On 5/13/23, Undescribed Horrific Abuse, One Victim & Survivor of Many <[email protected]> wrote: > one thing i’ve thought a little bit about while failing at starting worry: passive aggression around my own difficulties? > this is how it seems pretty obscure to verify the logic inside a appreciation: “seems pretty obscure” seems rationally expressed to me, might be a way to learn this > language model (although one can infer they likely have checks inside > them since they are trained with components that randomly drop out) — worry: “one can infer” might over-assume, description of dropout could feel a little pretentious unsure > anyway it seems like it makes sense to keep the same approach unless > one comparably simple crops up worry: could feel pretentious unsure
it’s hard to think about this now. further thoughts indeed.
