AI for Smartness: Moving Up the IoT Value Chain


Curated by Vinay Prabhakar Minj 

“We should not just focus on making products smart as the end goal, we should  also move up the value chain”

The combination of Augmented Intelligence (AI) and Internet of Things (IoT) is something that is very much talked about now and is touted to be the future of technology industry. Thanks to these emerging technologies, things will change a lot in the next decade, improving our standard of living.

Today we have already entered the fourth revolution of industry (Industry 4.0). If we look back, the maturity curve started with the steam engine and the industrial revolution. After the discovery of electricity, we started doing things in a better way. In the late 1700s, late 1800s and mid-1900s, that maturity curve continued to grow. And when we finally entered the fourth generation in the early 2000s, the entire misnomers that we were being previously talking about changed. Now, people talk about IoT, AI and cyber-physical systems that can self-correct, take decisions on its own and make a difference.

How IoT can add value to a manufacturing company

In several manufacturing facilities, you may see old machines that have been running for more than 40 years along with some not-so-old ones which got added at the end of every decade. They all belong to different generations. While some may have an excellent framework and some could be controlled from a cloud, there may be machines that need to be attended by a technician to make it work. But since this entire data is not stored in the cloud, it is not easy to replace such machines altogether without showing that there is a big value to it.

To solve the above problem, we first need the Enterprise Asset Management (EAM) system which has the maintenance log of that particular machine. All we have to do is to pull out all the previous maintenance records, take the data from these machines and put it into a simple time-series model.

By doing this, you can predict the maintenance cycles or warranty of the machine. This way, a manufacturing company was able to save about 17 percent operational cost. This is the value IoT can add to a manufacturing company.  

The term smartness cannot be generalised, but technology like neural networks can help in infusing more intelligence into the system and raise the level of smartness to a completely different level. 

IoT and data

The smarter ecosystem concept is all about data, and when you talk about data, it is all about IoT. We analyse the data to find patterns, which help us to create predictive, prescriptive and descriptive models, i.e. AI or Augmented Intelligence. And when the model becomes self-correcting and adoptive (meaning when it is able to correct the machine by itself pre-empting the error conditions), you have achieved a successful system. This collaboration makes the process much more seamless and intelligent.

Smarter cities

Everybody talks about smarter cities nowadays. The concept of smart cities started around a decade back. From the past two years, this concept is being talked about a lot in India. Still, this concept is at a basic level and hasn’t blown out of proportion in changing the world. Some reasons behind this are socio-cultural causes and disparate data sources.

A major reason hindering the growth of this concept is our inability to talk about it on the same platform. For example, a smart city proposal coming from Singapore will differ from the one coming from Bangalore. The two proposals talk at two different levels. The concept coming from Singapore will begin with providing fundamental facilities like 24 x 7 electricity and water. But this concept can only be an end-goal from the Indian perspective. So, how smart am I is my own definition and that’s where the cities differ. To be precise, every city intends to become smart in their own ways and so there cannot be a common smart city proposal and pattern.

The term smartness cannot be generalised, but technology like neural networks can help in infusing more intelligence into the system and raise the level of smartness to a completely different level.

How to realise the full potential of IoT and Industry 4.0

Smarter cities are all about instrumentation (IoT, sensors), interconnection (edge computing, cloud) and intelligence.

In order to realise the full potential of IoT and industry 4.0, we have to move from smartness to the value chain. We should not just focus on making products smart as the end goal, we should also move the value chain. To achieve this, it is to be ensured that the system or the technology we build is self-reliable, self-correcting and self-working. And it should work with human collaboration, not with human support. It should think along with us, and not depend on us to take a decision.  And we can achieve this with the combination of IoT and AI.

Achieving smartness with the help of different technologies in various industries simply comes down to taking data (implementation of sensors and intelligence), finding patterns (visualising the patterns with the help of data analysts and software) and putting it into use (preparing a time series plot/chart and making a prediction).

How to do it?

  1. Gather data for intelligence (anybody can do it these days as everybody has sensors).
  2. Visualise the patterns. You don’t necessarily need a good data scientist to do the visualising of patterns. Only a data analyst is required. There are also systems and software that can find patterns out of data.
  3. For repeating patterns, do the predictive analysis by putting the varying patterns in a time series plot. The everyday conditions have to be factored in and need to be self-corrected. This self-correction can be done by recurrent neural networks or by creating LSTM (Long Short Term Memory) models.

Automated Analytics Process for Saving Energy

First you need the domain knowledge (whether it is about traffic, smart cities, energy), then you have to map the metadata and create a cyber-physical graph (for modelling the system in the cyber world, also known as a digital twin). Finally, you have to put the knowledge graph (analytics) to complete the process.

Energy is always limited. India is energy and water-stressed country. The losses in generation and transmission of electricity are close to 27 percent – 30 percent in our country. The biggest consumers of electricity are buildings, malls, hospitals, homes, etc. Hence, builders are promoting the concept of smarter homes/buildings.

An initiative towards energy conservation is the 2030 plan, where 30 percent of electricity or energy saving is mandated by every building. Many countries are going after this because there is a limit to energy production.

In conclusion, IoT helps in providing data for taking decisions. But the intelligence required for using that data to make those decisions can be achieved through AI.

Q & A

  1. What is your take on AI ethics?

Ethics is a must have and we cannot do any unethical practices. We should not leave the decisions entirely on the machines. Rather, the machines should auto correct with human collaboration. Personally, I believe that every machine should have a humane touch as well. The intelligence of any system is dependent on how we are building and training it.

2. Biased data/training can corrupt the whole AI system. Do you agree?

True, biased data or wrong training can corrupt the whole AI system because the system does not have the intelligence to understand whether or not you are intentionally doing it. When computers were introduced in India or when the industries moved towards automation, there was a lot of resistance with a fear of huge job losses. But over the time, we have come to understand that jobs will need to be more and more specialised. So, the more you get specialised and the more you correct your systems, the more it will move up the value chain and behave better.


About the author

The above article is an extract from a speech presented by Alex Jojo Joseph, Program Director, IBM, at IEW/IOTSHOW.IN 2019.