Emerging AI tech is not a magic bullet, it is the latest bullet that requires hard work to use effectively.
Technology-driven industries have a habit of believing the latest buzzword and trend is the universal answer to all existing problems. It has happened in telecom with the promise of 5G and the best recent example has been the metaverse.
When it comes to AI, we must avoid falling into the trap of believing it is the answer without understanding the hows, why and the whens. We must also avoid seeing AI’s potential through just one lens.
From generative AI advancements to new modeling breakthroughs, AI has significant potential for telecom. It could quite possibly transform how we do business. But only if we change our thinking from AI being a noun to a verb or from an it to a do.
AI is not unlike the journey of automation, with a key difference being how much autonomy in decision making can be left to machine versus human interpretation, and how much we trust machine interpretations that are independently reached.
At its most basic core, AI is simply the next software algorithm, albeit the most powerful to date. It allows us to answer questions without knowing the question. It is a tool that will absolutely change the landscape of today and will be used to continuously go faster, cheaper.
AI is not new, but it is now more sophisticated and powerful due to the scale of the computing it can use to generate models and the new algorithmic approaches it takes. Previously, we made machines intelligent with rules-based automation using if/then statements in code. This will be replaced by AI models that have been trained on large datasets, and allow a different level of insight, intelligence and decision making.
Generative AI will eventually transform any process or task that relies on humans extracting knowledge from documents, leading to a more than 8x performance gain. Adoption of various AI modeling to automate networks will lead to level five autonomy in the future.
It is now within reach to automate the job of watching and reacting to network changes. The ultimate goal is to be able to analyze those same large, unified data sets. This is the fuel of machine learning and AI that will lead to the possibility of autonomy.
Automation is Rakuten Symphony’s holy grail. Openness and cloud are critical to what we do but they exist to advance automation efforts that help us go faster by making everything accessible and programmable based on a unified data understanding that AI models can be applied against.
It took us four years to make the progress we have on automation in Japan. We believe that others can combine AI and what we’ve learned to do the same in as little as one.
The most substantial progress we’ve made condensing timelines is leveraging time and commonality. We recognized the need to build a unified data model that was vendor-agnostic and gain the ability to interpret what different data signals meant. Without this, we wouldn’t be able to draw any kind of inference from the data as we’d essentially be trying to get different systems to speak different languages to each other while out of sync.
One of the most challenging elements associated with adopting AI on the journey to automation is building trust in systems meant to mitigate manual intervention. Developing the confidence to close your eyes and take your hands off the wheel is a journey in and of itself.
When we started, everything was open loop in terms of decision making because we were still establishing confidence in the system. Basically, humans checked everything.
Over time, we recognized patterns across network and operations management. We trained our systems to address these repeat requirements, eventually graduating to a mind state where we could trust the system to do exactly what it was designed to do.
This trust did not come overnight, but we believe it can be established faster for other operators seeking to do the same, especially in common deployment scenarios where we have already proven automation for work.
For operators seeking to make fast progress on this front, we suggest starting with automation's low-hanging fruit, identifying simple problems such as sleeping cell situations. In this scenario, the issue would not normally be detected right away and likely require an investigation. Eventually, the loop is closed, reducing it to simply a blip on a performance graph that doesn’t require human intervention.
It’s not about being the best, but being better every day.
There is no easy answer, for personal development, for company progress, for social improvement. There is hard work to understand the latest opportunities to be better. The key success factor is not the latest opportunity, but what somebody chooses to do with it.
Unfortunately, telecom doesn’t have five years to take its time to get this right. The industry needs to understand today where it is headed, how it will get there and take the first steps now.
Telecom has to ask itself: why aren’t we delivering twice as fast for the customer versus what was possible yesterday. Why are we waiting for customers to push us instead of finding the motivation to move quickly on our own?
A major mindset shift is needed. Aggressive goals should be set for telecom to fully automate its core business. So rather than spend 70-90% of time and energy running a network, we can spend it on solving customer problems with a network that runs itself.
Speed, and the fate of this industry, hang in the balance.
