logoalt Hacker News

sedan_baklazhan • today at 12:01 PM • 0 replies • view on HN

You can't prove that point,because model training data is opaque.