Artificial Intelligence in Telecom – From Hype to Reality – AI

AI is having enough hype through media, researchers and vendors. Innovative organizations are putting a lot of efforts in AI research to make full benefits of it.

We all know about Sophia a social humanoid robot developed by Hong Kong based company Hanson Robotics. She looks quite human like and has been in talk shows a lot. She has also been given citizen ship in Saudi Arabia as being first artificial intelligence based human robot. She is being consider as intelligent but in reality, she is thoughtless whose intelligence is very basic. She can read a manuscript, on stage can perform a speech and she can answer questions that are pre-programmed. But If we ask her question out of script she can’t answer. Why is it so? Why is the gap here?

Well there is a basic difference of implementation AI.

Artificial General Intelligence

Machines that almost same level of intelligence that humans have, are expected in AGI products. This is what we all are expecting.

Artificial Narrow Intelligence

Machines have ability to perform specific tasks extremely well. These are computers that are trained to do simple and basic level of tasks only more efficiently than that of humans. This ANI based products are more in utilization by industry now.  i.e a machine that is trained to identify objects in images

A computer that can identify brain tumor cannot detect tiger in a picture because it has not been trained on that.

And certainly that’s the kind of AI, organizations like Telecom operators are trying to implement.5G  , IoT and big data will be handled by Narrow AI models to perform small task efficiently and fast.

Operators can train AI models such that, input from big data and after processing give expected results about customers. Mobile operators are also training the network based on alarms to predict whether there will be a failure or not. We just need to in put this data to AI model to learn and give out put as per desired results. It will also learn the relationship that if a new related input is given then it can predict the answer.  Similarly a new customer is added then machine will learn that client will churn or not.

This is very basic application of AI today in mobile telecom industry.

One more interesting and quite effective application of AI in telecom is dynamic carrier allocation.

 Artificial Narrow Intelligence trained model can add or remove additional carrier based on learning from previous weeks trends using capacity. No more manual allocation and wastage of resources.

Another example of machine learning using AI is

Automatize Customer Care:

Machines are trained to identify patterns in the text and predict problems for better solution offers.  

This is pure automation, quick and competent customer support application.

Challenges:

Challenges that mobile operators are facing at the moment is, they are not having enough samples of issues to predict more accurately.Only those queries are addressed well where model is having high volume data.

These are very concrete examples that are being used in Telecom companies today.We are seeing that AI will be a part of all domains in near future to automate tasks and improve productivity whether in Network, Customer care, Marketing, HR or Finance etc.

How to Adapt AI in Mobile industry:

First of all companies need to upgrade the technology stakes, upgrade the infrastructure and competence to work with AI. Especially need to fix the first mile and last mile of AI.

1^st mile means data readiness. To be able to collect, store, process to make available the data for training AI models.

Last mile is infrastructure on which AI based models work and utilization of predictions from AI models to operations.

So deploying AI in Telecom sector effectively, mobile operators needs to work on refine enough big data and then focus on applying outputs in business operations.

Key barriers that must be solved and must have things for AI readiness:

• Agile data Access Processes.
• Experimental platform with the right tool.
• Culture for experimentation and failing.  
• Domain and AI expert collaboration / cross function operational model