Artificial Intelligence — The Revolution Hasn't Happened Yet

ai-hypemachine-learningintelligence-augmentationengineering-disciplinedata-science

Abstraction: Michael Jordan argues AI hype obscures the real ML/data-engineering challenge

Key points:

What is publicly called "AI" is largely Machine Learning (ML) / Data Science — a rebranding, not a new field; "human-imitative AI" (reasoning, cognition) remains largely unsolved
Jordan proposes three complementary goals: AI (human-imitative), IA (Intelligence Augmentation — search, translation), and II (Intelligent Infrastructure — societal-scale decision systems in medicine, transport, finance)
II systems must handle distributed knowledge, cloud-edge tradeoffs, long-tail data, data sharing across boundaries, and market incentives — challenges orthogonal to human-imitative AI
Backpropagation traces to 1950s–60s control theory, first applied to optimize Apollo spacecraft thrusts — not a product of the "AI" tradition
Current "AI" approach (gather data, deploy deep learning, demo narrow skill) deflects attention from unsolved classical AI problems: causality, reasoning, uncertainty, long-term goals
Calls for a new human-centric engineering discipline analogous to civil or chemical engineering, integrating social sciences and humanities, with emphasis on provenance, reliability, and safety

Connections: Michael Jordan Professor · Machine Learning · Data Science