orthogonal? Neural networks are a layered graph structure. When they talk about "parameters" they're referring to nodes and edges.
When they talk about sending a video they’re referring to TCP/IP. I bet they can’t even talk about elliptic curve cryptography!
GP here is out of their depth.
You don't need graph theory to understand NNs at all.
You need Linear Algebra and Differential Calculus, though.
You do need graph theory to understand and do Graph Neural Networks. But that's a subfield of modern AI and many AI researchers don't know/study/research Graph NNs at all.
You don't actually need to know graph theory to do RL, Vision, ANNs, NLP, etc. unless explicitly needed in your research/job.