There has been much discussion lately about 6G being truly AI-native, building on an increasing level of AI and machine learning (ML) being embedded in networks as the 5G-Advanced standards are implemented over the coming years. This raises the question of what difference an AI-native network would make for operators and users, as well as what is actually entailed. AI in the network context is primarily a means of harnessing the vastly increased computational power and storage density now available, compared even with a decade ago, and 5G/6G applications mainly focus on ML. In wireless communications, there are three pillars underlying AI/ML, which have been established, in principle at least, for at least a century. These pillars are: tensor calculus,…