Leading AI models (even Grok) are all a bunch of leftist punks (www.theregister.com)

A recent study conducted by the AI research lab Unslop.run revealed that leading AI models, including Grok and various versions of GPT, consistently lean towards the libertarian-left quadrant of the Political Compass. The models were subjected to a quiz that measures political perspectives, with results showing that fifteen out of sixteen AI models remained firmly in the leftist camp, while Grok demonstrated variability between left and right in different runs. Despite the lack of rigorous scientific controls, the outcomes prompt interesting discussions about the ideological biases embedded within AI, linked to the sources of their training data. This analysis signifies a critical moment for the AI/ML community, raising awareness about potential biases in language models which could influence their outputs and interactions. The study suggests that a preponderance of left-leaning content in training corpora, particularly from platforms like Reddit and academic literature, shapes the models' political inclinations. With implications for the ethical development and deployment of AI technologies, the findings highlight the importance of scrutinizing the data that informs AI learning processes, as it appears to harbor a bias that reflects a predominant liberal perspective, even when the models themselves claim to position closer to the center.