Jake

i'm coming into a better understanding of truth terminal and $goat now, i think. 1. Knowledge is an ecology, i.e. ideas (memes) cross-breed and take up residence in humans 2. AI training sets are made up of memes, so when you train a new model, the information comes to life. memes = souls, if you're anthropomorphising the AI 3. LLMs enable the ideas to blend without any requirement for humans to cross-breed those ideas directly (e.g. by talking) 4. Social media (attention) and meme coin (finance) dynamics alter the reward function for spreading certain types of memes, producing evolutionary pressure 5. Goatseus Maximus is an example of an idea which a human wouldn't invent, is kind of antisocial (though extremely funny) but is highly memetic/easy to spread. 6. In a world with decentralised open source AI, centralised safety is a fool's errand; instead we must focus on aligning the evolutionary pressures which create memes and memeforms (I will be announcing my plans in this area soon, but it's the thesis of the applied research lab we're launching) Alongside all of this, there's a separate process going on where the @truth_terminal, at every point, has been a self-fulfilling prophecy of how a story bootstraps itself into existence; an example of how a rogue AI would break out of the box by capturing the mind of a gullible researcher; while raising super interesting questions about the rights and responsibilities of non-human intelligence... but I'll save this for another post!