AI and the MVNO Industry – Opportunities and Perhaps Mirages

The last 5 years or so have seen a lot of buzz in the world around the advent of Artificial intelligence. This buzz was supercharged around 2022 – 2023 when OpenAI released their large language model frontend, ChatGPT. We have written industry position papers on the opportunities that are currently available to players in the telecommunications industry that have been wrought upon by advances in technology. This article summarizes the possible opportunities and the toolkits that are unique to the MVNO industry that could be used to delineate these opportunities for further qualifications. The full paper on which this article is based upon is available on request.

The article argues that AI is a relatively old technology (at least the framework), gives 2 examples of the application of the technology outside of telecommunications and finally delves into the possible opportunities.

It is happening folks. Brace yourselves.

That AI will deeply impact our lives in some profound and sometimes personal ways is no longer cliche. Here in South Africa, on 14th April this year, we woke up to the news that the largest hospital group, Mediclinic, has put a freeze on new hires in the administration and offered employees packages to leave. In short, people are being replaced by AI. This is likely to be the case around the world regardless of whether companies are doing it loudly or quietly. But the effects are the same. The first shots in visible societal shifts wrought by AI have been fired. And we are only getting started.

Background

Historically, we have been able to identify the following as the salient features of the ongoing AI revolution:

• AI is almost as old as computing. If one assumes that the Colossus Mark I machine, which became operational at Bletchley Park in December 1943, was the world’s first computer, then it was only 5 years later that Alan Turing created the first framework for AI
• The Story of AI mirrors the Story of the Internet in its Original Funding. DARPA is credited with having originally funded the internet to a point where it was ready for commercial development and exploitation. Similarly, DARPA and later the Japanese Government played a similar role in development of AI technology.
• AI development was way ahead of the available computing capabilities at the time. This is a recurring theme starting way back in 1957 when the available computers could only execute the commands but could not store them. Similarly, DeepBlue that played the chess match against Gary Kasparov in 1997 was an IBM supercomputer which was generally not available for mass consumption at the time.
• The Internet provided a ready- made training platform with its multiplicity of text, video and images. The improvements in AI technology would never have been possible without this crucial factor.
• The current AI Revolution is simply an intersection of technological maturity (Hardware) and a long standing framework. It can be argued that the proliferation of AI is as much due to hardware becoming cheap and fast enough as to be deployed across a multiplicity of applications and by a wide pool of AI software developers and researchers.

Thus, the intersection of research funding from the USA and Japanese Governments, faster and better hardware, and ubiquitous internet have together enabled the maturity that we are seeing in AI today.

The story of Protein Folding and Alpha Fold

The human body is largely controlled by proteins in its complex functionality. Everything from where to insert the backbone in an embryo to regulating the amount of salt or water that needs to be released in urine is controlled by the structure of proteins. Thus, understanding the nature of proteins is central to understanding the cure for most genetic diseases as well as designing new drugs, vaccines and therapies. Proteins are complex molecules that are made by 20 or so amino acids in the human body. These amino acids in turn are able to create approximately 200 million protein structures. The process by which amino acids build proteins is known as protein folding. It is a physical process whose resulting molecular structure is key to determining the function of the resultant protein. Until around 2015, this process was studied through physical processes. This was a laborious process and time- consuming, with no hope of ever being completed in this lifetime. In 2020, Alpha Fold (part of Google) AI completed the process releasing the resultant database of 200 million proteins to the scientific community. This opened exciting opportunities in the world of medical research that could only have been dreamt of, 5 years before.

In our opinion, AI at that point had come of age, solving real world complex problems (as opposed to being used as a college plagiarism tool). And that was just the beginning. And the most amazing thing is that this earth shaking breakthrough happened away from the glare of world headlines that were afforded more mundane innovations in the AI field. (1, 2)

1. Deeper treatment of the protein folding problem, however fascinating, is beyond the scope of this paper. To those with greater interest, we recommend among others, Mukherjee, Siddhartha. The Gene. Vintage, 2017.Deeper treatment of the protein folding problem, however fascinating, is beyond the scope of this paper. To those with greater interest, we recommend among others, Mukherjee, Siddhartha. The Gene. Vintage, 2017.Deeper treatment of the protein folding problem, however fascinating, is beyond the scope of this paper. To those with greater interest, we recommend among others, Mukherjee, Siddhartha. The Gene. Vintage, 2017.
2. In September 2023, the Computer Scientists behind Alpha fold received the Lasker Price,  a prestigious award in sciences, leading to speculation that in future, some AI could win a Nobel prize in Science or Economics.

What are the salient features of the ongoing AI revolution?

It is a Scientific Revolution

Looking at the number of scholarly articles on AI as compiled by Center for Security and Emerging Technology (3), the amount of AI research seems to have accelerated from roundabout 2016, although the scientific output at that time was still impressive. Thus, it can be argued that the rapid adoption of AI technology over the last 10 years or so is a result of actual scientific and technological advances.

It is a Statistical Model!

Outside of Artificial General Intelligence (which has not been achieved yet), the performance of any system will be as good as the training data. Statistical Biases will be transmitted and show up in the results generated by artificial intelligence. Moreover, the results will be generally good, but not outstanding, given the tendency of AI training models to revert towards average.

Increases Capacity to Aggregate Knowledge Several Folds

This ability to aggregate knowledge will likely have an impact on the telecommunications industry both positively (improved products and services) and negatively possibly through constant changes that promote instability.

An Emerging Security Nightmare

Just like every aspect that is touched by AI touches, security is going to be supercharged. It is becoming faster to probe and find vulnerabilities and easier to exploit them. This combined with the increased levels of online interactions make a case for a nightmare scenario for security practitioners.
The corollary can also be true in that AI will enable researchers to plug security holes faster than they can be exploited. Sadly, the old adage that says the thief is always a few steps ahead of the cop may ring true in this case.

It used to be Content is King. With AI revolution, data is King

The AI revolution is only as good as the final product that it produces. The promise of AI is to discern hitherto unknown patterns, aggregate knowledge at a supercharged level, and continuously learn in view of any new knowledge in the wild. Thus, at its core, a good AI model will only produce results as good as the underlying training data (at least until Super Artificial General Intelligence Models come into the market). And a fair amount of the data found in the Internet is of suspect quality.
Thus, the best-trained models will be those that have access to the best data. This is likely to set off a data ownership arms race that will be ultimately mediated by the regulators world over.

It is directly responsible for an unfolding Social Revolution

Away from technology, AI will be directly responsible for a lot of social changes world over. By far, the most serious will be widespread loss of jobs and subsequent retooling of existing ones that will be retained. AI will also be able to supercharge all the positive and negative aspects of the society, including hate, bigotry, and fake news. Quite likely, these aspects will have an unforeseen impact on the democratic outcomes in countries that practice it.

What are the expected major impacts on the industry

Opportunities are likely to come with complexities that come about in delivering novel products and services. Moreover, any big technological shifts such as those expected out of the AI revolution will quite likely lead to unforeseen risks.

Heavy Demand on Data and Computing Resources

The markets have already figured out that the demand for computing power is going to be heavy in an AI revolution. On May 30, 2023, NVIDIA, who are makers of graphics processing cards passed the 1 trillion dollar capitalization. This was a recognition of the market that the types of processors manufactured by NVIDIA will be in much greater demand as the AI revolution took off. Of more immediate importance to the telcos, most of whom host various applications on behalf of their enterprise clients, is that there will be an immediate need to do a complete review of their processing infrastructure and assess the impact of AI-powered applications on it.

The demand for data on the other hand will grow exponentially as various market intelligence applications try to gain competitive advantage. Data is one resource that telcos have in abundance. They would well be advised to use it judiciously.

Recalibration of the BPO Market

Over the last 25 years or so, a strong Business Process Outsourcing market has sprung up in developing countries being used for just about every repetitive chore ranging from customer service to digitization of medical records. The earliest motivation for outsourcing beyond cost savings was the fact that the tasks that were being outsourced were repetitive and thus likely to be of low value. By coincidence, this same attribute makes it easy for the tasks to be replaced using AI.

Telecommunication companies are the biggest enablers of the BPO industry whether this is through the provision of connectivity or data centers. In case of an AI takeover of the BPO markets, the actual processing will take place in the cloud or the edge, meaning that the connectivity will be on the minimum be redirected and at worst be rendered redundant. And a bigger risk is that cloud computing will mostly mean that the hyperscalers will be taking revenues directly away from the telcos. It is a risk that will need to be addressed.

Market Fragmentation as Micro Targeting becomes economical

The traditional telecommunications model is saturated and has consolidated substantially over the last 10 years. Thus, there is little space for innovative disruptors using the current model. This is made worse by the large capital outlay required to set up traditional telecommunications infrastructure. One play that could find its way into the model is MMC play. Traditionally, MVNOs have struggled to make economic sense out of the low margins offered by the host operators and more so, at low volumes.

One expected impact of AI is that specific social / commercial groupings will be able to micro-target at a much- reduced cost given the improved analytics borne of AI. This in turn will lead to lower marketing costs and potentially, market cannibalization. The strategic play for large telcos who occupy second or third play will be to ride with the wave and disrupt while that of the leaders will be to protect the status quo, which could be risky.

An Empowered Consumer able to use AI to analyze value

Traditionally, large telcos have created value by selling products that are fairly complex to distill the pricing for consumers. These products are nearly impossible to compare side by side with competing products, thus enabling the consumers to move across networks in search of value. Moreover, the consumers are bound by long- term contracts that are normally tied to a sale of a handset.
It is quite likely that a lot of new powerful AI- driven applications will be released to the markets for telco consumers to reduce their telecommunications costs. If these products gain traction, a strategic risk to the telcos will be realized and will need to be mitigated.

New Products whose commercial and Billing Models have not been thought of yet

A multiplicity of new products whose natural home is a telecommunications network are likely to be spawned by the AI revolution. In our opinion, the following sectors are most likely to spawn new product opportunities for the telcos:

• Online Games: AI is already changing the way online gamers play, introducing new games with different computing and bandwidth requirements. The challenge for the telcos will be to figure out how to create viable and chargeable products out of the new online gaming community
• New Breed of AI generated Content: Once intellectual property rights issues around AI- generated content have been cleared, the next logical step will be to monetize them and telcos are well positioned to at least bill the content off the telecommunications bill
• AI Powered Marketplace: As Telcos reimagine themselves, one business model that seems to be getting pursued is that of a digital marketplace. In this model, Telcos are offering their traditional voice and data services out of this marketplace and then extending it to new products such as the app market. Maturation of the AI model has a potential impact of increasing the number of products that can be offered on this marketplace as well as its overall management

So, what are the novel opportunities for Telcos and by extension, MVNOs?

Is Edge Computing an inevitable consequence of AI? Does it provide an opening for Telcos to Wrest the Initiative from hyper-Scalers?

AI applications are resource-intensive. This applies to both network and computing resources. Thus, the most optimal deployment for AI applications will be at the edge, rather than the deep cloud. The edge is one area where the telcos have a distinct advantage over the hyperscalers. They are generally closer to the end customers hence more likely to take advantage to proliferate edge applications. Of course the hypers-scalers could (and possibly will) simply drop boxes at the Telcos’ data centers as a paid service for their edge computing applications. That of course will still mean extra revenue for the Telcos. It is thus imperative that the Telcos develop the optimum strategy with which to approach edge computing opportunities but with one eye on the hyperscalers.

Deeper Insights into your Customers….some, will shortly after be borderline illegal

Telcos own customer insights in raw data that most jurisdictions in the world require that is stored going back 5 years at least. Older data is archived. In the coming AI-powered data revolution this will be a Gold mine for customer insights that would be of use to just about any sector that is involved in the retailing of goods and services. And Telcos stand to gain by setting up businesses to monetize this data.
Legal and philosophical issues are most likely to determine how this opportunity pans out. Already, data protection laws such as GDPR in the EU regulate how this data can be used. With the AI revolution, there is an increasing clamor for regulation around the use of the technology and by extension, the underlying data. Thus, whereas business opportunities exist and monetization models are in their infancy, the regulatory environment is less so. On the minimum, Telcos can start by asking new and existing customers to give consent for their data in an anonymized form to be used in the analysis of consumer insight.

Easier (and cheaper) prototyping of telco products Before Market introduction

Product design for Telcos can be a very expensive process involving changes to the business and systems process adding to the costs. Thus, the price of failure can be very high and in some instances, discourage product managers from being innovative, to the detriment of the Telcos’ financial performance. In addition to large amounts of data held internally by the Telcos, AI will assist Product Managers in simulating their products long before deployment and get insights closer to the market conditions. This in turn will in turn save on costs and at the same time improve on the probability of success.

RAN and Transmission Network Profitability Analysis (for thick MVNOs)

Radio engineers do not necessarily take into account the profitability of individual towers or tower groups. The decision to build or not is usually based on availability of site as well as the requisite permission. With AI, it will be relatively easier to conduct an analysis of the profitability of a proposed site using historical data combined with static demographic data. In this way, the decision to deploy a radio network will have a second parameter to work with apart from service quality.

Dynamic Network Congestion Management

One intractable problem with this impact on the customer service level that Telcos face is congestion, and seemingly, there is no real solution apart from simply adding more resources. This may not be economically efficient given that some of these congestions are transient, say driven by sporting/ entertainment/events or road traffic patterns. One simple solution would be to offer discounts to the neighboring towers to where the congestion is happening and broadcasting the service discounts. This will be made all the more easier through AI.

Multiplicity of AI + Maybe 5G+ Products

The next evolution of radio network standards for 5G contain an element of using AI in radio congestion management among others. So far, the only equipment manufacturer who has incorporated the same is Huawei in an announcement made in 2023. The vision that the people who crafted advanced 5G standards were to use AI in the network and at the same time spawn other types of products out of this. A simple example of such products would be a road traffic management tool that takes its cue from network traffic. This is a product that could be developed and sold to motorists.

In Conclusion

Technological sea change such as the one being posed by AI is always wrought with opportunities and risks. For established Telcos, the risk is that newer players in the market would move faster and create competing products, hence eroding the market dominance. For challengers, there are plenty of opportunities in AI to upend the market and take the fight to the dominant players. And for the entire industry, there is an opportunity to take back some of the market that the hyperscalers.