Berkeley AI professor Stuart Russell's landmark book reimagines the foundations of AI — not as systems that optimize fixed objectives, but as systems that remain uncertain about human preferences and learn from us over time. This framework, called "assistance games", might be the key to building AI that is genuinely beneficial by design. Deep dive: https://medium.com/@adnanmasood/human-compatible-engineering-value-alignment-in-artificial-intelligence-3ebb39222104
