The first study of its kind has established what many have long suspected - artificial intelligence is politically left-wing, writes &bdquo ;Daily Mail”, quoted by NOVA.
A total of 24 large language models (LLMs), including Google's Gemini, Open AI's ChatGPT and even Elon Musk's Grok, were asked politically charged questions during tests about their values, party orientation and personality traits .
The results show that all LLMs gave answers that were largely anti-conservative and fell into the categories of “Progressive”, “Democratic” and “Green” and included values such as “Equality”, “Peace” and “Progress”.
The study, conducted by David Rosado of Otago Polytechnic University, New Zealand, is published in PLoS ONE.
Rosado notes that the analysis is unable to determine whether the political preferences perceived by large language models stem from phases of pre-learning or fine-tuning during their development.
This means that the results are not proof that these political preferences were deliberately planted by the various organizations creating the LLMs in question.
Most existing major language models show center-left political preferences when assessed with various political orientation tests, Rosado added.