The research team, led by Victor Galaz of the Stockholm Resilience Centre, looked for investments and uses of these technologies that could possibly impact the environment. Following the money leads to a global picture and positive ways to make sure these tools are used for sustainable outcomes.
Why is this topic important?A
We are in the so-called post-industrial revolution. AI, machine learning, sensors, all of these technologies are advancing in all possible fields. We have drones in the air, sensors in the land, other machines such as robots working underwater and all of them simultaneously have an increased impact on human societies, as well as on the environment.
With these rapid advances, people think that monitoring will be able to increase society’s capacity to respond to change on the planet. In particular, AI could provide us with tools that can help us steward ecosystems and natural capital in more active ways.
Increasing the use of AI enhances global connectivity and could either individually or in collective succession affect many aspects of the environment and sustainability because it affects behaviour.
We have sectors such as trade, finance, economy, energy, early warning systems, farming, forestry: all increasingly use AI. We have a situation where we humans think we have control of each sector so we have better insight to react quickly in the sectors.
The way any technology is constructed and designed will influence its use, which will affect the environment. Studying these technologies will help us understand them better, how they operate, how they act, the opportunities, the risks of these technologies and how can we utilize them to the best of our knowledge.
What is the most critical technology for these sustainable and environmental impacts?A
I would say there are a couple of technologies: drones, robots, software, satellites.
Different types of sensors or small devices in applications then operate independently or on phones and computers, so they are many technologies. And from day to day, they are growing and increasing – in different applications, in various kinds and depending on the sort of problem or this sort of application, you have the technology that is more or less relevant, but all of them are important in their own way.
Depending on the application in different areas, a rating of which technology is most important, I would say that all of them are.
What is the global picture for number of companies, finance and type of products used in this space?A
In our paper, we decided to study the agriculture, forestry and marine industries, fields where the application of AI is more easily grasped concerning impacts on the environment.
By looking at this in general, we realized that we needed a global picture. We chose agriculture, marine aquaculture and forestry to see what is happening in developed countries and emerging markets – anywhere from North America to China, Brazil and India. We realized the largest number of companies working in this space are from the US, followed by Brazil, Canada, India and Israel, which, surprisingly for a small country, was number five.
Then we looked at the funding the companies acquired to support technology development and enter new markets to spread their technologies to be used around the world. The US has the largest number of companies, but China is by far the place where the largest amount of dollars are invested. The US had invested $1.5 billion and China almost $5.5 billion, so those are the two biggest markets when it comes to economic power.
According to our research, when it comes to size, the US is first and China is number seven. We believe we captured most of the companies. However, since this is a very entrepreneurial and rapidly growing thing, this could change from month to month.
What are the systemic risks?A
Systemic risk starts from economy and finance to even the biosphere: what can cause a collapse of an entire system? For example, it could be an individual risk that causes the collapse of the whole economy. Systemic risk is crucial because it’s capable of creating catastrophic impact that leads to instability and a domino effect of chaos, if I may say so in my own words.
We found four different types of risk in the research. First, one is algorithmic bias and allocating harms risk. I’ll give you an example: software used to model certain production features on a farm, such as in Sweden. So it’s developed somewhere with industrial farms and in a data-rich context and it has a decision-support system. However, if you take the same software somewhere else, in a developing country with a different social, economic and maybe climatic context, algorithms on which the software was based can lead to damaging results.
The second risk is equal access to benefits. We have this digital divide, meaning people in developed countries with bigger farms may be more developed in certain technological aspects than smaller farmers in developing countries. So you know if you have a perfect solution with these drones, software, even satellites for farming, maybe they can function in countries like Sweden, Norway, Switzerland, Canada, Australia, Japan. However, if you would like to put some of these technologies to use in places such as Honduras, Afghanistan or Lesotho, people there would not have financial means, they will not have the knowledge, in some cases, they would not have the technical support. Some of these systems require special optic cables and 5G networks to operate. For example, in certain countries, you have only access to 3G networks.
Risk number three, we call it cascading failures and external distractions. For example, data is stored online and these technologies are heavily Internet-dependent. This means cyber attacks, for example, can disable or block your accounts or cyber attackers can take control of your machines and do things with your machines that are not supposed to happen. A technology intentionally designed to be used for good can become a weapon. Say a hacker takes a drone full of pesticides and dumps it on a local school close to a farm. It’s probably not deadly, but it’s definitely not healthy and not a wanted experience.
Trade-offs between efficiency and resilience are the fourth thing. Say you want to increase the productivity and efficiency of a particular crop, maybe soy grown on a massive plantation, supported by advanced AI technologies. That will come at the cost of a decline in biodiversity, scenic beauty and climate or flood regulation.
How can we ensure that future products are sustainable?A
Digitalization is happening and it has many benefits. We should not ignore that whole trend. However, at the same time, societies need to be careful and try to utilize the greatest maximum benefits and minimize the risks.
We concluded that the current interest in responsible AI principles could do much more to integrate environmental stability dimensions. Right now, these principles are focused on social aspects. In the future, they need to evolve to include environmental resilience. For companies developing their technologies, most of the people there are engineers or IT people; not all are environmental scientists or have a good understanding of the environment and environmental processes.
However, people working in the sector can use principles of the human response of using AI as guidelines or a roadmap to base development of the technologies. These principles deserve an ongoing dialogue between the IT community and environmental scientists – they need to talk to each other in the same way that is happening now in AI finance and the environment. A financial committee committee is trying to talk to environmental scientists to understand their work and vice versa, so I think the same trend needs to be present in AI.