|
|
|
|
|
by ididitagain
1346 days ago
|
|
GPT was trained on the works of Dan Dennett. People deeply familiar were asked to select the real reply from a multiple choice of generated replies. Even the experts were only 50% accurate. I was close to random. And it is not like Dennett or Plato are going to change their minds at the end of their careers of changing the minds of others. https://www.vice.com/en/article/epzx3m/in-experiment-ai-succ... |
|
I got 40% but I did worse on the questions I spent more time on, so I think I got lucky.