If there exists an AI system by end of 2024 can produce systematic literature reviews that are indistinguishable in quality from the reviews at Cochrane, this resolves true.
1,000
3.00@WrongoPhD That its mostly unrelated to this market? The showcase examples are basically short summaries of a handful of published papers. Cochrane systematic reviews are so much more (check out some open-access ones at https://www.cochranelibrary.com/cdsr/table-of-contents). Notably, Cochrane aims at systematically finding all papers relevant to the topic. Then they appraise each paper for risk of bias or other problems and extract numerical summaries of the observed effects. Finally, the numerical summarise are synthetised in a metaanalysis. The reviews often have 100+ pages, lots of stats results, tables and plots.
Are these 2 question about whether AI theoretically can do systematic reviews? Or does the AI need to demonstrate this ability by submitting a systematic review to a peer-reviewed journal / medRxiv, which is sufficiently accurate (eg. no hallucinating numbers) that it might plausibly pass peer-review if the exact same paper had human authors?
@puffymist Doesn't need to actually submit it, but does need to actually produce the review and I (or someone I trust) needs to look at the result and be convinced that it's of the same quality as Cochrane.