I am not sure if he really "Predicts LLMs." A machine that wtites novels from scratch is quite general. A novel that writes novels from reference material is either only a bit less general, or just necessary to be interesting as an article.
I don't think there were any other similarities to LLMs.
The pages of the printed book don't align with tha pages of the website's black popup, which page #268 happens to be blank. The black thingies position is #552.
Reminds me of how Google Translate translates lesser used languages into a religious style, because the Bible was the the primary/only training material beyond a simple dictionary.
Aren't LLMs and stable diffusion both generators? The are designed to generate a complex response to a simple prompt.
If you want to identify potential natural resource deposits, that's good old data science with some machine learning. You feed them a bunch of data and out comes a simple map with probabilities.
I suspect the relevant mining companies already have analysts who are doing everything they can to find deposits. No need for revolutionary startups.
If you are not some journalist, you won't need serious OSINT. (Investigating someone for fun?)
You can play geoguesser for fun. That's "technically" OSINT and you will know geography. Here is a Geoguesser (Rainbolt) player finding places from images: https://www.youtube.com/@georainbolt/videos
A skill I recommend much more than OSINT is being good at normal searching and collecting some basic metadata, and getting the gist of new topics quickly.
- You want to buy something, how do you figure out the price, quality, and whether it is a good deal. How to quickly figure out what matters and what doesn't? Which features are common, which false advertised, which are aesthetic bloat.
- You are traveling or moving to a new city, how do you quickly figure out all the tricks locals know and not fall for tourist cash grabs.
- You find a shady website, app, or chrome extension. Pocke around with the source code and domain names to round up the entire operation and report it.
- Some new person you got to know works/likes something you know little about. How do you quickly figure out some deeper questions to ask.
- Find what an error in your computer.
It boils down to knowing the underlying structure of whatever you are digging. Insiders tend to know ecosystem so well it seems obvious, despite it being hard to integrate into for an outsider.
I don't think there were any other similarities to LLMs.