It was said by others (like lessdazed) but I insist on the difference between "very long to compute" and "not computable". Finding the winning moves in chess may require longer than the universe lifetime, but is theoricaly possible given enough computer power. Kolmogorov complexity is not computable, due to the impossibility of having an halting oracle. A full paragraph about those issues and how to get around them would be worth it IMHO. Reading the text without knowing that makes it feel that SI is just the same kind of "too long to be directly used" problem like winning at chess, while in fact it's a "level" harder (not even computable).