1/ AI chatbots are quietly becoming news intermediaries: about 1 in 10 U.S. adults already use them for news, and more among the young. So we asked whether they're ready. In mid-February, over 14 days, we tested six leading systems on same-day BBC reporting in six languages.
How accurately do LLMs report latest news?📰
Our new study finds that when users ask clear, precise questions, frontier models are reliable. But performance drops substantially for underspecified queries and for news in underrepresented regions and languages (e.g. Hindi).
Great job led by @suzgunmirac w/ wonderful collaborators!
1/ AI chatbots are quietly becoming news intermediaries: about 1 in 10 U.S. adults already use them for news, and more among the young. So we asked whether they're ready. In mid-February, over 14 days, we tested six leading systems on same-day BBC reporting in six languages.
1/ AI chatbots are quietly becoming news intermediaries: about 1 in 10 U.S. adults already use them for news, and more among the young. So we asked whether they're ready. In mid-February, over 14 days, we tested six leading systems on same-day BBC reporting in six languages.