Can AI Save Humanity From Climate Change? That’s the Wrong Question – Interesting Engineering

Artificial intelligence is among the most poorly understood technologies of the modern era. To many, AI exists as both a tangible but ill-defined reality of the here and now and an unrealized dream of the future, a marvel of human ingenuity, as exciting as it is opaque.

Its this indistinct picture of both what the technology is and what it can do that might engender a look of uncertainty on someones face when asked the question, Can AI solve climate change? Well, we think, it must be able to do something, while entirely unsure of just how algorithms are meant to pull us back from the ecological brink.

Such ambivalence is understandable. The question is loaded, faulty in its assumptions, and more than a little misleading. It is a vital one, however, and the basic premise of utilizing one of the most powerful tools humanity has ever built to address the most existential threat it has ever faced is one that warrants our genuine attention.

Machine learning the subset of AI that allows for machines to learn from data without explicit programming and climate change advocacy and action are relatively new bedfellows. Historically, a lack of collaboration between experts in the climate and computer sciences has resulted in a field of exploration that is still very much in its infancy.

Happily, recent years have seen the beginnings of a shift in that paradigm, with groups like Climate Informatics and the Computational Sustainability Network focusing on how computational techniques can be leveraged to advance sustainability goals.

Taking this notion a step further, a group of young experts in machine learning and public policy founded Climate Change AI in 2019, a non-profit that aims to improve community-building, facilitate research and impactful work, and advance the machine learning-climate change discourse.

There have been different communities working on different aspects of this topic, but no one community unifying the discourse on AI and the many different approaches to climate action, explained Priya Donti, co-founder and power and energy lead of CCAI in an interview with Interesting Engineering.

Climate Change AI has, in no uncertain terms, altered that landscape. In 2019, the group published a paper entitled Tackling Climate Change with Machine Learning, a call-to-arms for the machine learning community that presented 13 areas ranging from electricity systems and transportation to climate prediction and agriculture where the technology might be best utilized. Dozens of experts in the machine learning, climate change, and policy communities contributed sections to the paper and well-known figures like Andrew Ng and Yoshua Bengio provided expert advice on the project as well.

In the years since its publication, the organization has helped foster communication through workshops and other activities, ensuring that the people joining these events are a blend of computer scientists and those from other disciplines.

Encouraging this communication is neither easy nor without its difficulties, however, something that David Rolnick, one of the papers authors and co-founder and biodiversity lead of CCAI readily acknowledges.

The machine learning and AI community is very vulnerable to hubris, explained Rolnick in an interview with Interesting Engineering. Thinking we can solve the problems of other fields without [...] working with people in those fields, without having to leave our algorithmic tower. As in other areas of applied machine learning, meaningful work on climate change requires collaboration.

The interdisciplinary mingling the group promotes is beginning to bear fruit. Many of the professionals who engage in these events help facilitate dialogue between experts of varying fields who would otherwise have a hard time understanding each other, a prerequisite of any collaborative effort.

Were starting to see a lot more people who [...] are not 100 percent machine learning experts, theyre not 100 percent experts in the climate-change-related domain, [but] theyve done a really good job of doing work at the bridge between those two things, and as a result, are able to bring people together, Donti notes enthusiastically.

The team at CCAI believe that researchers and policymakers alike are beginning to alter the focus of their efforts as a direct result of the groups 2019 paper, and its broader efforts. Along with healthcare, climate change is now widely viewed as a key application of AI for the greater good, something that wasnt the case just a few years ago.

I think one thing thats inspiring is the number of people who have risen up to take on [the climate change] challenge, says Donti.

Crucially, though, that inspiration needs to translate to results, and that mentality underpins the teams efforts.

Whether Im optimistic or pessimistic, fundamentally, Im action oriented, and I think its important to do what we can, she underscores.

Ultimately, doing what we can to address climate change through AI (or any other technology) is going to be approached via two basic principles: limiting greenhouse gas emissions going into the future and responding to the effects of what levels of climate change we have, unfortunately, already locked in.

Research bodies, governmental institutions, and private companies around the world are beginning to take up the challenge on both fronts. Brainbox AI, for example, is a Montreal-based company that uses machine learning to optimize HVAC systems in office buildings and other kinds of real estate. This is a key area to focus on when dealing with potential GHG reduction, as the energy consumed by buildings accounts for a quarter of global energy-related emissions alone.

Given that real estate is a major contributor to greenhouse gas emissions, the decision-makers in the industry have a major opportunity to lead the charge, explained Jean-Simon Venne, CTO and co-founder of Brainbox AI in an email exchange with Interesting Engineering.

An AI-driven HVAC system can allow a building to self-operate, proactively, without any human intervention. It can ultimately evaluate the most optimal HVAC configuration for energy efficiency, saving money but also reducing the load on the power grid, keeping the buildings footprint low.

Adaptation will be just as crucial an effort, as extreme weather events driven by rising temperatures rapidly increase in frequency. Disaster response is one area already seeing the application of AI technologies, with machine learning being used to help people recover from natural catastrophes far quicker than in the past.

Such was the case during the 2021 typhoon season in Japan, when the U.K.-based company Tractable used its AI in partnership with a major Japanese insurer to assess external property damage caused by Typhoon Mindulle, helping homeowners recover more quickly. The company claims it can reduce the time needed for damage assessment from several months to a single day.

Just as neither of the goals of climate change mitigation and adaptation will be easy to make progress with, neither can be accomplished using AI alone. While the technology lends itself to flashy news headlines and compelling sci-fi narratives in literature and film, its far from the silver-bullet solution that its often made out to be.

Rolnick stresses that the practicality of what machine learning can and cant accomplish must be a primary consideration when entertaining the idea of applying the technology to any particular problem. Climate change isnt a binary issue, and we must mould our attitudes accordingly.

[AI] is not the most powerful tool, he emphasizes. Its not the best tool. Its one tool, and its a tool that I had at my disposal. Im not optimistic because of AI specifically, Im optimistic because climate change isnt an on-off switch. We get to decide just how bad it is. Any difference that we can make is a meaningful difference that will save lives.

The applications of machine learning are manifold, and both the groups 2019 paper and their recently-published policy report for the Global Partnership on AI are well worth an in-depth read.

The team at CCAI underscores that one basic use of machine learning in this space is its ability to help gather data, like how the technology was recently used to create a map of the worlds solar energy facilities, an inventory that will be of great value going into the future. Such datasets will help scientists better guide their research and policymakers make informed decisions.

Another area where it can make a substantial difference is in improving forecasting, scheduling, and control technologies that pertain to electricity grids.

The energy output of electricity sources like solar panels and wind turbines are variable, meaning they fluctuate depending on external factors like how much the sun is or isnt shining on any particular day.

To ensure consistent power output independently of weather conditions, back-ups like natural gas plants run in a constant CO2-emitting state, ready to fill in those gaps. Improving energy-storing tech like batteries could be a way to reduce the need for such high-emission practices, with machine learning being able to greatly accelerate the process of materials development and discovery.

Were seeing huge advancements in batteries in terms of cost and energy density, Donti says. Batteries are going to be a critical piece of the puzzle, and there are some companies using AI to speed up the discovery of next-generation batteries. One example is Aionics.

Aionics is a U.S.-based startup using machine learning to expedite battery design, which could, in addition to improving electricity systems, unclog one of the bottlenecks standing in the way of electric vehicle adoption on a large scale.

Using machine learning to help decarbonize the transportation sector on a larger scale is more difficult, however. Passenger and freight transport are notoriously difficult to decarbonize. If fossil fuels are to be replaced with batteries, for example, they will in many cases need to be extremely energy-dense. But thats only a tiny part of the picture, the bigger issue being the convoluted nature of the transportation sector itself.

In the electricity sector, you have relatively few, large players, and its rather centralized. What happens in terms of innovations is happening in fewer companies with more aggregate datasets, explained Lynn Kaack, assistant professor of computer science and public policy at the Hertie School in Berlin and co-founder and public sector lead at CCAI in an interview with Interesting Engineering.

In transportation, there are many more and smaller companies [...] often there is much less means, much less data to exploit. Where one can take the system perspective, trying to optimize routing, charging station placement, machine learning has interesting things to add, but its not always straightforward.

Kaack points to the example of how German passenger rail operator Deutsche Bahn is looking at maintenance optimization through machine learning. Technological failures result in delays, and delays have a big influence on whether or not passengers perceive rail as a viable alternative to driving.

Technical challenges are far from the only thing that needs to be overcome in the service of doing right by the planet. How these issues and their potential solutions are framed and perceived matters greatly.

The public sphere is prone to putting a spotlight on glitzy techno-cures that can divert attention away from simpler but potentially more actionable projects and technologies. Neither are research bodies or governmental agencies immune to such frenzy. Awareness here is crucial, as the lens through which AI is seen can play a role in dictating the direction research leans and where funding ends up.

AI can make certain kinds of action easier, but it can also lead to greenwashing, Rolnick warns. Techno-solutionism can lead people to think they are having a much bigger impact than they are, and even divert peoples attention away from lower-tech, but more impactful courses of action.

Working on unsexy problems is important. How even the most exciting technologies get integrated into the workflow where they will be applied is quite simply boring, essential work. Persuading the relevant parties involved in funding and finding a new solution often requires the right rhetorical touch.

For different innovations and solutions, we should think about who the audiences are who need to be convinced, who are the people who might be financing things, how do you make [the incentives] clear to private and governmental funding sources, Donti says.

By the looks of things, many appear to find the group and its goals compelling. Climate Change AI has had a direct impact on funding for programs like the U.S. governments DIFFERENTIATE program and Swedens AI in the service of the climate program, for example, and theyve just finished the first round of an innovation grants program thats allocating two million dollars to projects that will promote new work by creating publicly available datasets.

On a broader scale, how we leverage and manage AI is a topic that is increasingly being given the attention it deserves. Last April, the European Commission introduced the Artificial Intelligence Act, the first large-scale regulatory framework for the European Union regarding technology.

While some claim the framework doesnt do enough to protect civil rights and liberties, it is a step in the right direction, and the more central and common these high-profile discussions become, the better. Anyone and everyone involved in machine learning applications need to embed the ethical considerations of relevant stakeholders, not just investors, into the foundations of the technology as much as possible.

Taking all of this together, its not a stretch to say that AI can be utilized to address climate change. But the fact remains that the issue is an extraordinarily complex one, and even those directly involved in approaching it admit that the conversation of when and how we do that is an ever-evolving one, wherein the most effective path forward is never exactly clear.

Are you going to spend your time with practical applications and policymaking, helping people who are supposed to make decisions shape funding programs and inform legislation, or do you go back to fundamental research? Its difficult to balance them and understand which has the greatest impact, Kaack says.

While a difficult question to navigate, that its even being asked is nothing short of inspiring. Doing what is within ones reach stands out as an evergreen principle for achieving real, tangible action, even when dealing with something like climate change. The overall message is less of a, Do it with AI, and simply more of a, Do, period. In the face of a problem of this scale, one that often feels paralyzing in its insurmountability, that message is a refreshingly galvanizing one to hear.

Im not here to say that AI should be our priority, reiterates Rolnick. AI is a powerful tool, but climate action will require all the tools. The moral of the story for me is that it is important for people to think about how they can use the tools they have to make a difference on problems that they care about.

See the rest here:

Can AI Save Humanity From Climate Change? That's the Wrong Question - Interesting Engineering