Connected intelligent machines
The traffic in future mobile networks is likely to be dominated by connected intelligent machines communicating with each other. Networks so far have been serving a particular type of intelligence, and that’s us – humans. In the future, mobile networks will support new types of intelligent entities, like AI-powered intelligent machines talking to each other. Such a future will have a direct bearing on how we design mobile networks.
About the technology
Until now, humans have created and consumed most of the data. Soon enterprise applications relying on AI, machine learning, machine reasoning, automation and IoT are expected to stand for more than half of the traffic. A growing share of this traffic will originate from surveillance systems, autonomous vehicles, drones, and so on, and will be consumed by machines and computer vision systems rather than humans directly.
The performance of 5G networks is close to the limits of the human sensory system: Millisecond-level latencies and 1 gigabits per second data rates. However, connected intelligent machines are not limited by the capabilities of the human sensory system. Intelligent machines sense the environment in a different way than humans. As an example, a machine can use echolocation or even the radio waves as a sensing tool. In addition, the way they will be interacting with each other is different from the way humans communicate with other humans. Intelligent machines will benefit from latencies below 1ms, data rates of hundreds of gigabits per second, and additional machine-tailored services on top of the communication. The network will need to offer enhanced features and services to intelligent machines, for example semantics translation and joint communication and sensing.
Apart from extreme performance, intelligent machines will also require other services beyond connectivity, for instance mediation, trust enablement, AI-tailored services such as data discovery, publishing and subscription, monitoring and orchestration control, and secure software updates. All of these services facilitate the interactions with other intelligent machines, including the network, which is a large-scale distributed intelligent machine in itself.
Multisensing, deeply specialized, intelligent machines are used to perform predefined tasks in specific environments. These machines will evolve further, adapting their behaviors and tasks, to interact on an intent level and learn from other machines. This way, intelligent machines will eventually be capable of performing in entirely new and dynamic environments. Intelligence will become more and more decentralized, and with collective reasoning, machines will be able to self-manage and regenerate themselves, moving towards general artificial intelligence.
High levels of interoperability among intelligent machines will be enabled by data formats, for example video codecs optimized for machines rather than for human perception, and protocols optimized for machines. This will allow to move towards a semantic-driven communication among machines. To reach full interoperability across complex systems, new interfaces for exposing intelligence will be needed.
Intelligent machines will move from human-assistance to human-collaboration, in a scenario where humans and intelligent machines co-exist and interact in a harmonized way. Human-machine communication will be done in new ways. For example, machines will move from communicating with each other via human-designed protocols to dynamically evolving the protocols and languages they use for their communication. Machines will be able to understand human senses and gain the trust of humans, until machines and humans can co-exist in almost every aspect within society.
Systems of systems
At the start, intelligent machines will be grouped into systems, and systems will interact with other systems. Systems of systems need to be carefully orchestrated from external entities. Later, we will move towards dynamic membership of intelligent machines to systems of systems, and autonomous orchestration of these systems of systems. The network will facilitate the mediation among different systems, supporting their creation and self-organization towards achieving the final goal of full autonomy of systems of systems life cycles.
Trustworthy intelligent machines
Trust is a key factor for the full adoption of intelligent machines into our society. Intelligent machines need to be privacy-preserving and explain their decisions to other machines as well as to humans. They are societally aware, expected to behave in ethical and unbiased ways, and will be held accountable and responsible when carrying out their tasks. Overall, intelligent machines will aim to be fully trustworthy, being responsible, resilient, secure, safe, explainable, unbiased, fair, and ethical.
Enhanced network platform
Networks will play a central role as intelligent machines evolve and become cognitive. Networks will offer services such as data-handling services that are specifically designed for intelligent machines and will enhance the navigation and discovery of the cyber world. The network itself will become a large-scale distributed intelligent machine that provides connectivity, compute, storage, mediation, incentives, and other services to all other intelligent machines.
Example use cases
Examples of use cases that will benefit from latencies below one millisecond include closed-loop industrial control systems, industrial robots, eXtended Reality (XR) with real-time synchronous haptic feedback, and negotiated automatic cooperative driving.
Autonomous systems and robots assist and collaborate with human colleagues to solve simple or complex tasks and perform manual activities more safely and efficiently. To smoothly cooperate with groups of humans and among themselves, AI partners require high-performance networks that offer extreme reliability, low-latency communications, and so on, as well as additional features including extreme resilience, precise positioning, sensing, and AI trust and integration.