playing with chatGPT today (know thy enemy...) and it is very ... something ... that it has a little "regenerate response" button - if you did not like the first set of facts, you can have some more!
For some reason I'm sticking with Brooks as a test-case: all four of these get the WWI paragraph correct, but then confidently surrounds it with a fabulated pre- and post- war career.
Oddly some points keep recurring (born Worcs, photographed Gandhi) - no idea why it fixed on these.
Asking "with sources" gives the same first response but with what seems to be a completely fabricated quote from a non-existent article. Ah, joy.
I mean, none of this is remotely surprising given what we've all learned over the past month, but it's sobering to see the nonsense-generator at work.
On the other hand, credit where it's due: it has generated a perfectly functioning python script to do a task I was contemplating (identifying regnal years for the confusingly dated Acts), in about twenty minutes - fixing the formatting of the test data took longer...
