IJKLM -- the reason I'm going to regard i = sqrt(-1) as a special case is because its history simply doesn't support the idea that it was invented. It's well worth a read, and very enlightening, to study the history of that number through the 15th-17th centuries, when it was first discovered and investigated.
See also my earlier post -- to that I should perhaps add that, at the same time as sqrt(-1) was first encountered, the only solutions that any mathematician accepted as "real" were those in positive numbers. Descartes wrote in his textbook "La Géométrie", among other things, that "sometimes it happens that some of the solutions [to a quadratic equation] are false, or less than nothing". It stands to reason that if people don't accept the reality of negative numbers then there's no reason to invent their square roots either. What's the square root of meaningless bull? A meaningless bullock, perhaps.
Instead, as I say, the study of cubic equations led to the remarkable result that some equations whose solutions were definitely positive whole numbers could be expressed in terms of this bizarre, and completely meaningless, object called sqrt(-1). What nonsense! And yet there it is, staring you in the face, forcing you to accept its existence. There is no other word for it: this was a discovery. Nobody asked for it, nor did they demand it; the fact that this strange object turns out to be so useful is the purest of accidents.