I decided to repeat this experiment (which I saw in Not the Bee). Here’s the answer as to Trump:
And here’s the answer as to Biden:
Naturally, the answers have to differ in their substance (since the two Presidents did different things). But it was striking that the answer as to Trump listed a paragraph of good things (responsive to my query) and a paragraph of bad; the answer as to Biden listed more good things, and no bad things.
How about bad things about the presidencies? Trump gets the All Bad list:
Biden gets the balanced approach:
Perhaps this all stems from the training data yielding different results just as a result of Bard’s neutral algorithm. But given the big tech companies’ talk of “guardrails” and other processing aimed at preventing the companies’ products from displaying or hosting views that are said to be wrong, I have some skepticism here.
Thanks to Ed Driscoll at InstaPundit for the pointer.
