This article articulates the emerging paradigm, sitting at the confluence of computer vision and wireless communication, to enable beyond-5G/6G mission-critical applications (autonomous/remote-controlled vehicles, visuo-haptic VR, and other cyber-physical applications). First, drawing on recent advances in machine learning and the availability of non-RF data, vision-aided wireless networks are shown to significantly enhance the reliability of wireless communication without sacrificing spectral efficiency. In particular, we demonstrate how computer vision enables {look-ahead} prediction in a millimeter-wave channel blockage scenario, before the blockage actually happens. From a computer vision perspective, we highlight how radio frequency (RF) based sensing and imaging are instrumental in robustifying computer vision applications against occlusion and failure. This is corroborated via an RF-based image reconstruction use case, showcasing a receiver-side image failure correction resulting in reduced retransmission and latency. Taken together, this article sheds light on the much-needed convergence of RF and non-RF modalities to enable ultra-reliable communication and truly intelligent 6G networks.