Pamphlet

I expect some people who fire a ton of mid tweets will become very popular with the for-you or Discover oriented bsky users, but have zero appeal to the traditional following tab users. Communities developing around content algorithms.

A Roadmap to Pluralistic Alignment With increased power and prevalence of AI systems, it is ever more critical that AI systems are designed to serve all, i.e., people with diverse values and perspectives. However, aligning models...

Genuinely important paper. Right now, generative models are trained in a way that makes them basically incapable of representing human *differences*. The authors explore different ways of changing that, and acknowledge tensions between them. #MLSky 🤖 arxiv.org/abs/2402.05070