CONCEPTS & TURNING POINTS / 1 MIN READ
Statistical Learning and the Quiet Resurgence (1980s–2000s)
On this page
After the expert-system winter, AI shifted toward statistical and machine-learning methods that learn from data. This sub-article expands Section 5 of History.
Neural Networks Revived
- Backpropagation (Rumelhart, Hinton, Williams, 1986) enabled training multi-layer networks.
- Hopfield networks and Boltzmann machines advanced the theory of learning and memory.
Machine Learning as Engineering
- Decision trees, support vector machines, Bayesian networks, and ensemble methods became standard.
- Applications: spam filtering, search, recommendation, and speech recognition.
Speech and Vision
- Hidden Markov models dominated speech recognition.
- Computer vision moved from hand-crafted features toward learned representations.
The Internet Changes Everything
- Large datasets and web-scale products made ML economically central.
- IBM Watson (2011) beat Jeopardy! champions using applied question-answering.
Why It Matters
Statistical learning replaced hand-coded rules with data-driven induction, setting the stage for the deep-learning revolution.
Sources
- Rumelhart, D., Hinton, G., & Williams, R. (1986). Learning representations by back-propagating errors. Nature.
- Vapnik, V. (1995). The Nature of Statistical Learning Theory.
Back to History · Next: Deep Learning Revolution.