenty of examples that fit this pattern, which the author mentions, but I’m not sure that it really applies to TikTok. Mostly because it is not a typical technology company, driven purely by profit incentives.

TikTok and its parent company ByteDance are in part state-influenced information tools, and so have a whole set of ulterior motives to take into account. Of course, with this rather nefarious set of motivations, arguably TikTok is already “shit” for its users. But regardless, the “enshittification” as described by Cory may not apply to it. The incessant need to extract more and more value out of its users won’t happen the same way, because it is not just the monetary value of their users that they are interested in.

Stephen Wolfram’s Primer on ChatGPT

GPT-based large language models (LLMs) have definitely been in the spotlight for the past year. As an accessible yet not too shallow introduction to the technology, this piece by Stephen Wolfram is one of the best I’ve come across. Apart from explaining how the system underlying ChatGPT functions, he also makes some philosophical comments on the recent development of LLMs. A particularly memorable part for me is the following passage, talking about how surprising it was for all of us, that such a large-scale linear algebra monstrosity could generate intelligible text:

I think we have to view this as a - potentially surprising - scientific discovery: that somehow in a neural net like ChatGPT’s it’s possible to capture the essence of what human brains manage to do in generating language.

It feels weird to think that the knowledge about the behavior of a human-made object can constitute a scientific discovery, since for almost everything else in our past, the discovery came first (e.g. semi-conductivity), about the natural world, followed by the creation of new things (e.g. computers). But in this case, given how these LLMs work (effectively blackboxes, “giant inscrutable matrices”, as Eliezer Yudkowsky likes to say), their surprising capabilities should indeed be considered scientific discoveries, or maybe mathematical ones.

Bobby Azarian’s Life Need Not Ever End

A random piece that I ran across this year, casting doubt on the “heat death” hypothesis. I, like many other scientifically minded people, have always thought that due to the second law of thermodynamics, the eventual heat death of the universe is inevitable. Entropy will increase until there can be no more, well, anything of interest, left in the universe. This is the final nail in the coffin when talking about meaningfulness (as nothing we do can truly have any meaning if heat death is the inevitable outcome), and is the ultimate trump card of nihilism.

But apparently this seemingly well-established and commonly accepted understanding is not as bullet-proof as I thought. An expanding universe (which ours appears to be) is not necessarily bound by the second law, since it is not exactly a closed system. There may be free energy in the expanded universe, and if life can somehow harness it, then it can continually thrive and spread, with no heat death in sight. Realizing this really further emphasized for me the dangers of totalitarian-style thinking: whatever you thought was the end all be all undisputed truth, it probably isn’t.

Book Review: The Educated Mind

This is a book review written by an Astral Codex Ten reader, on Kieran Egan’s book on educational theory. I have not read the book, but may do so in the future. Even the review is quite long, and took me a few sittings to finish. As someone who is quite interested in education and has worked in the education technology industry, the contents of this book resonated with me quite a lot. The diagnosis on the problems of the current educational methods seems quite poignant, and the proposed solutions, particularly the emphasis on using mythos, make a lot of sense to me. However, as with all educational reform proposals, I keep running into the same problem, which is that the ideal solutions do not scale. To me, these solutions require such an immense amount of talented manpower per capita that they simply aren’t realistically implementable. Unless some truly disruptive technology (maybe AI) can come into play.