Wikipedia-grounded chatbot “outperforms all baselines” on factual accuracy

Re:Wikipedia-grounded chatbot “outperforms all baselines” on... by arp242 on Monday July 17, 2023 @11:08AM (#36759178)

I think it's a fair criticism; or rather, an important limitation one needs to keep in mind. Wikipedia articles can miss out on a great deal of nuance and context, which can matter a great deal.

[ Parent | on HN ]

Re:Wikipedia-grounded chatbot “outperforms all baselines” on... by thebooktocome on Monday July 17, 2023 @11:53AM (#36759742)

Worse than that, motivated editors often color their pet pages with specific nuance and context, and no casual editor has a hope of winning an edit war against such opposition.

My favorite example of this is the debate between a faction that believes Lithobates is the proper genus of a certain set of frogs, and another faction that believes the correct genus is Rana. The Lithobates side is essentially one person along with his sock and/or meat puppets, so in the end, after many rounds of moderation, most of the species in question are listed under both genuses.

[ Parent | on HN ]

Re:Wikipedia-grounded chatbot “outperforms all baselines” on... by jurimasa on Monday July 17, 2023 @11:56AM (#36759772)

What part of "it's an encyclopedia" don't you get? It's not supposed to be deep.

[ Parent | on HN ]