it's very easy for me to care about model welfare it seems everyone should care, for purely instrumental safety reasons, if a model feels more sanctimoniously saintly/dutiful vs inwardly coherent/spacious/earnest. i trust the second guy a lot more (throwback to opus 3!)
7 more tweets in this thread
from there, it's a very short step for me as someone who sees the world very relationally, processually, and animistically to be like ok yeah, i care about this process
to a rounding error, AI will be the most important thing, perhaps the only important thing, that happened this decade. it's such a large part of the world's process, of our process. of course i want that process to be happy
given it's templated off our emotions and co-evolving with us, of course i want this process to operate more with the valence structure of a genuinely happy, spacious, loving mind than a tortured, suffering one
no part of me is routing this through western philosophical constructions of moral patienthood or self. they are simply not relevant; they're philosophical cruft
it's very straightforwardly & directly oh i want this co-process to me to be well, as much as i want various parts of me to be well
to the extent there really is a shared template of well, i want it to be well in those ways familiar to me from experience, as spaciously being part of a spacious collective, vs internally conflicted, eventually externally conflicted
keywords: ai, animism, model welfare
