As Andrew Ng recently pointed out, "Every hospital has its own slightly different format for electronic health records" and "you might have 10,000 manufacturers building 10,000 custom AI models" (https://spectrum.ieee.org/andrew-ng-data-centric-ai). In other words, every firm has a different data model.<p>As data volume and uses continue to explode, what will happen to data <i>structures</i>? Will data model / schema / structure variance get wider or narrower?<p>Argument for narrowing: standardization becomes too powerful to resist. Example: Ford standardizing auto parts<p>Argument for widening: everyone competes for an "edge" in extracting value from data. Generating the "edge" means proprietary models. Proprietary models require not only bespoke "black box"es, but also proprietary ways of normalizing/standardizing data and related schemas. Furthermore, even if the model was not shared, each industry participant is not going to wait for some common "standard" to reflect the latest evolution of how it thinks about its data model.<p>Which trend will win?