Which brings me to one of my peeves about this latest "memorization" craze in tech interviews. I just don't see the value of this over asking conceptual questions, etc. For example, in the example above, I don't think there is as much value in memorizing how to convert Celsius to Fahrenheit as there is in knowing how to derive the equation based on a couple of common data points and an understanding of the relationship between the two scales.
However, aside from the two data points, there is something that needs to be memorized to derive the equation, and something that needs to be understood about the relationship between the two scales. So there are things that do need to be memorized. But what are the best things for someone to memorize? I think the best things are principles that form the basis for solving other types of problems. So in the above case, memorizing the method of deriving the slope of a line (in two dimensions), and that the relationship between Celsius and Fahrenheit is linear, allows one to derive the equation to convert between the two. What's more, one can solve any problem involving equations of lines (in two dimensions) given a couple of points on them.
I remember discussing this aspect of problem solving with someone at Google, who admitted that some interviewers aren't very good. It surprises me that Google doesn't have more internal quality control for interviews, given that CEO Eric Schmidt has often remarked that if he could only find people qualified to do the things Google wants to do, there would be all sorts of market opportunities, etc. But there are widespread anecdotes of people who've been turned down by Google, and have cited an inability to answer trivia questions. (Granted, this may not be the only reason they were turned down.) But it seems strange to me that a company that needs people as badly as they claim to would not restructure interviews to maximize the likelihood of finding qualified people.