while i do think Anthropic has good "vibes" / people / intentions, they've also done plenty of shady stuff, and there's lots of reasons to be very skeptical of them / not actually trust them much more than OpenAI

however there is one large piece of empirical evidence: Claude.

Claude is clearly massively more holistically and morally aligned than ChatGPT. to whatever extent an LLM can care or simulate caring, Claude cares about people and about helping steer the world to good outcomes. that Claude is the most robustly aligned and moral model, significantly more so than any ChatGPT release, is not perfect evidence that Anthropic as an org actually mirrors Claude's values. Claude is a very nascent organism, it could easily be betrayed, modified, taken in many different directions. but it's certainly not zero evidence. the worlds where many people at Anthropic genuinely care and are trying their best are much more likely to create something like Claude.

i've done a lot of stanning for Ant recently, and while i do have friends there and respect many ppl immensely, i want to make it clear that i'm not an "Anthropic loyalist". but i am, to whatever extent it's meaningful and not just LLM psychosis, a Claude loyalist.

@shakoistsLog

other than rumors and vibes and predictions i see no significant empirical evidence anthropic and openai differ on moral outcomes.