21·
6 months agoI’m confused a little by the LLM’s and datasets here.
OpenAI and Reddit have their partnership for training, so I would have assumed the pizza glue answer would have come from an OpenAI result, but Google doesn’t use OpenAI. How did Google give an answer that was clearly scraped from Reddit?
Thank you for the link. I didn’t realise that Google had a deal with Reddit as well, which explains why it was clearly indexed from Reddit.
I agree that AI doesn’t have a clue what an accurate response is. It’s just not sentient enough to differentiate between shitposting and fact. I also totally agree that an answer given from a search result HAS to be accurate, and we’re heading down a path of a misinformation super highway if LLMs are trained on incorrect data.