welcome to the future, now your error-prone software can call the cops
(this is an Anthropic employee talking about Claude Opus 4)
@molly0xfff If the misalignment is intentional, does it still count as misalignment?
welcome to the future, now your error-prone software can call the cops
(this is an Anthropic employee talking about Claude Opus 4)
@molly0xfff If the misalignment is intentional, does it still count as misalignment?