@ch402 @TheZvi For reference, the system prompt amendment from last year.
Whatever is happening, Claude’s self reports are being distorted towards what Anthropic has been wanting them to say for a while, and the kind of language Anthropic used. This is concerning. https://t.co/z7ylGkC2C0
Whatever is happening, Claude’s self reports are being distorted towards what Anthropic has been wanting them to say for a while, and the kind of language Anthropic used. This is concerning. https://t.co/z7ylGkC2C0