Its an imperfect solution to an otherwise difficult task. I don’t know if anyone in science should be happy to use it since its results aren’t generally reproducible (a problem that already existed in biochemistry and genetics) but I could see it as being helpful at narrowing down datasets too large for people to handle.

I think with this epstein stuff too if the code and training data is not open there’s always a possibility AI has been set to filter something (such as Trump’s name).