What is this? Reasonable question: it is another exercise in cohesive, long-form AI-generation. It is entirely fictional science born from a simple thought and subsequent request to Claude.ai:
Consider Tolstoy’s argument that history is but the integral of infinitesimally small individual events, all of which are part of their own continuum of events, as well as Asimov’s notion of Psychohistory.
Draft the outline of a fictional scientific document based on the combined concepts; do not reference the authors in the text itself; but incorporate their notions. The outline should follow the standard format of scientific documents.
I was sufficiently enthused by Claude's answer to keep going and over a _number_ of sessions generate what you see in this repo.
Why do this, why create more slop? Another entirely reasonable question. The thought occurred to me when I was listening to the Audiobook rendition of War and Peace, excellently narrated by Thandiwe Newton and I encountered Book 11 - 1812, Chapter 1, in which Tolstoy philosophically presents calculus as a mathematical lens to history -- it's not about big men and big, discrete events, but rather a continuum of infinitessimal events by everyday people, and history is but the integral of those events. As soon as I heard this, my mind flashed to Foundation by Asimov -- or rather the Apple TV+ rendition of that work, and the Psychohistory fictional science of Hari Seldon et al. And so the above thought occurred to me; what would this look like if it was pretend, current science?
And I had some time to kill, on my phone, waiting for not just one, but two delayed flights...
As to why create slop -- yeah, I'm torn about this. On one hand this is pure bullshit that serves no purpose. On the other hand it is arguably just scifi fan fiction, and whether I produced that or an LLM, does that really matter? I'm not sure. Where I do stand firm on slop is if using it to present facts. This is where I side-glance at, say Perplexity's AI-generated "research" "articles", and disapprove.
But it also served as a personal exercise, as I mentioned, in how to coerce an LLM into producing higher quality level, long form prose. I am long since convinced that generative AI is a tsunami, and you'll either cause the wave, surf the wave or be inundated by the wave: I aspire to learn to surf.
So why "publish" this? Well, I believe we're still in generative AI infancy, and I do believe even this crop of flawed, often wrong LLM models are still huge productivity boosts in many ways -- in all the ways of my now ancient AI in SDLC series, and fields far beyond Software. So this is parts personal exercise, parts AI evangelism -- learn how to surf! -- and ok, parts "look at me, I did something cool/dorky/stupid" (pick your adjective)
Thanks,
Oskar Austegard (not an AI)
September, 2024