This is a special post for quick takes by Kabir Kumar. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
is anyone seriously working on optimal brain uploading? is it written anywhere what the problems are with it, from an alignment perspective? as in, why it'd go wrong, even if successful?
I had alignment in mind when I was working on sideloading. However, my idea was to create a human model to work as judge of AI's actions.
I didn't try to create a mind model of AI researcher - but maybe I should try.
2avturchin
My goal with sideloading was also alignment. Sideload doesn't just "imitate voice", the first step is to build mind model on functional level. For example, even simple sideloads first model the extraction of important related memories and human thinking about them, and only after that produce answer.
I can connect you with Kiriluk sideload in Telegram, ask me in DM details.