It's silly human bias. David Cope's program composed music which people claimed was great. As soon as they were told it was made by a computer they suddenly thought it was terrible. In a test between an actual composer and his algorithm, the audience thought the generated music was composed by a human.

It's true today's algorithms might not be completely original, but then again neither are most humans. And this isn't a strict limitation, computers could learn to experiment with new patterns, or generalize further than mere imitation.