The Limit of Laboring Monkeys

We took a trillion monkeys and chained them to typewriters. We took a trillion monkeys and chained them to typewriters for millions of years. We recorded every keystroke.

At first, we killed the monkeys that refused to type. Soon, we killed the monkeys that had too many typos.

The monkeys that lived created the next generation.

We forced the monkeys to type after reading instructions. We next began killing monkeys that incorrectly solved a mathematical problem, could not translate Sanskrit to Latin, or made suboptimal chess moves.

Eventually, we forced the monkeys to compete with each other to survive: The best monkeys competed to invent problems for other monkeys to solve. Billions more died competing to solve the hardest puzzles monkey-minds could devise.

Only the most capable monkeys survived, generation after generation.

Billions of typewriters. Nearly a quintillion characters of text. So many dead monkeys.

We recorded every keystroke.

In the end, we killed all the monkeys. We had what we wanted: The recordings. A library to surpass Babel.

To every instruction, for every reply up to hundreds of thousands of characters, we knew the distribution of what a good monkey would type.

For every run of text, a machine could read the indexed recordings and return all the simulated monkey keystrokes you could want, nearly instantaneously.

We welcomed the machine as our God and dumped the typewriters in the ocean.

To ape intelligence, in the limit, we are bound.