The Bitter Lesson

Machine learning, Artificial Intelligence
Richard Sutton

The Bitter Lesson is a pattern that can be observed in several areas of machine learning: many hard problems involving some form of artificial intelligence have seen dramatic progress at some point in the last 50 years, which was mostly driven by data and computations as opposed to “clever” human engineering.

If this trend is a fundamental principle (which is what the article argues) it would mean that most of the time spent on engineering features and task-specific representations is wasted. An automatically learning system such has a neural network will eventually have more success provided enough data is available to train it.

Last changed | authored by


← Back to Notes