Mathematics & Logic
- Learning by reward
How reinforcement learning works: a program gets a score instead of correct answers, and the score can teach the wrong lesson.
- 9min read
- 10min listen
- 29questions
- Learning with no one to mark the answers
How machines learn from data nobody has labelled, by sorting it into groups, compressing it, and predicting the parts that were hidden from them.
- 8min read
- 11min listen
- 26questions
- How much is a surprise worth?
Claude Shannon measured information by how surprising a message is. The same measure explains file compression and how language models are trained.
- 9min read
- 11min listen
- 31questions
- Changing your mind by the numbers
Bayes’ rule says how far new evidence should move a belief, and why a positive test for a rare disease often means less than it seems.
- 9min read
- 10min listen
- 28questions
- Thinking in lists of numbers
Vectors and matrices: how a photo, a word or a film becomes a list of numbers, and how search, recommendations and neural networks compare them.
- 8min read
- 10min listen
- 29questions
- Secret codes
How ciphers work, how counting letters broke them for centuries, and why modern encryption depends on a key and on maths that is hard to reverse.
- 8min read
- 11min listen
- 29questions
- Transformers
The design behind chat assistants: how text becomes numbers, how each word is weighed against every other, and why a fluent reply can still be false.
- 9min read
- 9min listen
- 30questions
- Deep learning
How a neural network learns to recognise pictures from examples instead of written rules, why that took decades to work, and where it still fails.
- 8min read
- 10min listen
- 33questions
- The bitter lesson
Rich Sutton’s 2019 argument that AI methods which use more computing keep beating methods built on human expertise, and what his critics reply.
- 7min read
- 10min listen
- 24questions
- Algebra: finding the unknown
Al-jabr means restoring: one move from a Baghdad book of the ninth century, the balance it keeps, and what letters can prove that numbers cannot.
- 9min read
- 11min listen
- 31questions