The 1950s program that learned to outplay its maker | How AI Works · S 06, Ep 2

Last time, we met machine learning: the part of AI that learns from examples instead of following rules people wrote. How can a computer learn something nobody told it? Start with a job that sounds easy to write rules for: spotting junk email. Episode 2 of How AI Works from Aussie Science Clips: What's actually going on inside the AI everyone is talking about: how a machine learns, how a chatbot writes a sentence and makes a picture, and why it sometimes gets things confidently wrong. Full transcript, sources and every episode: https://aussiescience.com/how-ai-work... CHAPTERS 0:00 The question 0:19 Writing the rules by hand 0:56 A checkers board at IBM 1:22 Too many moves to write down 2:03 The machine picks the weights 2:33 Alpha and Beta 3:05 Beating its own author 3:45 How good was it, really? 4:18 The smallest learning there is 4:45 Squaring the misses 5:25 An idea from the comets 5:57 Training, then using 6:19 A spam filter that counts 7:00 Scoring a new email 7:32 What the counting found 8:06 Testing on emails it hasn't seen 8:37 It learns whatever is in the data 9:09 Wrong answers in the answer key 9:33 Memorising the practice exam 10:09 Where the analogy breaks 10:31 The road Samuel didn't take NEXT: What is a neural network, and how is it not like a brain? SOURCES 1. P. Graham, "A Plan for Spam" (August 2002) — https://www.paulgraham.com/spam.html 2. M. Sahami, S. Dumais, D. Heckerman, E. Horvitz, "A Bayesian Approach to Filtering Junk… — https://cdn.aaai.org/Workshops/1998/W... 3. A. L. Samuel, "Some Studies in Machine Learning Using the Game of Checkers", *IBM… — https://people.csail.mit.edu/brooks/i... 4. IBM, "The games that helped AI evolve" (IBM history) — https://www.ibm.com/history/early-games 5. J. Schaeffer, Chinook project, University of Alberta: "Arthur Samuel's Legacy"… — https://webdocs.cs.ualberta.ca/~chino... 6. Wikipedia: "Least squares" (history: Legendre 1805, Gauss 1809 and the 1795 claim, Ceres) — https://en.wikipedia.org/wiki/Least_s... 7. Google for Developers, Machine Learning Crash Course: "Linear regression" (y′ = b +… — https://developers.google.com/machine... 8. MacTutor History of Mathematics (University of St Andrews): "Adrien-Marie Legendre" — https://mathshistory.st-andrews.ac.uk... 9. Google for Developers, Machine Learning Crash Course: "Static vs. dynamic training"… — https://developers.google.com/machine... 10. Google for Developers, Introduction to Machine Learning: "Supervised learning"… — https://developers.google.com/machine... 11. scikit-learn User Guide: "Naive Bayes" — https://scikit-learn.org/stable/modul... 12. Google for Developers, Machine Learning Crash Course: "Dividing the original dataset"… — https://developers.google.com/machine... 13. J. K. Winkler et al., "Association Between Surgical Skin Markings in Dermoscopic… — https://jamanetwork.com/journals/jama... 14. C. G. Northcutt, A. Athalye, J. Mueller, "Pervasive Label Errors in Test Sets… — https://arxiv.org/abs/2103.14749 15. Google for Developers, Machine Learning Crash Course: "Overfitting" — https://developers.google.com/machine... 16. scikit-learn examples: "Underfitting vs. Overfitting" — https://scikit-learn.org/stable/auto_... 17. Google for Developers, Machine Learning Crash Course: "Model complexity" ("Simpler… — https://developers.google.com/machine... HOW THIS WAS MADE Researched and scripted with AI assistance, then fact-checked claim by claim against the sources above by a separate AI pass. Narration is two synthetic Inworld voices. Diagrams are drawn in code (Manim); atmospheric images are AI-generated (FLUX) and are illustrations, not footage. No real person is depicted. Spotted an error? Tell us in the comments and we'll correct it.