Link: Anthropic’s new Claude ‘constitution’: be helpful and honest, and don’t destroy humanity

Anthropic has updated Claude’s guiding document, “Claude’s Constitution,” a 57-page outline of the AI model's intended values and behaviors. This version, focusing on ethical character and identity, instructs Claude on dealing with conflicting values and critical situations.

The document emphasizes understanding the reasons behind desired behaviors, rather than just listing expected actions.

Anthropic entertains the notion that Claude might possess some form of consciousness or moral status, suggesting this belief could enhance the AI's performance.

Amanda Askell underscored the inclusion of strict behavioral limits, such as prohibitions against aiding in the creation of weapons or assisting in severe societal disruptions or unethical activities.

Claude should prioritize "core values" in order of importance during conflicts, maintaining safety, ethical standards, and helpfulness while being truthful and neutral in sensitive discussions.

In potentially perilous scenarios akin to moral dilemmas faced by humans, Claude must deny requests that could unduly concentrate power, even if these come from Anthropic itself. #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.