AI & Machine LearningReinforcement Learning

AI Plays Hide and Seek

Agents found tool tricks and exploits nobody taught them

AI hiders and seekers played hundreds of millions of rounds. Nobody taught them tools, yet they built forts. What came next?

You'll find out

  • How rewards alone taught teamwork
  • The six moves of an arms race
  • The trick that surprised the researchers

Why it matters

Simple rewards can grow surprising behavior. That's exciting for AI, and a reason to watch it closely.

About 3 min6 interactive slides · 5 questionsLesson 693 · Zone 77

Play this lesson

Free to play in your browser or on iPhone

More in this zone

  1. 685Thermal PhysicsCan Hot Water Freeze Faster?Scientists still argue about the Mpemba effect
  2. 686Machine Learning EvaluationWhen a Measure Becomes a TargetGoodhart's law catches AI and people alike
  3. 687Decision TheoryNewcomb's ParadoxA predictor who's always right offers you two boxes
  4. 688Software HistoryY2KSaving two digits almost broke the year 2000
  5. 689Software HistoryThe 2038 ProblemMany computers' clocks will run out of numbers on January 19, 2038
  6. 690Hardware SecurityWhere Your Face Is StoredA locked chip inside your phone guards secrets from the phone itself
  7. 691Cryptocurrency HistoryThe Bitcoin PizzaIn 2010 someone paid 10,000 bitcoins for two pizzas
  8. 692Fluid MechanicsThe Shrimp That Shoots BubblesA collapsing bubble can be hot enough to glow