While there is talk of centralization of the Internet and breaking up of Big Tech, this discussion is often limited to the application layer of the Internet. What about the infrastructure? This research paper by Moura et al. at Internet Measurement Conference (IMC) 2020 discusses the centralization of the Internet based on measurement of DNS traffic sent to .nl, .nz and b-root servers. Centralization affects end-users in both good (e.g., deployment of QNAME-minimization) and bad (e.g., single point of failure) ways.
While this paper looks at who (large firms and elite universities, they say) gets papers into top AI conferences, I found Figure 2 interesting as it has a few security and networking conferences. I wonder why, in the early 2000s, the proportion of papers with at least one author from a tech firm drops off significantly at USENIX Security. For some reason, in the appendix, the authors classify INFOCOM as a visualization conference.