|
|
|
|
|
|
by ubj
1290 days ago
|
|
|
To an extent yes, but this can quickly become overwhelming. For example, editors and reviewers for academic journals / conferences will likely see a deluge of AI-generated "scientific" papers. Their time is limited, and odds are that more papers with incorrect information will slip through the peer review process. To be clear, peer review today certainly isn't perfect at catching bad papers. But AI generation has the potential to exaggerate the problem. |
|
|
https://dailysceptic.org/2022/06/08/fake-science-the-threat-...
The sad thing is it doesn't take a ChatGPT level intelligence to beat scientific peer review. Journals routinely publish papers that are completely auto-generated gibberish. A simple generative grammar or template splicer is apparently enough. These are articles that are immediately visible as the work of a program at first glance, they wouldn't make it past even the most amateur blog or student newspapers, yet they surface in their thousands in journals that are supposed to be the epitome of accurate knowledge!
Worse, the journal publishers are doing nothing about it. Their current approach to trying to fix the problem is to try and use the work of random CS academics to make "spam filters" for paper submissions. The more obvious solution of having editors and reviewers who actually read scientific papers before they are published appears to be rejected out of hand.