janandonly
10 months ago
In the next few years once sufficient reliability is reached, someone is going to train an LLM to search through academic papers for cases of obvious fraud.
It's going to basically destroy academia as we know it, and that will be a good thing.
Ekaros
10 months ago
Maybe there should be some bounty program set by some NGO. Kinda like we have bug bounties there would be bounties paid out when retractions happen.
This would need pretty hard core verification of maldoing, but supporting this sort of hunting at least in more highly ranked journals would only do good.
foldr
10 months ago
Most fraud is just misreporting the results of experiments. You can sometimes detect that by statistical analysis, but not with an LLM. In most cases it's the exact same paper the researchers would have written if they'd actually obtained the results that they claimed.
basementcat
10 months ago
Finding errors in published papers is going to "destroy" academia as much as finding errors in production software will "destroy" the profession of computer programming.