Can AI accept mistakes and learn from them?