Among conversations about implementing AI-native networks, Dr. William Webb, IEEE fellow, CEO of Commcisive, Former Director at Ofcom argues for greater consideration
Despite efforts by various companies and even theories on how to implement it, there is still no single, clear definition of what an ‘AI-native’ network actually is. In fact, the very notion of something being AI-native is unclear, as its application remains vague, particularly with regards to its use cases within various industries.
For example, one of the key theoretical use cases behind 6G is that it should utilise this ‘AI native’ concept, with the intent being that AI can elevate the functionality of the network, resulting in greater capacity and lower operating costs. Of course, this is assuming that this cannot be achieved on the current 5G network, due to its design not allowing AI-driven optimisation.
However, in spite of the theoretical benefits offered by an AI-driven network, it is important to emphasise that this remains theoretical. The telecoms industry still lacks clarity on how to effectively implement these changes, as the potential benefits also come with potential drawbacks.
Additionally, with the recent advent of AI entering the public lexicon, many are keen to leverage this new technology. More and more industry experts are trying to find problems for AI to solve, as opposed to identifying current issues and coming up with the most viable solutions.
The case for an AI-native network
There is apt opportunity for AI to transform a variety of sectors, with customer service at the forefront. By using chatbots to provide customers with responses, automating this role often proves to be more cost-effective. Furthermore, AI can play a role in other areas where human error is typically common. One such example is fraud detection, where AI can identify unusual patterns of activity and flag them for evaluation by a human expert.
Likewise, there are also clear instances where AI can drive positive change within telecoms, with theorists having identified huge potential for AI native networks to drive transformation in two distinct ways:
- Improving traffic management: by forecasting when there will be a signal jam and there and identifying and re-routing the more important traffic
- Making networks more precise by enhancing network efficiency
Concerning network efficiency, AI has been identified for optimisation in three sub- areas. These include its ability to reduce the overhead in sending channel state information (CSI), as well as allowing for more precise beam pointing which further saves time and reducing errors by predicting exactly where error rates may increase consequentially the coding scheme ahead of time.
However, before experts even think about looking closely at these factors, it is important to firstly weigh up whether these benefits outweigh its complex implementation.
It is also worth noting that these benefits don’t carry much importance. In fact, as growth in data usage begins to slow down, it will soon reach zero. Therefore, it is not possible to see any new congestion, nor will there be a need for increased network capacity. If these are the only advantages offered by an AI powered network, then its value is not likely to offset the costs, which are likely to be significant.
Unpacking the need
Although AI has its values, this is not without additional costs. With AI systems mainly being ‘pattern recognisers,’ they function by analysing a sequence of data and drawing conclusions from them. However, these models are nowhere near foolproof and have the tendencies to make mistakes, also known as ‘hallucinations’. These tend to arise when the best pattern match is not the correct answer. While most of the time these mistakes are inconsequential, in telecoms, this is less so.
Additionally, even when working correctly, the insights that AI-powered networks bring – for example, by predicting congestion – tend to be of little value if there is nothing that can be done about it. With networks unable to allocate resources when needed, the best that they can do is to prioritise important traffic, thereby blocking lower priority traffic. AI cannot help in any way, as this is a task that requires human intelligence – as opposed to artificial – to determine what data is valuable.
As it happens, many other systems are already running as efficiently as possible, with CSI already compressed and ciphered. In addition to the advent of beam pointing algorithms, the correction coding systems already in place remain flexible and are, therefore, able to adapt to the current, evolving error rates.
This is all notwithstanding the implications of inaccuracies, which are significant. If the CSI information is flawed, then channel capacity will deteriorate, and if beams do not point in the right direction, then mobiles will lose their connection. An AI algorithm that is capable of reducing the CSI by 20%, but then renders the information 10% less accurate, is likely reduce the overall network capacity.
What does an AI-native future hold?
All in all, it still remains relatively unclear exactly what, if any, tangible gains AI can provide within these areas, and even if it does, they are likely to provide very little value.
However, this shouldn’t discourage leaders within the field from further exploring the role of AI in telecoms networks or other industries. In fact, it is always possible for additional advantages to be uncovered, for example, in other areas such as power reduction. However, making ‘AI native’ one of the core reasons for a new generation of cellular technology is a very loose justification for its implementation.
‘Author of the ‘5G Myth’ and ‘The End of Telecoms History’ Dr Webb is widely known as one of the most capable individuals in the communications arena. He was previously an adviser and Director of Corporate Strategy at Motorola, and former director at Ofcom, where he managed a team providing technical advice and performing research across all areas of Ofcom’s regulatory remit and led Ofcom’s Spectrum Framework Review. He now acts as a trusted advisor, strategist and change-agent for CEOs and senior Government members.
There’s plenty of other editorial on our sister site, Electronic Specifier! Or you can always join in the conversation by commenting below or visiting our LinkedIn page.
