Signal City Archive Desk · the argument behind the first AI winter · back to the desk
A program that "looks at everything" has to check every possible line of play. Each turn multiplies the work by the number of choices. In 1973, Sir James Lighthill named the combinatorial explosion as one big, general reason AI had disappointed. Turn the dials.
| A machine checking a thousand positions a second | |
| A million a second | |
| A billion a second | |
| A trillion a second |
Rough numbers on purpose: real games have more or fewer choices at different moments, and the speeds are round figures, not specific machines. The shape is the point. Add one more turn and the work multiplies again. A faster computer only buys you a turn or two. The field got out of this trap by not looking at everything: smarter search, rules of thumb, and eventually learning which lines are worth looking at. Hear the episode: History No. 4, The First Winter. Everything runs in your browser.