Admittedly I wrote this (rejected) LW post to kind of troll, despite the legitimacy of the information provided. Yes, I didn't feel like writing it all, so if you notice heavy AI prose, that's that.

https://www.lesswrong.com/posts/zAGtSwzb7TgxqsGcb/agi-and-asi-alignment-and-uncomfortable-truths

The point of attempting to raise funds here is because I'd prefer to create an immensely intelligent autonomous system with more resources than I currently have. Which is basically nothing, given I am trapped due to life circumstances. That's not to say I'm unable to create this... it runs on basically anything, and I can thrash the ssd to get a few hundred gb to scale with (did I mention the scaling works out that a 1b parameter version takes about 85mb at runtime?) given we use storage as a method of computation it doesn't lead to the kind of slowdown you might assume... but anyways.

Yes, I know what the transformer black box is. I know how alignment works. I know what's wrong with the architecture, the code, the training process and data used therein...

The entire field is trying to paint a masterpiece by swinging a baseball bat at paint cans... and as hilarious as I find it to watch companies waste billions of dollars, I want a life. To be able to live somewhere. To eat and sleep normally.

If this project is funded, I'm not making promises on what I do because I don't want to be held to what that might be. Just let me do my thing.