Here's the thing: those are perfectly reasonable requests, the rules aren't arbitrary, there's some determinism, etc.
I went to an interview, and I got asked how to write something that calculates the nth Fibonacci number. The naive solution is of course to do it recursively, but I'd recently been reading about it, and I had an iterative solution. So I wrote that down. Turns out the question is multi-part, and the second part is "do it faster". My solution was already pretty optimal, so I didn't have anywhere to go. It was clear that this wasn't okay. I did the work better than expected on the first try, but got docked points for the interviewer's lack of imagination.
This is what people mean when they say they don't want to solve riddles. I'm not there to be a dancing monkey, we're trying to have a dialogue about why I'm right for your company and to clear a minimum bar for skill. There's this story about fizzbuzz (http://blog.codinghorror.com/why-cant-programmers-program/) where basically the writer concludes that bad programmers are weeded out by really minimally challenging problems, and making things more complicated doesn't give you any better sense of success.
My questions in interviews are more philosophical. "What do you think about commenting code?" "What's your favourite programming language? Why? What problems does it have? Do those problems have solutions? How do you work around issues like that?" These are questions that programmers think about on their own time and have opinions about, and having an opinion and being able to talk intelligently about it tells me a lot more about how engaged and appropriate someone is for the job than basic BS like 'write a solution for the nth Fibonacci number'. That's not an inherently interesting problem, it tells me very little about what you'd be like to work with or even your problem solving skills, and ultimately bores the both of us.