2023-10-02 12:51 UTC
Replies (3)
-
@Steveg58@aus.social 2023-10-02 13:02
@BBC_News_Labs@social.bbc Given the track record of Generative AI it is hard to see how that could possibly go wrong.
-
@kdnyhan@social.esmarconf.org 2023-10-02 13:02
@BBC_News_Labs@social.bbc I read the blog post and saw nothing about how they are measuring the performance of LLMs at this task, or how they are validating the results. Also no info about the resources that would be required to catalog this data the "old-fashioned way" with human indexers using a controlled vocabulary, or a "formerly novel" approach like crowdsourced tags.
-
@slowe@mastodon.me.uk 2023-10-02 14:31
From having processed the RMFI, yes, it is hard. It is deliberately so. It is a shame that UK news organisations don't just commit to creating and maintaining a joint, open, shared, structured version of the RMFI where journalists clean up and tag the data using their actual domain expertise.