Scientists from the University of Toronto and MIT discovered that AI systems trained with descriptive labels may make harsher decisions than humans. The research reveals profound flaws in AI algorithm training, which could affect decision-making in areas such as housing, loans, and surgeries, leading to catastrophic consequences for humanity. The experiment used a dog image dataset, showing that descriptive labels resulted in harsher judgments of non-compliant dogs, potentially leading to social injustices. Algorithms may reinforce social biases in areas such as PhD applications and hiring, exacerbating inequality. Researchers urge for prompt corrections to this issue.