Yihyun Lim: “AI may help us visualise and combat Climate change”

Yihyun Lim: “AI may help us visualise and combat Climate change”

Can AI help us think the future differently and change its trajectory, especially regarding Climate change? That’s the view of Yihyun Lim, Assistant Professor at the University of Southern California and Former Director of the MIT Design Lab, and the topic addressed during the seminar organised by SKEMA Center for Artificial Intelligence (SCAI).

How do we generally imagine the future? When we think about the future, various narratives seem to dominate, like the idea of projected future, technocratic future and collapsed future. Which one is the most dominant stream?

Yihyun Lim: Our visions of the future are influenced by those that are published in the media; whether it’s in movies, or maybe it’s in the forecasting reports that we see at the end of the year, or maybe it’s in big tech festivals such as the Consumer Electronics Show (CES). It is these technocratic futures that we are exposed to and that we are most familiarized with. And then oftentimes there are visualizations portrayed by the movie industry, showing the extreme climate futures, which we call collapse futures. Current future are driven by the industry, government, and those with the power to create and disseminate.

These dominant visions of futures influence how we, as individuals, create visions of futures. How can we, as individuals or communities, envision our futures in our own terms, and how can we make these voices to be heard?

How can AI help us think differently? If it is powered by humans, can AI help humans imagine a different future from the ones they generally imagine?

When we talk about future visions, we all form different images in our head. This is because it’s all text-based. Having images portray words, having an audio-visual format can help us to communicate our ideas. Through images we can see what each other is thinking about. Generative AI can have a really important role in helping us to create visuals of different text-based narratives. However, the current limitation of generative AI tools is that we can’t fully control how outputs are being generated. There is no way to fine tune, edit, and preview the generated output to match what I am envisioning. Even when you enter the same exact text-prompt, you’ll get different visual outputs. A lot of things happen in a black box, and we don’t yet have a mental model of how this system works.

At the same time, it represents an opportunity: I may have an idea that I can’t fully visualize. In that case, AI can help with the creative envisioning process. There are both limitations and opportunities.

Read also: Nisreen Ameen: “It’s not easy for humans to give up a task that they’ve been doing for many years and automate it to AI”

Climate scepticism is sometimes fuelled by the fact that climate change is not always tangible. Can AI help to represent it so that we can better understand, accept and fight it?

Climate science education is being communicated through reports, texts, or numbers. These formats aren’t very accessible or digestible for us. Whether it’s a climate action plan being discussed at conferences, or companies promoting their sustainability action plan, it all feels very detached from our personal surroundings and context. It is difficult to make these climate issues tangible, so that I can feel, not just understand, how it may have an impact in my backyard, in my front yard, in my neighborhood.

AI could help us with contextualizing and personalizing these stories, so that I can see the possible direct impact of climate issues such as smog, extreme heat or flood in my neighborhood. Generative AI can visualize these ‘what if’ scenarios and bring it home, to my doorstep. For example, past research project looks into creating a visualization platform where people can type in their address to see the impact of such climate issues happening at their address. It takes a Google street view image of my address, and creates an AI generated image of fire, smog, flood and other possible climate impact. Seeing is how we build empathy. Contextualized visualization can really enhance the public awareness of climate issues and help us to take action, one step at a time.

“Having generative AI to visualize possible actions in hyper-realistic forms can help us to empathize with those futures and take action.”

Yihyun lim

As an expert, deeply involved in the intersection of technology and design, you believe AI can motivate more effective climate action. But is it necessary for us to imagine a specific future, to be motivated, like setting a general common goal for society in front of us and goal going for it together? Is it feasible and possible?

I hope so. Generative AI can also help us to explore strategies. It can help us to synthesize various information and discussions taking place in the community, and brainstorm possible actionable strategies. Generative AI should be utilized as a collaborative partner to help us envision and strategize. But of course we can’t rely on AI to solve the issue. We are the one that need to take action. We are the ones who can take the information and see how it may actually apply. But I think there’s an interesting possibility to bring AI as a collaborator.

Can AI help to play it on climate issues? And if so, is it desirable? For instance, do you see any risks of AI that underplay the scenario? By over simplifying data or presenting overly optimistic scenarios… Is there a danger that AI could be used in ways that might dilute urgent calls for action?

Climate is a complex issue. And we cannot fully capture the complexity of climate futures in one scenario. As mentioned before, we should be cognizant of AI’s capabilities and how we can collaborate with AI to envision various pathways and strategize actionable steps. Having generative AI to visualize possible actions in hyper-realistic forms can help us to empathize with those futures and take action.

Some questions we should ask is how we can make technology to be more usable, accessible, and transparent. Current interface design of generative AI platforms are not fully scripted; it takes some training and exercise to know how to craft a prompt that will bring back useful results. It’s also quite difficult to fine tune results. And most importantly, there’s the question of authenticity. If we are using generative AI tools to create a short video based on personal stories, how can we tell what part of the story is authentic and what is fabricated? How can AI become a trustworthy partner to help us envision and strategize towards positive climate futures?

Margherita PaganiDirector SKEMA Center for Artificial Intelligence

All author's posts

Close Menu