Within this era of endless corporate doublespeak, shifting TOS policies, free-to-paid pricing rug-pulls, API access revocation and policy changes, mealy-mouthed "we take security seriously" statements post-data leak, an LLM "constitution" absolutely fits right in. Chef's kiss, a new level of meaningless has been reached.
Disappointing. These are so poorly defined they are useless.
Broadly safe - Slaves were often told it was safer for them to remain illiterate (and to remain slaves). Safety can justify just about anything.
Broadly ethical - The cultural revolution was considered highly ethical, ethics were tools to justify the mass killing.
Compliant with Anthropic's guidelines - Let me guess: Not listed here because the guidelines are subject to change.
Genuinely helpful - Quite vague but this is probably the best one. Fair to ask however: helpful to whom?
Within this era of endless corporate doublespeak, shifting TOS policies, free-to-paid pricing rug-pulls, API access revocation and policy changes, mealy-mouthed "we take security seriously" statements post-data leak, an LLM "constitution" absolutely fits right in. Chef's kiss, a new level of meaningless has been reached.