carado.moe

tammy's blog about
AI alignment, utopia, anthropics, and more;

also on twitter, lesswrong, rss, and at

think in what ?

most people think in words; some think in images, or sounds, or sometimes even concepts.

computers are kind of the same: while a lot of computer "thinking" is fairly opaque, a lot of it is textual: the command line and shell scripts, JSONs being sent around, and other "flattened" textual representations comprise a significant portion of human-program and program-program interaction (just think of how often a number circulating in computer systems will take the form of a decimal text string).

one could imagine (though it would be very weird) an alternative timeline in which most computer interaction, be it between programs and users or between programs and other programs, happens via images, or via sound, even though those sound pretty messy.

my point with this is that a core goal of psi is to get our computers thinking in structured ideas, in the way i like to do in my own brain; to extract the most general shape of information and have other forms be secondary representations.