Responsible AI for the water sector?

The world's leading Artificial Intelligence summit, 9-10 October, Zaandam

Artificial Intelligence has penetrated many aspects of our lives, often without (most of) us realising that this is the case. Society is starting to push back at the AI technology push, mostly as a result of scandals such as the Cambridge Analytica case and the growing unease for things like automated facial recognition in public places. The AI industry is becoming and has become very aware of this.

This was made very clear at the World Summit AI (“the world’s leading AI summit, 9-10 October, Zaandam), because if a central theme were to be distilled from the presentations at this event, it would be responsibility. And even if there is a business motivation in addition to an ethical motivation for this increasing awareness, it is beneficial nonetheless.

As part of the first World AI Week in Amsterdam, the wsAI event brought together thousands of AI researchers and practitioners from around the world. Many high profile companies (Microsoft, Facebook, Airbus, Shell, to name a few) were presenting their applications of and successes with AI but were also making clear that they are part of and have an impact on society. Not only in the opening speeches (Prince Constantijn van Oranje referring to responsible use of AI, Accenture’s Irine Gaasbeek speaking of an “AI summit for good”, Marc Carrell-Billiard, also Accenture, naming “more responsible AI” as one of the three challenges ahead), but also in many other, more in-depth presentations. Most notable was a presentation by Dagmar Monnett from the Berlin School of Economics and Law, who presented a framework for creating responsible AI.

But of course, there was also a lot of talk on technology. New York University’s Gary Marcus warned that the current AI systems work based on correlations rather than comprehension and that too much is entrusted to technology that is immature and incapable of performing well beyond a small set of particular tasks. A compromise between two AI traditions, the classical based on abstraction, and the current paradigm based on learning, is necessary – they need to be brought together to develop the broadly and safely applicable AI of the future.

Several presenters showed exciting results from the application of AI, ranging from medical applications (Kirontech), through fraud detection (Mastercard) and deforestation monitoring and water management (Airbus), to the detection of emotions in pictures of humans (Affectiva).

So what does this all mean for AI in the water sector?

So what does this all mean for AI in the water sector?

There are two critical aspects to the question of the responsible application of AI in the water sector. To begin with, the sector is lagging behind other sectors in the application of AI. The upside of this is that we can learn from the mistakes that others have already made. The second aspect is that as the water sector offers public services (and is publicly managed in several countries), there may be fewer financial incentives for “irresponsible but profitable” applications of AI. Nevertheless, the sector has had its issues with the data that can be fed into AI (for example security issues with smart water meters), so continued vigilance is necessary.

The focus of current AI research and development is very much on machine learning, as was also evident at the conference. However, other branches, such as Optimization, Expert systems, and Robotics (referred to in the presentation of Marc Carrell-Billiard) are at least as necessary for the water sector. It is clear that machine learning may play an essential role in, e.g. identifying and reducing leakage. But optimisation is starting to help us design better-performing systems for less. Expert systems may be a way to consolidate the knowledge of a knowledgeable but ageing workforce moving towards retirement. And robotics will allow the sector to obtain information on the condition all its assets, including the large fraction that is buried, thus generating the data which can be fed into machine learning algorithms. The movement to join the classical, abstraction based paradigm and the learning paradigm that is now is fashion, as advocated by Gary Marcus, can be expected to boost these applications in the water sector.