Why does computer vision data need better infrastructure than AWS S3?

The next big thing in AI is computer vision. It has the potential to take over most real-world human tasks. However, today there is no data infrastructure built to support computer vision projects. The majority of us store vision data (images/videos) in S3 or Google Cloud Storage buckets. The problem is that it is not easy to search, preview, or use vision data stored in these buckets. This is because they are not designed for computer vision, but rather general purpose storage containers. Please comment your thoughts and how you solve this problem.