Digital Water Transformation: The Promise of Artificial Intelligence
Will Sarni, Founder and CEO of Water Foundry
Camilo Huneeus Guzman, UNDP Perú, Gruber Fellow
Foto: Mónica Suárez Galindo / PNUD Perú
This article reviews the AI status in the water industry sector, proposes new AI applications for Digital Water and envisions a road-map for the future of AI in the sector.
1.0 What is artificial intelligence?
Artificial Intelligence (AI) and Machine learning (ML) are becoming part of our daily lives, but we don´t take a minute to study and understand it, letting the alarmist media spread fear and miss information. In this article, we will demystify what AI is by explaining it in plain English, and later we will share what we think are the opportunities and benefits that it offers to the water sector and companies that use water across their supply chain.
Artificial Intelligence is a broad concept that encompasses any activity performed by computers that humans would deem as ‘intelligent’. Machine Learning is the collection of statistical methods used by data scientists to ‘train’ and enable to “learn” without have been programed to perform a specific task. Train and learn, in this context basically means to adjust and fit data to statistical models.
AI is essentially pattern recognition and classification. The simplest form of AI, are the well known linear regressions, where a trend-line is generated and one can say -with some degree of certainty- if something belongs to the pattern or trend.
For example, The most common AI applications are targeted advertisements. When Amazon, Google or Facebook offers us an adds what they are doing is detecting our patterns of behaviour, following our clicks and ‘understanding’ what we like or dislike and using that data, AI can predict, predicting with a certain degree of uncertainty, what products we would buy. that we would buy the product.
More complex AI tools are Artificial Neural Networks. They are inspired in the way the brain works as millions of ‘nodes’ are interlinked. A node, is a unit or ‘point’ in a network that processes data. The simplest version was invented in the late 1950s, the perceptron. A perceptron takes inputs that are simply 0 (no) or 1 (yes) , adds them up and evaluates if it’s higher or lower than a threshold (bias). If it is higher, it will spit out a 1 (yes), if it’s lower, a 0 (no)
For example, Facebook can use a perceptron to decide if it should show you an add about cars based on the evaluating if:
You are a Nascar fan.
You live in a city.
Since Facebook knows that you are more likely to buy a car if you are an avid Nascar fan, this parameter will weight more than living in a city. Let’s says it weights the probability at 70 percnet. So, if you are an Nascar fan and you live in the countryside, the perceptron will do the following operation...
To read the entire article, please click HERE
This article was also published in PNUD Peru.
Click HERE to read the Spanish edition.