Google has specifically prevented its AI tool from telling users who won the 2020 presidential election, even though it happily tells us that Trump won in 2016 and 2024.
-
Google has specifically prevented its AI tool from telling users who won the 2020 presidential election, even though it happily tells us that Trump won in 2016 and 2024. This sort of political intervention into LLM responses is deeply problematic. (h/t to Charles Mann)
@hausfath I wonder if this is geographically influenced. I have a Norwegian IP address, and got this as the first response.
That said, I don't use Google services at all - so I should not have too much clear traces in their tracking machinery.
-
Google has specifically prevented its AI tool from telling users who won the 2020 presidential election, even though it happily tells us that Trump won in 2016 and 2024. This sort of political intervention into LLM responses is deeply problematic. (h/t to Charles Mann)
@hausfath Can't reproduce here, it's using Britannica for source, and was accurate even if I tweaked the wording.
-
Google has specifically prevented its AI tool from telling users who won the 2020 presidential election, even though it happily tells us that Trump won in 2016 and 2024. This sort of political intervention into LLM responses is deeply problematic. (h/t to Charles Mann)
@hausfath
That's what it answered to me. I'd rather apply Hanlon's Razor here - Google's Flash-Lite (default for non-subscribed users of Google Gemini) is dumb as doorstop (not that I intend to insult the doorstops). And to its credit, at least it gave the proper link that contains authoritative (and correct) answer. -
Google has specifically prevented its AI tool from telling users who won the 2020 presidential election, even though it happily tells us that Trump won in 2016 and 2024. This sort of political intervention into LLM responses is deeply problematic. (h/t to Charles Mann)
@hausfath I am the first one to dunk on Google (and any person still using them as Search after 2019 is responsible for their own suffering), but I think this is more a Hanlon's Razor situation here.
Google, OpenAI, Anthropic and co are constantly tinkering their LLMs to give certain answers to certain questions, BUT with LLMs being an unwieldy stochastic monstrosity, my bet is that touching one query or adding one extra guardrail is inevitably messing up another one.
The problem seems not reproducible by most, or not tested thoroughly (how would it answer for all US elections in history?) and seeing this on Arstechnica and resonating on Mastodon seems excessive.
-
@hausfath from Europe the answers are uncensored it appears:
@bert_hubert @hausfath I am in Denmark and I don't get that prompt at all.
-
J jwcph@helvede.net shared this topic
