That, and also it's just a completely different approach which might later on turn out to be useful. People should remember that artificial neural networks were developed decades before they were useful. People were doing all kinds of other approaches to ML like support vector machines before advances in hardware made deep neural nets feasible and therefore interesting again. ANNs were never obsoleted by SVMs.