In this blog, Stephen Unger (former CTO at Ofcom and Chair of the CW TEC 2018 Steering Committee) outlines his impressions of "The Inevitable Automation of Next Generation Networks".
The annual CW Technology & Engineering Conference was held last week. These conferences take a technology which is attracting attention, and pick it apart a bit in order to understand it better, and see what the practical applications might be.
This year we looked at the potential for recent advances in artificial intelligence to automate the operation of next generation networks. One of the key trends in the technology sector at the moment is the development of applications which combine advanced networks with smart devices, from connected cars to smartphones with embedded AI. We wanted to see what the potential is for the networks themselves to get smarter, so that they can optimise their own performance, or protect themselves against security risks.
We had a good set of speakers, from major telcos and equipment manufacturers to a variety of interesting tech startups. They gave us plenty of practical examples of how AI is already being used, and how it might be exploited in the future. These practical examples are essential when dealing with a subject which attracts as much hype as AI. Importantly, we also had a great audience, keen to ask difficult questions. And I learnt a lot!
The first takeaway is that current commercial developments are not primarily about fundamental advances in AI algorithms, though there is of course some interesting work in this area. The main engine of growth at the moment is increased processing power, which makes the exploitation of existing AI techniques practical, a point made for example by Tero Rissa (Nokia). The fuel which powers this engine is data, in a variety of forms. This means that one of the key commercial opportunities is in the development of systems which organise and analyse large datasets, and several of the speakers at the conference were working in this space (e.g. Richard Baker from Geospock, Iris Barcia from Keima, Arthur Meadows from Fetch.AI).
My second takeaway is that AI systems can, on occasion, be quite stupid, especially when they are operating on unfamiliar territory. Several speakers gave practical examples of this and there is a consensus that full automation is some way off. This has a couple of practical implications. The first is that whenever AI is being used in a mission critical application, some form of check is required on the results. This point was made for example by Peter Haigh (NCSC) in the context of his talk on network security. The second implication is that it is important for human beings to be able to understand and interact with the results produced by AI, and Ben Azvine from BT demonstrated some very impressive data visualisation tools which his teams have developed.
My third takeaway is that, despite the above, a degree of automation is inevitable. Several speakers (e.g. Mansoor Hanif from Ofcom, Dave Salam from BT, Chris Murphy from Viavi, Dean Bubley from Disruptive Analysis) commented that the complexity of next generation networks will require a degree of automation, and others described how this capability is already being incorporated into technical standards (e.g. Yue Wang of Samsung, Ray Forbes of Huawei). Automation will deal with the 80% of issues which are straightforward, and will eliminate human error as a common source of network failure - which it currently is. Human beings can then focus their time on the long tail of network issues which will continue to require human judgement
There’s more detail in the speakers' presentations where permission has been provided for publication. Many thanks to everyone who helped make this year's conference a success!