I imagine the point of this algorithm, like a lot of algorithm research, is to prove the upper bound of complexity for this problem. Not to be used in practice (despite what this article seem to suggest).

On a similar note, there's a lot of work put into optimal matrix multiplication algorithm. We know the lower bound is N*2, the obvious upper bound is N*3, the best (complexity wise, not practical at all) current algorithm is N*2.37, but we don't know how fast can it really get. Is it possible to write N*2 algorithm? We don't know.