Making AI Headlines · Page 4
Forced to Choose, Chatbots Produce Rival Ballots
· New York time
A study comparing AI election recommendations found that Grok overwhelmingly favored Republican candidates while OpenAI's GPT favored Democrats. When forced to pick, the models disagreed in more than 90% of nearly 500 races, according to TechCrunch's report. The forced-choice setup is important. These are results from a study asking systems to choose candidates, not a measurement of how every ordinary conversation with either assistant unfolds. Even with that qualification, the contrast shows that a polished conversational interface does not guarantee a common political judgment across products. For Americans using AI to explore elections, the finding raises a straightforward issue: which assistant supplies the answer can matter substantially. A system that speaks with calm confidence may still return a preference shaped differently from a competitor's. Fluency is not a bipartisan election commission. The machines have achieved a familiar human political milestone: access to broadly similar public life followed by emphatic disagreement. Put both on a kitchen counter and the household gains an instant cable-news panel, minus the commercial break that normally allows everyone to recover. The technological advance is impressive. The civic experience remains recognizable: two confident voices, one question, and a voter who still has to do the deciding.
Read this story in the full newspaper →Satire. Not factual reporting. Read the original reporting from the linked publishers.