Connectionism and the problem of systematicity (continued): why Smolensky's solution still doesn't work