TURIN — From vacuum-cleaner robots that clean your house all by themselves to virtual assistants that can keep track of every appointment in your work agenda to self-driving cars that will soon be circulating on the streets of our cities: One of the most important applications of Artificial Intelligence is the ability to make our technological devices ever more autonomous, able to understand on their own how to accomplish a task. Yet it is curious that while many such devices connected to the internet are learning to run on their own, the system that supports this transformation — network infrastructure (routers, hubs, servers, etc.) — continues to largely be operated manually.
Recently, though, the situation has begun to change thanks to the efforts of two leading companies in the industry: Cisco and Juniper. The first launched its Intent-Based Network a few months ago, introducing a new generation of networks capable of independently learning how to handle the flow of data; the second responded with the Self Driving Network project, which takes inspiration from autonomous cars in its aim to completely transform the operation of networks. "Our goal is not to make some part of the network autonomous, but to make it so that it's able to handle it entirely on its own," Kireeti Kompella, Senior Vice President and CTO of Juniper, told La Stampa. The company, which has reached $5 billion in annual turnover, includes among its customers the world's top 10 telephone providers.
We must build greater responsibility into the machines.
Just as Cisco's Intent-Based Network is in its early stages, the road to Juniper's Self Driving Network is also long: "We have all the technologies that we need, but we still have to make progress in the field of machine learning (the technique of artificial intelligence) applied to networks," explains Kompella. "This technology is making giant strides with regard to voice and image recognition, and in many other fields, but as far as networks are concerned, we are still at the beginning."
How long will it take before we have networks that can be configured on their own, to understand how to manage themselves, to solve problems and perhaps most crucially, to defend themselves from intrusion and cyber attacks? "It's hard to make a prediction, but it might take about five years," explains the Juniper CTO. "At that point, it will be possible to anticipate problems and prevent them, quickly spotting the signals that indicate something is not right. It is an evolution that is essential, especially if we think of the exorbitant number of devices that will be connected to the Internet of Things."
AI giving a hand — Photo — Samuel Zeller
Just as autonomous cars have raised several ethical dilemmas, self-driving networks are not immune to such questions. "Obviously, in our case, it is never a matter of life and death, as can be the case with autonomous cars. Nevertheless, the networks handle very sensitive data, and therefore we must build greater responsibility into the machines," says Kompella. "Whenever we increase their autonomy, we reduce our own and lose control of the details. In the world of technology there is a bad habit of moving ahead with each progressive step without stopping too much to reflect on the consequences."
Obviously, one of the consequences of network automation is jobs: What will happen to network engineers? "Human workers will be relieved of the most pressing and repetitive tasks, and will have time to focus on what is really important. In the future, however, new skills may be required, while service design will become increasingly important," says Juniper's CTO. "But it's also true that this is a problem that all of society is facing: Even elder care, considered one of the most protected professions, is automating."
In Silicon Valley, and elsewhere, many are talking about "universal basic income" as a solution to the problem of mass unemployment; but at that point another question arises: What will we do with all of our new spare time?"