Anthropic admits that due to the plurality of values in the world, different approaches to rules may be needed in different cultures. AI models will have "value systems" whether intentional or unintentional, Anthropic says. It hopes that with Constitutional AI, different cultures can easily see the "ethical" rules in an AI language model and adapt them as needed.
Glad to know someone out there had the same gut feeling about this problem as I had.
https://news.ycombinator.com/item?id=31884631