Contact person
Aleksis Pirinen
Forskare
Contact AleksisAleksis Pirinen, AI researcher at RISE, will hold one of RISE's acclaimed Learning Machines Seminars on January 27 (2022). His presentation is about using AI to mitigate the effects of climate change, but also potential risks and limitations with this technology in an increasingly climate-destabilized and uncertain era.
A very timely and important topic. We therefore asked Aleksis some questions before his presentation.
Before I started at RISE, I completed my doctoral studies in computer vision and machine learning at Lund University. In my thesis, I investigated methods for improving various tasks in computer vision, such as object detection and semantic segmentation. For example, me and my collaborators developed methods for detecting objects in images more efficiently (this work was partly inspired by how we as humans quickly perceive a scene or an image), as well as approaches that reduce the amount of training data needed for an AI model to become accurate at a given task.
In my research at RISE, I develop AI methods that can be used for climate adaptation measures and adjacent areas to try to mitigate the effects of climate change. An example topic that partly overlaps with my expertise from the doctoral studies is about how AI can be used for intelligent maneuvering of drones in a context of landscape analysis and the like. It may for example be about using drones to keep track of how different crops are doing, so as to be able to take appropriate measures to, for example, reduce the risk of them dying as a result of drought.
AI can play several important roles here. With AI's ability to learn from large amounts of data, there is potential to use it for all sorts of predictive tasks, such as predicting risks of drought and floods (based on, for example, historical and current satellite data), or to predict and thereby streamline different sectors' energy use based on external factors such as weather and season. There is also a growing interest in combining AI with physics-based climate models, for example to improve the understanding of cloud formation and the role of clouds when it comes to weather and climate.
Yes, there are several risks, but let me highlight a few examples. One is that an excessive belief that technology will "solve" the climate crisis may lead to the lack of necessary political decisions - especially decisions which may be unpopular with large sections of the population - because political leaders can justify the lack of action by saying that technology will solve the problems.
Another risk area is the role and impact of AI in an increasingly climate-destabilized and turbulent era. Important discussions are needed here on how to reduce the risks that AI exacerbates the ongoing development. Examples of such issues could be: "How does AI-driven social media (including filter bubbles) affect public discourse on climate change?" and "Can and should anything be done to reduce the use of AI in wars and conflicts (for example, to carry out drone attacks)?", and so on.
I hope that the seminar will initiate thoughts and conversations about the role of AI (both its benefits and risks) in an increasingly turbulent and uncertain time.
As I highlight in the seminar, many are now emphasizing the importance of talking about a potential (and largely climate-driven) global societal collapse - something which is already a reality for large parts of the world's population. So far, I have not seen discussions about the role of AI in the event of such a development on a global level. My view is that most conversations held in public spaces tend to be characterized by an optimistic outlook on the future where the climate crisis will be curbed and civilization stabilized, and that conversations are missing about the scenario where the future is not better than the present but instead radically worse. This is despite the fact that a collapse of our civilization may now very well be the most likely scenario.
I hope that we will be able to change this and that we will dare to have conversations about these more pessimistic scenarios, so that we can try to reduce and delay suffering and harm even during such developments. Note that one does not have to exclude the other - in contrast, these conversations need to be had in parallel with accelerating work to reduce greenhouse gas emissions.
Learning Machines Seminars gathers experts in AI for an open weekly seminar every Thursday. Seminars include presentations on a current topic on machine learning. Meet people and listen to AI experts at RISE as well as invited speakers from academia and industry.
No registration required and the seminars are free of charge.