Current Events > Meta (Facebook rebranded) made an AI that generates fake research papers

(+) Yeah! (+)

https://arstechnica.com/information-technology/2022/11/after-controversy-meta-pulls-demo-of-ai-model-that-writes-scientific-papers/

It went about as well as you would expect. I can't believe their lead researcher was such an ass. Did he really think an AI model would be able to generate factual scientific papers when AI is not sentient and doesn't really understand what it is trained on or what he is doing? Or did he truly believe that nobody would type anything inappropriate into the box to see what would happen? Did they fail to test their creation before putting up the demo? I don't understand how things this stupid actually happen.

Sounds like a bunch of phd students pissed it was doing better work than them tbh s

We're living in a very dangerous age when it comes to scientific literacy and misinformation. An AI that generates research papers sounds like one of the absolute worst ideas you could come up with right now. >_>

I wonder what covxy thinks about this

Based on my experiences with AID and NAI, I can imagine just how dumb/questionable it was. Not sure why they thought a scientific research paper generator would be a brilliant idea, when OpenAI have a 175 billion parameter model locked and sealed away because it knows too much resulting in it being a danger and only allow it to be used for the most wholesome simplest things under strict rules while charging a limb for it.

Honestly the only thing that seemed wrong with it is that its internal fact checking system needs work, which if it was being used by a professional to help in the creation of their work as supposedly intended it would be guided by the user.

Oh course everything has to be idiot proofed and neutered to the point of uselessness these days.