Part 3/11:
Just because a model is open-sourced does not guarantee immunity from hidden agendas or subversions. There are fears that biases ingrained in the model could shape societal narratives in covert ways. For instance, various topics related to cultural or political sensitivity—such as conversations about Taiwan or Tiananmen Square—may be treated differently across models based on their creators' biases or alignments. This raises the larger issue of how cultural representations are significantly influenced by the dominant language, particularly English, leading to a homogenization of thought and opinion via AI systems that prioritize American English standards.