The massive sensing data generated by Internet-of-Things will provide fuel for ubiquitous artificial intelligence (AI), automating the operations of our society ranging from transportation to healthcare. The realistic adoption of this technique however entails labelling of the enormous data prior to the training of AI models via supervised learning. To tackle this challenge, we explore a new perspective of wireless crowd labelling that is capable of downloading data to many imperfect mobile annotators for repetition labelling by exploiting multicasting in wireless networks. In this cross-disciplinary area, the integration of the rate-distortion theory and the principle of repetition labelling for accuracy improvement gives rise to a new tradeoff between radio-and-annotator resources under a constraint on labelling accuracy. Building on the tradeoff and aiming at maximizing the labelling throughput, this work focuses on the joint optimization of encoding rate, annotator clustering, and sub-channel allocation, which results in an NP-hard integer programming problem. To devise an efficient solution approach, we establish an optimal sequential annotator-clustering scheme based on the order of decreasing signal-to-noise ratios. Thereby, the optimal solution can be found by an efficient tree search. Next, the solution is simplified by applying truncated channel inversion. Alternatively, the optimization problem can be recognized as a knapsack problem, which can be efficiently solved in pseudo-polynomial time by means of dynamic programming. In addition, exact polices are derived for the annotators constrained and spectrum constrained cases. Last, simulation results demonstrate the significant throughput gains based on the optimal solution compared with decoupled allocation of the two types of resources.