<ul><li>AI systems reflect the data they are trained on, interpreting and generating new reflections based on patterns within the data.</li><li>Training bias in AI systems, resulting from over-representation or under-representation of certain perspectives in the training data, can lead to a warped reflection of reality.</li><li>Understanding AI training data is crucial to interpret the output of AI systems, as their responses may contain factual errors or biases originating from the distorted training data.</li><li>The development of AI systems requires a consideration of training data, including who selects it and the values guiding the selection, to build better and more comprehensive mirrors of humanity.</li></ul>

The Mirror’s Edge: What We See When AI Looks at Us

Discover more