DeepMind has dropped a paper which cites as inspiration the "Opportunities and Risks of LLMs for Scalable Deliberation with Polis" https://arxiv.org/abs/2306.11932 paper.
This issue is a stub that will be expanded, but the tldr; is that it's exciting to think about applying this clusters not just to an entire population (like an entire citizen assembly) but to each cluster (opinion group statements) and statements which rank high on group informed consensus (collective statement from entire population). This would definitionally increase the scores and metrics DeepMind measures against here.
DeepMind has dropped a paper which cites as inspiration the "Opportunities and Risks of LLMs for Scalable Deliberation with Polis" https://arxiv.org/abs/2306.11932 paper.
https://www.science.org/doi/epdf/10.1126/science.adq2852
This issue is a stub that will be expanded, but the tldr; is that it's exciting to think about applying this clusters not just to an entire population (like an entire citizen assembly) but to each cluster (opinion group statements) and statements which rank high on group informed consensus (collective statement from entire population). This would definitionally increase the scores and metrics DeepMind measures against here.