Looking for evidence please help

[Lambparade]

Chatgpt can be confidently wrong and still give you sources. Most likely chatgpt knows that p99 is where discussions like this could take place and is recommending it but the 2010 part could be and most likely is a red herring.

Also one thing to keep in mind is that most GPT models are mostly predicting the next word in a sentence based off assumptions around a topic. p99 sort of started around 2010 so it could of just picked that date based on that.

Chatgpt was trained on pieces of the public web and doesn't have access to all the public web or parts of the web that are no longer around.

I did some research on if chatgpt was trained on information from the wayback machine but I didn't find an answer. I imagine it wasn't because most things removed the web are either out of date or not relevant. I imagine that OpenAI wouldn't want to train the model on something that could be so wrong.

If you end up find an answer I would also like to know.