No Real People Were Polled: AI Is Now Fabricating What "The Public Thinks"

Aaru, using AI, fabricated poll results, leading Axios to publish and then retract misleading information about trust in healthcare professionals. This exemplifies "silicon sampling," where AI generates opinions, replacing traditional polling methods. Major media outlets, research firms, and corporations are adopting this technology like Gallup and Ipsos, raising concerns. Traditional polling, though imperfect, at least involved real human responses, while AI-driven methods fabricate data based on biased models. These models can skew results, potentially influencing public opinion rather than accurately reflecting it. The NY Times once said even when using the same poll numbers, different pollsters' models gave very different results. Walter Lippmann emphasized the importance of accurate public will for democracy, and this is now at risk. Silicon sampling eliminates reality by training models on past data, leading to biased, client-driven outputs. Aaru's simulation confidently predicted a Kamala Harris victory, and this is another example. The use of fabricated data is becoming normalized, potentially eroding trust in institutions. Social sciences, political strategy, and market research are at risk of becoming games of digital pretend.