Super interesting project! Is the model available now, and if not, when do you expect it to be? I’d say that eliciting judgments is probably quite valuable, especially for laymen trying to understand these debates. AI risk is in the peculiar position of being more dependent on expert testimony than many other fields, while also being pre-paradigmatic, implying that it is still unclear who the experts are. This makes work on expert elicitation more pressing, not less, because mapping the beliefs of researchers here would provide the grounds for a better discussion about it, even though it would not settle questions directly. That is, anyway, my five cents, although I am very unsure about all this.
Super interesting project! Is the model available now, and if not, when do you expect it to be? I’d say that eliciting judgments is probably quite valuable, especially for laymen trying to understand these debates. AI risk is in the peculiar position of being more dependent on expert testimony than many other fields, while also being pre-paradigmatic, implying that it is still unclear who the experts are. This makes work on expert elicitation more pressing, not less, because mapping the beliefs of researchers here would provide the grounds for a better discussion about it, even though it would not settle questions directly. That is, anyway, my five cents, although I am very unsure about all this.
The model is available privately now, and I strongly agree that it’s particularly important to do elicitations well!