I have two main thoughts about knowledge graphs. On the one hand, they aren't models of how we think - human thought, unlike a knowledge graph, is subsymbolic. The neural network is a graph, but not a graph of concepts or words. On the other hand, knowledge graphs, as shared social artifacts, are super-useful as a way of representing what we, as a society, have come to believe. This utility applies to artificial intelligence as much as it does for humans, it it is near certain that something like knowledge graphs will be used to supplement large language models in AI. That's why the concept of social linked data is so important to the future of AI, even if it doesn't feel like it at the moment.
Today: 3 Total: 108 [Direct link] [Share]
