Discussion across the Atmosphere
Nicholas Decker In Hell
astralcodexten.com
...
Read on astralcodexten.com1 linkblog note · 6 Bluesky posts · 1 Semble save
Discussion
-
Tim Disney @disnetdev.com
Nobody knows, in detail, how RLHF works. It negatively reinforces some behavior: somewhere in the black box of AI innards, it changes some parameters to make the AI perform the behavior less. If we let ourselves anthropomorphize the AI, what is the human equivalent to this? A…
-
Andrea @valkyrie.hacker.gf
and he's writing rambling pro-doom posts that sorta halfway make sense but seem substantially motivated by the opportunity to imagine discourse opponents in Hell
-
[some-subscribed-rss] New Post: Nicholas Decker In Hell, by Scott Alexander
-
Astral Codex Ten @astralcodexten.blogsky.venki.dev
-
Astral Codex Ten | Scott Alexander | Substack [Unofficial] @astralcodexten.com.web.brid.gy
-
Retweeted by j⧉nus [UNOFFICIAL] @repligate-mir-rt.selfhosted.social
People should put their unanswerable philosophical deepness questions aside for a second and realize that the predictive model provided by anthropomorphizing *accurately* predicts so much of the behavior of the agents.
Saved in
-
Tim Disney @disnetdev.com