Machine-Learning Maestro Michael Jordan on the Delusions of Big Data and Other Huge Engineering Efforts

Abstraction: Michael Jordan critiques deep learning hype and big data statistical risks

Key points:

• Deep learning is largely a rebranding of 1980s neural networks; backpropagation (its key success) is not brain-inspired and the brain analogy is misleading
• Computer vision has solved a narrow class of problems (image classification) but humans vastly outperform machines on cluttered scenes, relational inference, and interaction
• Big data danger: number of possible feature combinations grows exponentially, guaranteeing spurious correlations — "billions of monkeys, one writes Shakespeare"
• Predicts a "big-data winter" within 2–5 years from unchecked hype and missing error bars, analogous to past AI winters
• With $1 billion, Jordan would invest in natural language processing, specifically question answering (e.g., negation handling which Google handles poorly)
• Turing test is a media event, not a scientific demarcation; intelligence will emerge gradually across many systems

Connections: Michael I Jordan · Machine Learning · Big Data · Deep Learning · Statistical Inference

Source: http://spectrum.ieee.org/robotics/artificial-intelligence/machinelearning-maestro-michael-jordan-on-the-delusions-of-big-data-and-other-huge-engineering-efforts