There is a lot of data on this question, and it flatly contradicts Asimov's claims. IQ scores correlate not only with academic performance but also with job performance as well as others' intuitive impression of how "smart" a person is. That's why the army gives intelligence tests to new recruits: high-IQ soldiers (and mechanics and carpenters) do better, on average, than their low-IQ counterparts. Other types of tests, such as the hypothetical test devised by Asimov's mechanic, either have no predictive power or are predictive only as far as they correlate with g, or general intelligence, which is currently best measured by IQ tests. The physiological basis of g is unknown, but it is highly heritable and correlates with a number of physical variables such as brain volume and myelination.

If you want to learn more about the subject, I suggest this book, which concisely disposes of these and other popular anti-IQ arguments:

http://www.amazon.com/Question-Intelligence-IQ-Debate-Americ...