The growth of AI is based upon our models of cognition, which are incomplete. So until an accurate model of cognition exists, the supporting hardware limits the implementation. One could argue that our own biological hardware is limits our cognition. Since we are the architects of AI, this then creates a feedback loop of limitations.