Do we need another G? or is this the inflection point for Telco/techCos to become the “Networks for AI” and position themselves as the lynchpin of Sovereign and Private Industry AI. As AI distributes across Data Centre, Edge and Device the performance and security of a combined wireless/wireline network will be table stakes. This could be the killer use case for 5G/5G-A. The convergence of compute and communications is inevitable. #MWC24 Spirent Communications
Yes another series of G's is necessary to extricate the telco's from telecommunications through decentralization. > Eliminate simcards, Hardware attested authenticators are supported on all user devices. Perfect for decentralized identity. > Similarly decentralize routing put the user in control of discovery via a W3C DID-document / equivalent. Allow peer-2-peer RTP/other negotiation. > Eliminate spectrum ownership it's a inefficient method of allocation / usage. > Develop a scheme that enables users, orgs and communities to stand-up their own antennas without being constrained to CBRS. > Also rethink the complexity of TCP/IP/HTTPS/APIs to enable sharing of unstructured data as fully authenticated transactions with end-to-end encryption.
There might not be a single next G. :-) Convergence? Depends on what you mean and at what level. Network convergence meaning something very different from a more forward view on the interplay, interworking, convergence between and of cloud computing and communications services. I'm finding a lot of the thinking around "Network for AI" (seemingly confused with network for massive AI supercomputing) is half-baked and largely devoid of consideration of what is happening on device, across edge infrastructure, and related economics of distributed AI computing. BTW, looking good, Stephen Douglas!
It's fashionable to dump on telcos while praising everything that comes out of the valley, entire consultancies depend on it. Bit it's all bits now regardless.
Convergence of telecommunications with computing has been evangelized since the 90s.
next G is AIG
I disagree. Compute & comms coexist, but aren’t really converging (that’s a line used by policymakers to try to regulate cloudcos). DCs & devices make a lot of sense for training / inferencing respectively, but the edge seems a bit of wishful thinking… except for running the network itself. I’m not yet seeing much evidence of AI impacting traffic on the access network, especially for consumers or on mobile vs. fixed networks. Curious if you’re seeing signs to the contrary