@JohnRennie If so, then a natural question to ask is what is the criteria that a finite number of observations become sufficient for induction? And I'm sure the measurements in the observations will not accurately reflect the law. So how to choose the range of acceptable error?
And induction is quite problematic. For instance, I can say that "all swans are white" if I just happen to not find a black swan.