AI and its Significance to 5G
Perhaps not too surprisingly, news coming out of Las Vegas last week (CES 2020) tended to focus on the mass rollout and uptake in 5G, and the evolution of ML (Machine Learning) and AI (Artificial Intelligence) as it begins to encompass all aspects of our life. As such, I thought I would take this opportunity to take a brief look at how these two technologies may combine.
Fuelled by Hollywood, the common perception today is that AI enables computers to be programmed to actually be a human mind; to be intelligent in every sense of the word, to have perception, beliefs and the general cognitive states attributed to humans. This is “Strong AI” and to date, remains solely in the minds of the film producers and directors.
Instead, the AI which is in widespread use today is termed “Weak AI”, examples of which include: the automated personal assistants of Alexa and Siri which conduct natural language processing; autonomous vehicles and their ability to interpret their environment; computer vision, which is able to conduct facial recognition or analyse medical scans. It is this type of AI which is being incorporated into 5G.
What is Artificial Intelligence?
There is no universal definition of Artificial Intelligence however it can be expressed as the ability for a machine to perform cognitive functions which we associate with the human mind, such as perceiving, reasoning, learning and problem solving.
A conceptual view of AI[1] can be described as an “intelligent agent” which in turn consists of three main elements; sensors, operational logic and actuators. The sensors collect raw data from the environment whilst the actuators act to change the state of the said environment.
Clearly, the power of the intelligent agent is the operational logic as it must, for a set of defined objectives, take the input data from the sensors and provide output to the actuators. These could take the form of recommendations, predictions and decisions that in turn can influence the state of the environment.
The operational logic element is comprised of various algorithms which are used to build a model of the environment and then to interpret this model to make predictions, recommendations and decisions etc. These algorithms form the basis of what we now describe as Machine Learning and Deep Learning.
Machine and Deep Learning
Most recent advances in AI have been achieved by applying Machine Learning to very large data sets. Machine Learning algorithms detect patterns and learn to make appropriate predictions and recommendations by processing data and experiences, rather than receiving explicit programming instructions. Moreover, the algorithms also adapt in response to new data and experiences to improve their efficiency over time. In other words, Machine Learning provides both predictions and prescriptions.
Therefore, Machine Learning relies on algorithms to analyse enormous data sets to be both predictive and prescriptive however it cannot think, feel or display any form of self-awareness or exercise free will, regardless of what Hollywood would have us believe. That level of AI is still some years away! What Machine Learning can do however is to perform predictive analytics much faster than we humans and without getting bored and thus should be viewed as a tool to help us rather than replace us. Thus, Machine Learning can perform the analysis, but we humans must still consider the implications of said analysis and therefore make the required moral and ethical decision.
Deep Learning on the other hand can be defined as a type of machine learning that is able to process a wider range of data resources, requires less data pre-processing by humans and can often produce more accurate results than the more traditional machine learning techniques. In Deep Learning, interconnected layers of software based calculations known as “neurons” form a neural network which can ingest vast amounts of input data and process it through multiple layers that learn increasingly complex features of the data. That is, the network will make a decision about the data, learn if its decision is correct and then use what it has learned to make better decisions in the future with regards to new input data.
5G and AI
The 2020s is very much the decade of 5G and even though 2019 saw the launch of numerous 5G networks around the globe, the majority of these were based on “Non Standalone Architecture” and as such, utilized the 4G core network. In other words, they simply provided data capacity boosts through a new 5G radio network whereas control remained firmly in the world of 4G. However, true 5G is a lot more than this and during the early 2020s, we will begin to see more “Standalone” deployments capable of supporting the key differentiators of 5G; Massive Machine Type Communications (MMTC) and Ultra Reliable Low Latency Communications (URLLC). Once these are truly available, along with the services that really require them, need for AI to process the data will be critical if we are going to support the 5G services that we as an industry, have been enthusing about for the past few years.
As we enter this new decade, we need to move from the “Internet of Things” to the “Intelligence of Things”; or in other words to support connected intelligence. Service providers cannot simply be focused on moving the data from “A” to “B” but instead, they must focus on the value and significance of said data. They must deploy AI to not only manage and optimize these complicated 5G networks but to also generate new revenue streams. History has taught us that if we simply build a data pipe, someone else will quickly develop the service to run over the top and thus capture the revenue. AI and 5G is no different.
Interestingly, the coexistence of 5G and AI appears to be happening already as indicated in a report from Ericsson[2]. This states that 53% of service providers expect to have adopted AI within their networks by the end of 2020 with a further 19% looking to do so within the next three to five years. Furthermore, the report indicates that the reasons service providers are doing so is to not only build new revenue streams (35%) but also to reduce capital expenditure (48%) and optimize network performance (41%) moving forward.
5G and AI working together …
5G and AI will form a “symbiotic” relationship in that each will enhance the capabilities of the other. For example, the integration of Multi Access Edge Computing (MEC) into the 5G Service Based Architecture (SBA) alongside the significant reduction in latency brought about by the new 5G air interface will enable Augmented Reality (AR) and Virtual Reality (VR) to become mainstream across both enterprise and consumer markets. Thus, video data from the devices will be rapidly transferred across the network enabling machine learning algorithms operating on the edge to interpret the information and return “intelligent” responses in near real-time.
Conversely, the complexities of the 5G radio will necessitate the need for Machine Learning and AI in order to dynamically optimize the millions of 5G small cells, not to mention the issues of introducing Massive MIMO and beam forming / beam steering. Network Slicing, a key enabler introduced in 5G, can also benefit from AI as the algorithms will enhance automation and adaptability by processing real-time information on network performance, quality of service provision etc, thereby supporting a more efficient, dynamic provisioning of a network slice.
Therefore, as we head into the 2020s and the widespread adoption of 5G, it is fair to say that AI will play an ever increasing role in supporting not only network operation but also the way we interface and utilize the new services that will run across it.
If you would like to learn more about 5G, our expert instructors provide live classroom and on-demand 5G Network Data Analytics course.
[1] OECD – Artificial Intelligence in Society (2019)
[2] Ericsson – Employing AI Techniques to Enhance Returns on 5G Network Investments (2019)
