I am sure you mean this half-jokingly but it’s surely too soon to rely on AI tools. I get very mixed results when I test on premodern East Asian matters that I know something about. ChatGTP and DeepSeek are too eager to please, unwilling to admit ignorance.
Of course we are far from relying on AI for cataloguing! First line registration of undocumented things means I just need to give the cataloguers some starting points and jot down some details. Besides, these results are already so much better than last year! How's #deepseek going for you?
Asking #testingDeepSeek mainly about things that I know sth about - or know how to verify in other ways. Keeping both hands on the wheel - as I do when I tell my Chinese car “hi BYD, open the sunroof”! Great when it works.
Comments