You are here

Artificial Intelligence and the Challenge of Sustainability

Written by Faizal Rohmat, a 2017-2018 Sustainability Leadership Fellow and Ph. D. Candidate for the Department of Civil and Environmental Engineering

The challenge of sustainability

Today we are faced with the monumental challenge of sustainability. By the middle of this century, the United Nations predicts the earth will be inhabited by more than 10 billion people [1]. This high rate of population growth puts serious pressures on ecosystems, both wild and agricultural. One example of how our population strains resources is the increase, nearly 100%, in our grain demand [2]. Attaining sustainable outcomes is even more important because the world is being impacted by our population today. Unless something unimaginable happens, human population growth is inevitable. To accommodate such growth, humans must be able to manage their interaction with the ecosystems in sustainable ways. This means that current demand on resources must be met without compromising the supply of resources that will be needed for future generations to meet their needs, as sustainability is defined by WCED [3].

Pryshlakivsky and Searcy mentioned that numerous efforts to implement sustainability over the past two decades on a national, regional and organizational scale have been generally less than satisfactory [4]. Researchers refer to this failure of to implement sustainable development, or sustainability in general, as “complexity and uncertainty of natural and social phenomena” [5], “seemingly random interpretations and competing frameworks” of sustainability [6], and the “difficulties experienced in implementing any holistic approach” [7]. The challenge of sustainability is a wicked problem because it lacks clarity, incomplete, contradictory, and has system-of-system complexity [4]. We need to solve this sustainability problem in a way better than we have for the past two decades.

What is AI, its history, and where we are

Artificial intelligence (AI) could be the way to solve the wicked challenge of sustainability. AI is the intelligent behavior performed by the machine, as opposed to the behavior of natural intelligence performed by living things. The AI discipline first appeared in 1956 through the Dartmouth Summer Research Project on Artificial Intelligence [8]. In the next decade, research on the idea of AI blossomed, followed by various disappointments in the late 1970s, then reached its nadir point at the "AI winter" in the late 1980s [9]. Followed later in the 1980s - 1990s with research focusing on machine learning, i.e., ways to achieve artificial intelligence. Then in the early 21st century, with advances in computer software and hardware, promotes we saw the growth of AI research, development, and use in several disciplines. Today, AI has become a part of our daily lives, such as Apple's Siri, Amazon's Alexa, IBM's Watson, and even the driverless cars that will soon fill our streets [10].

The AI forms we have used in our daily life, according to Stephen Hawking, are still the primitive forms of AI. Nevertheless, it is enough to worry him about the dangers of intelligence, one of which includes the has the power to either destroy humanity [11]. Even back in March 2016, the world was shocked by the defeat of Lee Sedol, 18-times world champion Go chess game, by AI AlphaGo with a score of 4-1 [12]. Deep blue had beaten human chess grandmaster two decades ago, but the Go victory was special because Go chess is more analytically hard to crack for the computers and requires a more human-like way of thinking [13]. This breakthrough in the ability of a machine to compete against and beat humans in a game that challenges human analytical ability brings anxiety regarding evil AI robots like those in the Terminator films.

On the other hand, the emergence of AI also presents us optimism. We have a long history of success being innovative. Fire, domestication, steam machines, electricity, cars, smartphones are just a few examples where we have used technology to benefit our species. We had fears of strange and powerful things until we eventually managed to tame them and utilize them to achieve our objectives. Now, we are going to do the same and use AI to solve the wicked challenge of sustainability.

AI, machine learning, and how we can utilize them

The first cause of dissatisfaction with sustainability efforts is the complex nature of sustainability issues and our lack of understanding of how ecosystems work. The ecosystem is very complex; everything interacts with everything. We do not know exactly what the consequences of what we do. But, as Pedro Dominggos says in his book "The Master Algorithm", with more sensors and more data we have now through the big data blast combined with better machine learning, we are able to create better models so we can understand better about how the whole ecosystem works [14]. Dominggos even calls "the automation of discovery" as one of the definitions of machine learning. Machine learning helps us automate the discovery by learning from the existing knowledge, fill in the gaps, and systematically reduce uncertainties [14]. This is basically traveling down the road of discovery by cars instead of walking. Machine learning, as a subset of AI, can amplify what we already know and accelerate the pace of discovery to make us better understand how the ecosystem works.

The next point about the challenge of sustainability is the difficulty of implementing a holistic approach to addressing the challenges of sustainability. The approach that has been done to address sustainability challenges is through computer modeling to gain understanding and simulate scenarios of the specific components of the ecosystem. Currently, we already have computer models of specific components of the ecosystem, such as the model of ocean currents, atmospheric models, or models of animal diseases. However, we still lack a great way to combine these specific components into a holistic model. AI, especially through machine learning approach, with the characteristics of input-output mapping, adaptivity, very-large-scale-integration implementation, and neurobiological analogy [15], can help us combine individual models into a more holistic ecosystem model. An example is what Triana did in Lower Arkansas River Basin where he simulated the effects of agricultural practice scenarios to the sustainability of the irrigation valley by utilizing machine learning techniques for coupling groundwater and stream water models [16]. His research has proven to be way faster than the classical approach of groundwater-stream water model coupling. This example of coupling two individual models can be scaled to any numbers of individual models. With a more holistic ecosystem model, we can better understand ecosystems as a whole and see what happens to the ecosystem if we do different things. We can model scenarios that are of great benefit to us while minimizing the side effects on ecosystems without doing them in a real situation. With the application of AI, we can see further down the road and gain public support for proposing the sustainability solution.

In conclusion, we need to address this sustainability challenge in a better way than we have done so far and AI has great potential to help us answer those challenges. AI can amplify the knowledge we already have about the ecosystem, accelerate the pace of discovery, incorporate individual models into holistic, and help us simulate what-if scenarios so that we can make better decisions to answer the challenge of sustainability.

[1] https://esa.un.org/unpd/wpp/

[2] Alexandratos, N., 1999. World food and agriculture: Outlook for the medium and longer term. Proceedings of the National Academy of Sciences, 96(11), p. 5908–5914.

[3] WCED (1987) Our common future. Oxford University Press, Oxford.

[4] Jonathan Pryshlakivsky and Cory Searcy. 2013. Sustainable Development as a Wicked Problem.

[5] Midgley G, Reynolds M (2004) Systems/operational research and sustainable development:

towards a new agenda. Sustain Dev 12(1):56–64.

[6] Frame B, Brown J (2008) Developing post-normal technologies for sustainability. Ecol Econ

65:225–241

[7] Espinosa A, Harnden R, Walker J (2008) A complexity approach to sustainability—Stafford Beer

revisited. Eur J Oper Res 187:636–651

[8] https://www.dartmouth.edu/~ai50/homepage.html

[9] http://www.ainewsletter.com/newsletters/aix_0501.htm

[10] https://www.theguardian.com/technology/2016/aug/22/google-x-self-driving...

[11] http://www.bbc.com/news/technology-37713629

[12] https://www.theguardian.com/technology/2016/mar/15/googles-alphago-seals...

[13] https://research.googleblog.com/2016/01/alphago-mastering-ancient-game-o...

[14] Domingos, Pedro. The master algorithm: How the quest for the ultimate learning machine will remake our world. Basic Books, 2015.

[15] Haykin, Simon S. Adaptive filter theory. Pearson Education India, 2008.

[16] Triana, E., Labadie, J. W., Gates, T. K. & Anderson, C. W., 2010. Neural network approach to stream-aquifer modeling for improved river basin management. Journal of Hydrology, 391(3-4), pp. 235-247.

Comments

Add new comment

User login

Featured Contributor

Climbers & Bats GCRT

This research team creates a working group of rock climbing interest groups, CSU biologists and human dimension specialists, and CSU students to strategically collect information on bat roost locations and share bat conservation information with the climbing community. View details of their GCRT here and their blog entry here.

Recent Comments

Join the Conversation