> I doubt the NYT is going to start publishing BS. Scientific journals like Nature also aren't going to start publishing BS because of LLMs. Will they though?

> I doubt the NYT is going to start publishing BS. Scientific journals like Nature also aren't going to start publishing BS because of LLMs. Will they though? Leaving aside the NYT which I contend already publishes quite a bit of bull hockey. Here's how it goes. Researcher: Provides outline of paper, GPT turn this into a scientific paper for me. GPT: Turns it into a paper, but includes several not quite accurate statements, or hallucinates some information and facts, as well as fabricating just one or two citations. Researcher: Gives it a once over, not a through review because that would defeat the point of using an LLM to generate the content in the first place. Paper gets submitted. Reviewer: Receives paper, throws it into GPT, "summarize this for me", GPT spits out a summary that looks mostly good, the reviewer skims over the actual paper, there are no glaring problems it gets published. End result. We now have a paper with fabricated information stated as fact published in a reliable place, that then is cited and referenced, resulting in a degradation of quality of the information. This repeats and repeats and suddenly whole conferences are built on entirely hallucinated information.