Recently, the TM Forum Community site, (our discussion forum for members and non-members) held its first ‘Ask the expert’ session. Artificial intelligence (AI) was the topic of choice, not least because of the huge expectations placed on AI in enabling digital transformation.
Our next session discussing ‘managing digital ecosystems’ will be held on Tuesday December 19 at 10 a.m. EST/3 p.m GMT. Put it in yout calendar and join us on the Community site within the Internet of Everything community, to ask questions and see how the session unfolds.
Our expert Brian Levy, Special Advisor to TM Forum and a telecom industry veteran, answered questions on AI in a 38 reply thread. Here are some of the highlights:
Tom Flynn, Neural Technologies: Are there any safeguards being discussed to address or prevent AI from learning and taking actions based on inaccurate or intentionally inaccurate data? Is it conceivable that an AI could evaluate the veracity of information before taking it into account, and would this even be desirable?
BL: This is a significant problem; we have seen AI systems being taken down for exactly this reason. There are companies now that offer data-cleaning services for training data. What is needed is domain knowledge and to build in some governing values. On what basis does an AI system evaluate the truth of data? This is a complex question. How do we evaluate fake news…not that well. Let’s hope an AI system can do better.
Richard Goldsworthy, Telstra: For our industry, where quality is measured in PPM and ‘always on’ is a non-negotiable service expectation, what are the control mechanisms being designed to ensure AI-enabled network management guarantees improved received customer experiences?
BL: Today the effectiveness of AI systems are measured by humans. Many AI systems are closed loop and are trying to constantly improve. However in the wider context the technology needs further development.
Sridhar Kuppurao, Infosys: As a process consultant, I am more driven by governance processes around any new IT trend. AI is relatively a new kid in the block, though old enough for some institutes. What are the governance mechanisms and bodies which control the AI at this point in time? To avoid any scare (as we see in the movies), how it is being governed? What are the rule engines in place as we speak and what is changing in these mechanisms for adopting the rapid growth and AI applicability across the world?
BL: I do not know of any controls on AI directly at this point. There are controls on data usage and access such as GDPR. However these do not stipulate AI access… As usual, lawmaking is behind technology development.
Krishna Basudevan, RankOne Consulting: Most of the AI machine learning also requires a significant amount of data for creating a learning model. Telecom has the availability for that mammoth data, but I have not heard of any telcos deploying a bot for responding on customer care help desk or interactive voice response (IVR) portals at any of the customer interaction points. Is it that for the transformation of customer experience, using AI not their cup of tea? It still seems they are submerged into their operational and tactical fight while OTTs are changing the industry landscape.
BL: Some telcos are actually deploying chat bots see this from Vodafone. But, it is early days for all this. I see three key initial areas for AI: customer experience, service management, and network management and optimization.
Anuradha Udunuwara, Sri Lanka Telecom: Do we really need AI? How to avoid AI ruling us?
BL: This is a complex question. In the more general sense do we need Facebook and social media. I think the genie is out of the bag now with AI and there is no way to put it back. It is coming and it will be very hard to stop it. The changes that AI will make to society are profound – self driving cars, call centers with no people, automated operations, automated doctors, etc. It will be more disruptive to society than the industrial revolution; it can be for the good of society or the bad, that is our choice at the end of the day.
For the telecommunications networks of the future, AI will absolutely be needed to dive automation, however the effects on society of AI will be far more profound than this
Can it be controlled? Well I think the first question will be: How can we avoid being too dependent on it? Imagine a world of self-driving cars, delivery vans, etc., and the system goes down. In the longer term, it will very hard to understand the answers that AI delivers, so working out how to control it may well be beyond our capacity.