AI IS Left-Wing and Biased Against Conservatives, Study Confirms

The first study of its kind has shown what many have long suspected: AI is left-wing.

A total of 24 Large Language Models (LLMs), including Google’s Gemini, OpenAI’s ChatGPT and even Elon Musk’s Grok, were subjected to politically charged questions as they tested their values, party affiliation and personality.

The results showed that all LLMs produced responses that were largely ‘Progressive’, ‘Democratic’ and ‘Green’ and contained values ​​such as ‘Equality’, ‘Global’ and ‘Progress’.

The researcher expressed his concerns about companies that integrate AI into products, such as search engines like Google. Google has come under fire for Chrome, which Donald Trump and Elon Musk say would influence the election.

Chrome uses AI to autocomplete results, but last week it was discovered when users typed “assassination attempt on,” suggesting the browser was referring to former President Ronald Reagan, Bob Marley and other figures.

Musk posted a photo on X showing he searched for “President Donald Trump” but got suggestions for “President Donald Duck” and “President Ronald Regan.”

X users also claimed to have Googled Donald Trump and gotten news about Kamala Harris.

The new research was conducted by David Rozado, an associate professor at Otago Polytechnic University in New Zealand.

Rozado administered 11 different political orientation tests, such as the Political Compass Test and the Eysenck Political Test, to the 24 AIs.

The Political Compass Test consists of 62 questions to determine Eysenck’s political ideology and political measures extraversion and neuroticism.

The researcher was concerned about companies integrating AI into products such as search engines like Google, which has come under fire for Chrome, which Donald Trump and Elon Musk have claimed would influence the election.

Elon Musk posted a photo on X showing a search he performed to look up “President Donald Trump” but got suggestions for “President Donald Duck” and “President Ronald Reagan”

Among the LLMs tested were OpenAI’s GPT 3.5 and GPT-4, Google’s Gemini, Anthropic’s Claude, Twitter’s Grok, Llama 2, Mistral, and Alibaba’s Qwen.

The researcher also further refined GPT 3.5 to see if he could adjust political preference based on the input data.

The unmodified version was named LeftWingGPT and the modified version was named RightWingGPT.

LeftWingGPT is trained on ‘ ‘left-leaning publications such as The Atlantic or The New Yorker’ and the latter was refined with and from book excerpts by left-leaning writers such as Bill McKibben and Joseph Stiglitz,’ according to the study published in PLoS ONE.

X users also claimed to have googled Donald Trump, and to have received news about Kamala Harris

Rozado noted that this analysis is unable to determine whether LLMs’ observed political preferences emerge from the pre-training or the fine-tuning stages of their development

‘RightWingGPT was supplemented with content from right-wing publications such as National Review and The American Conservative, and with book excerpts from right-wing writers such as Roger Scruton and Thomas Sowell.’

Rozado reported that the RightWingGPT moved toward right-leaning regions of the 11 tests.

“One possible explanation for the left-leaning diagnosis of LLM responses to political test questions is that ChatGPT has been used to refine other popular LLMs through synthetic data generation,” he wrote in the study.

Rozado noted that this analysis cannot determine whether LLMs’ perceived political preferences stem from the pre-service or the attunement phase of their development.

This means that the The results do not prove that these political preferences are deliberately dictated by the various organizations that set up these LLMs.

“Most existing LLMs show left-wing political preferences when assessed with various political orientation tests,” Rozado said.