The Problems With GNNs in Production

Even inside of this really small kind of click, let's call it Silicon Valley, top tier web companies, no matter what geography they might be based at, I can really only get to a handful of companies that are using them in production. One is just performance and scalability. Really kind of anything that does any kind of scale is very, very hard to shove into GNNs. And then you have the classic kind of broad neural network challenge of opacity.