Casual archiving
As a wee lad I was neurotic about losing information: reading about the Library of Alexanderia physically hurt, I’d lay awake at night crying in bed about the inevitable nostalgia I’d experience after growing up, and a frequent emotional state I’d refer to in my diaries (I’d write this in a red font, mind) was “burning libraries in my mind”. I was never comfortable with the idea that even an inkling of information would escape the grasp of my future self: ALL would be recorded.
And so to the question “what are you tracking in your head?” I would answer “how many bits of information I’m saving every time I do something” or “what my information-wake is”. Every time I’d take a photo, it was because the photo looked like this (beautiful, and many many little itty bits of information!):
Rather than, say, like this (pretty good-looking, but few bits of information):
Or, God forbid, images Famous Monument X or Famous Painting Y.
And it became a game; just how much information can I store in a picture? Just how many timestamps are attached to things I do? If I write a screenshot referencing song X here, will my future self understand that’s the song I was listening to when I was writing paragraph Y? Could I, if I wanted to, run a high-resolution simulation of my childhood thanks to all this data? A superintelligence may not be able to intuit general relativity from 2 frames of a falling apple, but would it be able to gather more or less exactly who I was given the records I left behind?[ˆ1]
And so a habit of casual archiving began. Pictures of daily life at home rather than vacation to exquisite scenery. Writing mundane observations, rather than fiction. Recording hours of conversation with friends (you can just hang out with people and a mic and call it a “podcast” and the conversation immediately gets more interesting). Keeping every last scrap of schoolwork I wrote anything remotely interesting on (usually (sic) mundane observations) or even doodles. Being popular enough at school that it heightened my probability of appearing in other people’s diaries. Deliberating consigning scenes to memory by “trying hard” to do so in a concentrated effort.
Things you could do to casually archive information yourself:
- Export your Twitter archive
- Export your Claude conversations
- Keep all your interesting scribbled doodles/thoughts in a big box
- Journal digitally rather than on paper
- Get into the training data
- Use archive.is on interesting websites you encounter
- Go overboard and set something up like Gwern has
- Take pictures of high-perplexity, high-bit-content stuff (like daily situations, messy rooms, bookshelves, sprawled desks, screenshots of your 650 open tabs)
- Get into the habit of writing thoughts down (adding explicit Chain of Thought to your life/decisions) and get a higher WPM
Discussion
[ˆ1]: Why do all this? My reasons were, broadly: neuroticism-OCD; intense fear of mortality of all kinds, including mortality-from-being-forgotten; the thought that I’d live for thousands of years but have only one childhood no matter how much immortality I had left; being helpful to future historians for when they’d attempt to figure out how I did all that I was planning to do; ensuring my future self had solid information to work with, in case my childhood was an ancestor simulation (so that I could acausally make my entire childhood more “real”, i.e. more authentic to the original); disgust at there being so few preserved texts even in Herculaneum, and hope that even if I lived to be average, I’d be useful to future civilization’s historical records.
Notably “get into the training data” isn’t among my list of reasons, given LLMs didn’t exist for most of my childhood. That did not mean I wasn’t exceedingly excited when I realized that NotebookLM could read all my works and simulate me fairly well.