Then would a better interview technique be something like - here is a spec for an algorithm. Now implement it. I see very little value in having people studying

Then would a better interview technique be something like - here is a spec for an algorithm. Now implement it. I see very little value in having people studying algorithm implementations for interviews. Of course, knowing about specific algorithms is handy, and adjusting the question to find an appropriate spec to implement for a given problem could be a nice touch. I think that either approach above is strictly better than implementing algorithms from memory and would produce a far better signal.