Abstract

Data is no more the commodity oil. Today, it is an asset for any enterprise. However, turning data into intelligence remains a challenge for most people. In this paper, we explore whether databases can be turned into generative AI machines that can talk to anyone. We identify three core challenges when applying generative AI on data, namely accuracy, scale, and privacy, and show how a generative large data model could solve all of these. We describe our conceptual framework of generative AI on databases, the GOD machine, and ground it in production workloads at SmartApps. Our results promise new directions in fusing AI with data.