AI & Machine LearningMachine Learning

The Spam Filter That Learned

Counting words beat hand-written rules for junk mail

Spammers kept slipping past the rules people wrote against them. Then a filter started counting words. How did counting win?

You'll find out

  • What makes a word look like spam?
  • How do many clues become one verdict?
  • Why did the filter keep improving?

Why it matters

Spam filters and fraud alerts still weigh evidence in a similar way: many small clues, one probability.

About 3 min6 interactive slides · 5 questionsLesson 355 · Zone 40

Play this lesson

Free to play in your browser or on iPhone

More in this zone

  1. 352Light PhysicsSunglasses vs Your ScreenLight has a direction your eyes can't see, but sunglasses can
  2. 353ProbabilityWhy Airlines OverbookSelling more seats than exist is a probability bet
  3. 354Data CommunicationWhy Dial-Up Modems ScreamedThose noises were two machines negotiating
  4. 356Visual IllusionsThe Müller-Lyer IllusionArrows make equal lines look different
  5. 357Permutation TheoryEight Perfect ShufflesEight perfect shuffles put a deck back exactly where it started
  6. 358Procedural GenerationThe SeedThe same seed number always grows the same Minecraft world
  7. 359Procedural GenerationEndless Worlds From NoiseGames build mountains, caves and coastlines from smooth randomness
  8. 360Content DeliveryNetflix Lives Inside Your ProviderPopular videos are stored in boxes near your home