The world of IoT and digitalization has fast become a reality for the water sector, with much excitement and hype surrounding the possibilities this brings. However, it has also created misconceptions surrounding artificial intelligence and its machine learning capabilities. ATi’s Executive Director, Garry Tabor, discusses whether expectations have been set too high and what crucial foundations must be laid to make digitalization a true success.
With an ever-increasing frequency, new buzzwords or acronyms burst onto the scene and quickly become part of every-day language, promising exciting and innovative ways to improve the way the world operates. The most recent examples include “smart”, “cloud”, “AI” and “digital transformation”, but unless we stop to define the true meaning behind these fuzzy terms, it creates misunderstanding as to how these concepts will work within the water industry. People then feel slightly embarrassed to admit their lack of true understanding because everyone else ‘seems’ to know. A classic case of the Emperor’s New Clothes.
Many have heard of the fairy tale where people pretend to admire the new clothes of the Emperor, despite the fact he isn’t actually wearing any. However, there is a real possibility that AI in the water industry could become much like the well-known fable, appearing to promise spectacular capabilities that many want to believe is true, yet the reality is often a misconception. Nobody wants to think of themselves as the Emperor, however, when it comes to the much buzzed-about topic of AI, too many of us are unwittingly playing that part.
For many companies in the water sector, collecting and analysing performance data in order to unlock operational insights is the driver of efficiency and innovation. Yet getting to the pot of gold at the end of the AI rainbow isn’t as easy as some would want you to believe, with too many myths that are creating unrealistic expectations.
AI is typically a platform that creates algorithms, but the water industry is not quite ready for the autonomous intelligence this requires – at best we are only at the stage of machine learning, pattern recognition and reproducibility. As all water specialists know, water is a complex and dynamic chemistry that is fluid and ever-changing. It is not a mathematical algorithm; it holds to the universal law of cause and effect. Moreover, there is also mother nature who often defies predictability. In a similar way to weather, water is inherently difficult to predict and manage by computers alone – whatever you put in needs to be interpreted by a data expert. Water companies cannot simply rely on the correct coding to monitor and measure water quality, and this is the first misconception of AI.
Just like the way humans need to walk before they can run, machines need to be taught how to learn. Algorithms need to be trained to do what is needed; but to train an algorithm you need to feed it with clean, reliable and accurate data from strategically placed smart sensors, along with the expert knowledge of data specialists to interpret it. In short, if you put rubbish into the smart system, you will only ever get rubbish out and even the smartest analytics platform could not succeed without the perfect combination of quality data, smart people and efficient processes to take those insights and turn them into meaningful actions.
For these powerful analytics tools to have any beneficial impact, they need to harvest high quality data throughout the process from source to tap. Purely mathematical models have many data blind-spots; but a neural network of smart sensors can provide the solution for this, creating a complete picture and enabling effective water quality management. Carefully selecting the right industry-leading smart sensors as the foundations, such as ATi’s digital, smart water quality M-Nodes, is vital – only then can digitalisation start being built.
There are plenty of factors that might limit the adoption of AI, including skills, operating model and accessibility of data. AI apps and services rely on data, and a lot of it. Accessibility and quality of data is key to make the journey to digitalisation a success and involves:
Discoverability. Do we know what data we have and how its structured?
Access. Can we get to this information with reasonable ease and economically?
Data Exploitation. Do we have the tools, skills and technologies by which to process the data for our chosen purpose?
Although these may sound obvious the water industry, with its legacy of technologies, needs re-think its strategies and become innovative in its approach, moving towards these data foundations goals.
Making the future of digitalisation a reality
However, due to the amount of analytics programs and data visualisation tools being marketed, a lot of confusion has been created and many water companies still claim they are ‘data rich, information poor’. This is either because their sensors are not as ‘smart’ as they claim, or it is due to a talent gap in specialist data analysts who can interpret and action the data. Water utilities know that it is not more data that they need, but clear, actionable insights. Automated data collection, through the use of smart, reliable sensors, and the ability to upskill staff to analyse and understand it, represents a change in this approach to unlocking insights to drive efficiency. In other words, to make digitalisation a success, the industry needs to focus less on the myths of AI and more on the famous PPT paradigm: People, Processes and Technology.
The future adoption of digital water could be at risk by AI expectations being set too high. Innovation and the adoption of smarter digital solutions will bring about rapid and dramatic benefits, but if not challenged, these misconceptions risk slowing the deployment of proven technology down and ultimately risks destroying the promise that such innovation offers.
So, is AI the future of smart water quality management? If we change the meaning of AI to “Actionable Insights” rather than “Artificial Intelligence”, then the answer is yes. While the vision of digitalisation may take time to materialise, it is the decisions made today, laying the foundations with the right smart sensors, collecting data and applying pattern recognition, based on specialist industry knowledge, that will determine how quickly we get there.