I just watched an interview with Geoffrey Hinton, one of the pioneers of neural nets and deep learning and he suggested that it needs more than language (multi-modal). So you know.. probably not.