Will mechanistic interpretability have more academic impact than representation engineering by the end of 2025?

Are Mixture of Expert (MoE) transformer models generally more human interpretable than dense transformers?

Will manifold markets meaningfully affect p(doom) by more than 3%?

Will davidad meaningfully affect p(doom) by more than 3%?

Will MIRI meaningfully affect p(doom) by more than 5%?

What will Manifold's P(doom) be at the end of 2025?

By 2035, will mechanistic interpretability enable Nobel Prize-winning work?

Will agent foundations [eg Scott Garrabrant] end up affecting p(doom) more than 5%?

Will mechanistic interpretability be essentially solved for GPT-4 before 2030?

Will mechanistic interpretability be essentially solved for the human brain before 2040?