Module · The machines that came first✅Knowledge checkCheck: can you make a search affordable?5 questions · Pass at 60%1. Alpha-beta examined 18,297 positions where minimax examined 549,946, and returned the same move. What was given up to get that?Nothing — it proved those branches could not change the answerThe guarantee of perfect play, in exchange for speedAccuracy in the endgame onlyA small number of wins against weak opponents2. A* uses f = g + h. What goes wrong if h is allowed to overestimate the distance still to travel?The search becomes slower but still finds the shortest routeIt can be talked out of the true shortest route and return a worse oneIt stops being able to find any route at allNothing — overestimating makes it more decisive3. The first A* run in this lesson saved almost nothing, because huge numbers of cells scored the same total. What fixed it?A finer gridLetting it revisit cells it had already examinedBreaking ties towards the cell nearest the goalA larger open list4. Stopping two moves deep and estimating lost 6 games in 400 and did roughly two thousand times less work. What is the general shape of that trade?More memory bought less computationCertainty was given up to make the problem affordableAccuracy was given up to make the code simplerNothing was given up; the estimate is exact5. Both ideas in this lesson depend on a guess — the distance still to go, and how good a chess position is. Where did those guesses come from?They were learned from recorded gamesThey were derived from the rules of the gameA person sat down and invented themThey were tuned automatically by the search itselfCheck answers← Back to Introduction to AINext topic →
