Just a few days before Earth Day, Google led a seminar and Q&A at its building near London’s King’s Cross, hosted by Camden Clean Air.
The session was part of EarthFest 2024, a sustainability festival that brought together thought leaders and industry experts through interactive demonstrations, talks, and workshops.
As environmental sustainability becomes an increasingly pressing concern for industries worldwide, the Google session showcased how AI can play a crucial role in transforming how firms approach the journey towards achieving Net Zero.
Sectors like Real Estate and Conservation, typically bereft of digital innovation, are joining the technological revolution with AI.
Yet, as these industries navigate their digital transformation, they face challenges. Crafting sector-specific AI models, harmonising AI’s capabilities with sustainability objectives, and overcoming data scarcity are some hurdles to a greener future.
Anna Williams, geo-sustainability lead at Google EMEA, kickstarted the session with an optimistic statement: “A report by Google and the Boston Consulting Group shows that AI has the potential to help mitigate 5-10% of global greenhouse gas emissions by 2030.”
According to Williams, that’s the equivalent of the total annual emissions of the European Union.
The real estate sector, responsible for 40% of those emissions, will likely face fundamental changes as the global economy decarbonises to meet climate goals.
Ranjeet Bhalerao, CEO and co-founder of MapMortar, a virtual modelling and simulation platform for decarbonisation retrofit planning, highlighted the scale of the problem.
“To meet those climate targets of 1.5°C, we need to decarbonise one New York City worth of buildings every single week for the next 30 years,” he says, which would cost $1.3 trillion.
According to Bhalerao, one of the primary challenges in Real Estate decarbonisation is the lack of comprehensive, structured data on existing buildings — which is crucial for accurate analysis.
“You’ll find data in PDFs and Excel sheets, and even old files sitting in the basement, which nobody ever looks at,” he said.
AI will enable users to structure data in a way that makes it usable, viewable, and understandable.
It can quickly extract valuable insights from various data sources, which would take humans a lot more time to complete.
“With a simple click of a button, AI can analyse complex graphs and charts to provide information such as your building’s EPC, areas where you might not comply with regulations, energy usage, and associated costs,” said Bhalerao.
Another issue the industry faces is that many Real Estate owners can’t easily find the energy performance data of their buildings, “It’s a huge problem in the industry,” according to Bhalerao.
However, MapMortar uses image recognition, powered by AI, to create building profiles based on predictive data. “We’re training our models to look at images and start giving information — like, that building is six stories tall, has so many windows, has this type of facade, built in this year — and therefore potentially has these characteristics.”
“When you have loads of complex parameters that affect a building, AI could start predicting how the building is going to perform or is performing today,” he continues, “We can use surrogates to predict the end performance and the carbon performance of the building.”
From urban landscapes to the wild expanses of nature — AI is covering all bases.
The same technologies that can parse complex data to optimise building energy efficiency are also unlocking new frontiers in conservation.
In conservation, AI is proving instrumental in monitoring and protecting biodiversity. Robin Freeman, head of indicators and assessments at the Institute of Zoology within the Zoological Society of London, explained some of the ways that AI is transforming processes.
One of the biggest challenges in conservation is the sheer amount of data and the extensive manual labour required to process it.
Comprehensive analysis of biodiversity data from camera traps, audio recorders, and other sources “takes a long time to process manually,” says Freeman.
“A single person labelling those images might take two and a half months,” he continues. One example he gave was Mega Detector by Microsoft, which automates species detection in imagery and audio and reduces the time required to process data.
“Just using that to detect whether an animal is in the image, let alone what that species is, reduces the time it takes to process that data tenfold.” He continued, “We’re now deploying this at scale globally.”
The uncertainty in predictions of future biodiversity loss, which limits the ability to determine the best strategies to mitigate the loss, also poses a challenge.
“We looked at things like logistic regressions and convolutional neural networks to try to find papers relevant to our data. We were so excited last year when we could ask AI to find us the papers,” said Freeman.
He added, ” I think there’s an opportunity for us to use foundational models to look for text relevant to biodiversity change and build datasets that allow us to understand how biodiversity has changed.”
Freeman concluded that the interacting and complex drivers of biodiversity decline, such as climate change, habitat loss, and land use change, must all be addressed together to potentially see recovery.
“Only in cases where we did all of those things together do we see biodiversity begin to recover above the baseline. There’s a lot of uncertainty there. But the idea that we will only begin to see biodiversity recover when we do everything we can is quite fundamental.”
The advent of foundational models ushers in a new era, according to Drew Purves, sustainability and biodiversity co-lead at Google DeepMind. He notes, “The net effect is that as a downstream user, you can do more than ever before with less data, less compute than ever before, and lower skills barriers.”
“If someone read one book on natural history, you wouldn’t necessarily think they’re an expert. But if they read and remembered 10 million books about natural history, you probably would.”
This democratisation of AI technology enables a broader range of stakeholders to engage in sustainability efforts, making complex environmental solutions more accessible and feasible.
Purves points to several AI democratisation applications from Google DeepMind that underscore the scale of its impact.
“For the first time in history, we can now take a DNA sequence, turn that into a sequence of amino acids, and then work out the protein’s shape. This has dramatic implications for environmental sustainability,” he said.
Another notable mention was Google DeepMind’s breakthrough in weather forecasting, which achieves unprecedented accuracy with minimal computational resources.
“Not only are those forecasts more accurate than all other previous forecasts,” he claimed, “but you can make these predictions on a laptop.”
As Purves puts it, the AI revolution is not just reshaping our tools and techniques but redefining the boundaries of what’s possible in our quest for a sustainable future.
These advancements’ implications extend far beyond scientific research and energy production; they represent a paradigm shift in how we approach environmental challenges. “And that sounds a bit too good to be true anyway, but that’s what technological revolutions do.”
Subscribe to our Editor's weekly newsletter
Share
