Latest AI Is Too Unpredictable to Behave According to Human Goals Scientific American January 27, 2025 0Comment AI “alignment” is a buzzword, not a feasible safety goal