Knowledge 141 You might argue we could devise a more sophisticated mechanical filter, something that contains an algorithm with an understanding of the rules of language. The problem is both the size of the task and the nature of understanding. If I devised some really good appreciation algorithm which did not delete all the creative words of the English language, it would still have to read and appreciate the huge quantities of input until it hit upon something good. There is no way for any machine to read all this information in the age of our Universe; the numbers are just too large. And there is no way for a machine to understand all the rules of language, they are not written down and constantly evolve. These descriptions should give you an intuitive feel for nature of the creative problem. If you try to deconstruct it into mechanical steps you end up with either a mechanism that needs to be infinitely specified or one that lets through an infinite quantity of nonsense. A human could never sift through all that garbage to find the occasional pearl of wisdom. Until the beginning of the 20" century, most people thought knowledge and creativity must be just a matter of scale. A big enough, fast enough machine should be able to solve any problem. But early in the 1930s two mathematicians — Kurt Godel and Alan Turing — showed knowledge was not so simple. Let me give you a feel for why. Knowing When You Know The essence of creating knowledge, is to know when you have done so. In a sense, counting from one to infinity means I know everything, and merely counting to 50 million creates every piece of significant symbolic knowledge that will ever be written — all the books, plays, mathematical theorems you could possibly want. But, if I were to list all these numbers in an enormous imaginary book it would hardly constitute knowing everything: I would be awash with numbers but not with knowledge. The essential feature of ‘knowing’ is to have a small number of steps th