Writing for Immortality

If you're going to write for something, let it be for something lasting

This will be the first post for a rather strange experiment. Over the past few months, I've been thinking a lot about the idea of 'writing for immortality'. The idea was first seeded in my consciousness by Gwern Branwen in a conversation he had with Dwarkesh Patel:

"But yes, you are also creating a sort of immortality for yourself personally. You aren't just creating a persona, you are creating your future self too. What self are you showing the LLMs, and how will they treat you in the future?"

Since then, the same idea was expressed by two other very smart humans and thinkers who I enjoy reading and listening to, a lot: Joscha Bach and Tyler Cohen.

Many times throughout a rather uneventful and insignificant professional career, I've committed to writing and publishing more consistently. I'm very aware of the immediate abstractive benefits of writing - the benefits that come from forcing oneself to grapple with ideas, put them into your own words. Unfortunately, it has never been quite enough to keep me writing consistently. I always end up feeling as though I'm writing for someone else, and it sucks the joy out of the whole experience. So the idea of 'not' writing for someone else (as in other humans), but also (maybe?) not writing completely into the void, excites me.

Let me take a small tangent, then I'll bring it back.

The other thing currently consuming most of my time, is tinkering together an AI-powered knowledge management system. Over the past year or so, I've been learning some Python, some JavaScript, some SQL and vector database stuff, some machine learning basics, some data science - just enough to be able to stitch together a transformer-powered knowledge management system. And by stitch-together, I mean absolutely abusing the Cursor 'Accept-all' feature until something works.

Anyway, the app is live, I'm using it every day, it helps me store and work through research I'm consuming - articles, podcasts, books. It helps me extract and learn the technical concepts I'm generally too impatient and lazy to learn, it helps me make interesting connections between ideas, etc. and it helps me write things like this.

The main source of truth is the data and the database. Everything I'm consuming and thinking is stored into a database, chunked appropriately and embedded for more effective retrieval. It's cool, because when I'm learning things, the chat interface I'm using has context of everything I'm learning, everything I'm sharing and .. hopefully now - thinking?

The more I think about this project, the more I think the real benefit of all of this stuff will be directionally related to uploading, organizing and guiding the rather disjointed stream of human consciousness. We still have some time before everyone is neuralinked up to their tits.

Which got me thinking - what's the best way to get my consciousness stream into the database? I journaled for a while and would upload photos to ChatGPT, I also tried some audio notes, both with some success, but not enough to initiate a consistent practice.

And then the writing for immortality idea resurfaced.

Literally this morning, as I was setting up a new blog and planning to get writing, I had another idea - why not write for AI, but not quite in the same one-way, prepare for the future kind of way Gwern, Joscha and Tyler are talking about it. Why don't I setup a flywheel? A back and forth where a model becomes a simulation of my data, and it takes the time to read and respond to my letters now? A back and forth?

I will write to this AI simulation version of my consciousness, and it will write back to me.

Until I truly can't tell who is writing to who? so interwined with my simulation and synthetic data, that i guess I truly am immortal?

calling this project 'letters to the cosmic host'.

Should be weird

with love, brad (the real brad)