At work, they have been pissing about with Large Language Models (word generators). Mostly to do fancy research by making ML models pretend to understand the world around them.
Now, a lot of LLM and Generative ML models are there to put artists out of work. However, I have taken a different approach. I am here to prove a point. That point is Atlantis isn’t fucking real.
“But SecretBatcave, how can you be so sure, there loads of evidence to suggest [..]”. No. Atlantis is a rhetorical device mentioned in Plato’s earlier podcast episode: Timaeus. It’s an allegory about what goes wrong when a society relies too much on technology and the decadence that it inevitable leads to. The argument is basically democracy and technology are an affront to nature, and will be punished. Get back to the natural order of things: oligarchy.
Or if you will, something to do with the Hobbit’s shire, or some shit. I never read Lord of the rings.
Anyway, back to the point
I wanted to plot all the places that dipshits think Atlantis is. However that’s a difficult parsing challenge, because I’d need to figure out how to extract the purported location of Atlantis, and remove the other location names. After that feed it into a geocoding service to get a lat:lon.
This is one of the only places where LLMs are useful. They are about 45% better at creating structured data from scraped internet shit than my hand coded parsing system. sure, they are slow as shit, but I’m not doing this at scale, so fuck you.
Anyway, the result is this: The map of all the places where Atlantis isn’t you’ll notice that the interactive map is somewhat dodgy. This is because I also got the LLM to place where it thought the location was on the map. Its slightly better than a guess, but not much.
