On the thresholds of knowledge