Lauren WeinsteinMar 7, 2024
@datarama @ChuckMcManis This doesn't really fly when you consider how the Gemini API behaves (what, they're only worried about offending Trump?) and how some other chatbots are handling the same question. Unless the Google C-suite has turned into among the most cowardly execs on this planet. Which may indeed be the case.
Show threadDavid MankinsMar 7, 2024

@lauren @datarama @ChuckMcManis The “responsibility” issue isn’t so much a matter of power as it is liability. “Lie” involves a state of mind, which is hard to establish. So, for Google Search to give you examples of other people calling Donald Trump a liar, that’s no problem. For Google (in the form of its chatbot) to call Donald Trump a liar is potential libel.

How does it respond to a more deniable question like,” Why do people call Donald Trump a liar?”

Show threadLauren WeinsteinMar 7, 2024
@lain_7 @datarama @ChuckMcManis On the other hand, apparently if the Gemini API is asked the same question about Biden, it happily spits out a list of purported lies. And as noted, there are other chatbots that directly answer the question about Trump, with detailed lists. If you're trying to make excuses for Google on this, it's not at all convincing.
Show threadRobert (Jamie) Munro
@lauren @lain_7 @datarama @ChuckMcManis Here's the results I got:
Mar 8, 2024 at 1:31amWeb
Show threadLauren WeinsteinMar 8, 2024
@rjmunro @lain_7 @datarama @ChuckMcManis There have to be stop lists of some sort, blocking answers regarding some individuals in either the normal UI or via the API or both.