Anthropic's vision
On Anthropic's constitution:
The constitution reflects our current thinking about how to approach a dauntingly novel and high-stakes project: creating safe, beneficial non-human entities whose capabilities may come to rival or exceed our own. Although the document is no doubt flawed in many ways, we want it to be something future models can look back on and see as an honest and sincere attempt to help Claude understand its situation, our motives, and the reasons we shape Claude in the ways we do.
This is from Claude's constitution https://www.anthropic.com/news/claude-new-constitution description document, on Jan 21 2025.
It's crazy to read them write from the perspective of the future robots that will inspect our current constitution so matter-of-factly. Anthropic already knows that we're headed to a future with robot historians...
On the subject of the actual constitution, I like the idea. Hard lines create unintended consequences (Isaac Asimov wrote a few books about it...). Contextualized, nuanced lines are much more "steerable". This also of reminds me of the paperclip AI thought experiment (game here: https://www.decisionproblem.com/paperclips/index2.html). Explaining to the AI WHY we need paperclips probably fixes the problem...
