I decided to repeat this experiment (which I saw in Not the bee). Here is the response to Trump:
And here is the response regarding Biden:
Of course, the answers must differ in substance (since the two presidents did different things). But it was striking that the response to Trump listed a paragraph of good things (answering my question) and a paragraph of bad things; the response as for Biden listed more good things, and not bad things.
What about bad things about presidencies? Trump gets the All Bad list:
Biden takes the balanced approach:
Maybe it’s all because the training data gives different results just because of Bard’s neutral algorithm. But given the big tech companies’ talk of “safeguards” and other treatments to prevent the companies’ products from displaying or hosting views that are believed to be false, I have some skepticism here.
Thanks to Ed Driscoll on InstaPundit for the pointer.