All neural models I've developed can operate autonomously and learn either unsupervised (learning from inputs alone) or supervised (learning with feedback about predictive success); perceptual and cognitive processes are designed to be general-purpose—they can respond adaptively to any input pattern in their domain and to changing environments.