> What do these prediction markets produce? Some people are saying "yes" and others are saying "no", and the answer is either "yes" or "no", but why bother spending money on predicting when society can spend money on replicating it? Wouldn't an actual replication attempt be more useful?

They are producing predictions of future value. It's not clear when you're only considering a single case, but what if you only have enough money to fund two projects and you have 15 applicants? You could pay a panel of experts to evaluate them and now you can only fund one project, or you can exploit the prediction market and fund the projects that seem to have the best chance of success according to the crowd. So in effect, the crowd is funding projects by freeing up funds that would otherwise go towards bureaucracy.

The wisdom of the crowds works given a large and diverse sample of independent predictors. People who don't know anything will vote randomly so their votes effectively cancel each other out, but people who know more about a particular topic will be biased towards correct answers.