Imagine you are scanning satellite data for signs of illegal loggers hacking down trees. The volume of data available is colossal, the equivalent of tens of thousands of HD movies by my very rough estimate. By the time you have scoured all of that, if it were even possible, the criminals are long gone.
What if you could use all that data to predict where the deforestation was most likely to happen next? And you had data from a network of sound sensors providing alerts for the sound of chainsaws and another monitoring social media for mentions of deforestation activity? This is all possible thanks to AI, in particular the machine-learning variety of the technology. To learn more, I contacted three experts.
“AI is particularly good at detecting patterns among very large amounts of data that no team of humans could process,” said Dr Alina Patelli, at Aston University in Birmingham. “It learns to identify correlations that are sometimes subtle, even counterintuitive, and can easily escape even the trained eye of highly qualified human experts.”
Leveraging vast amounts of weather data is a good example, she said: “Running AI to correlate predicted droughts with the water demands of industrial farming will yield insights into the likely fluctuations in global food supply over the coming decades, giving scientists a head start in coming up with solutions.”
AI has already delivered weather forecast tools that are faster and cheaper than current supercomputer models – some can even run on a laptop. “That’s a huge development, because it has implications for disaster risk management, agriculture and energy infrastructure,” said Dr Mohammad Hossein Amirhosseini, at the University of East London. In other words, it can improve life-saving extreme weather warnings, help maximise crop yields and make electricity grids more efficient and resilient (pdf).
Running a grid is a delicate balancing act of matching supply to demand, and AI can anticipate and optimise changes, which is particularly useful as inherently variable renewable energy ramps up. The same applies in your home: balancing your heating, cooling and charging needs with weather forecasts and price data.
Dr Andrew Rogoyski, at the University of Surrey, said: “We don’t want to bet our future on the promise of future technologies, but AI does seem to offer the possibility of making big advances in key sciences and technologies that will help mitigate [against] the impact of climate change.”
“For example, there have been recent advances in nuclear fusion, where AI is being used to better control complex instabilities in the fusion plasmas, bringing the promise of very low-cost energy that bit closer,” he said.
There are lots of other examples: Wildbook uses computer vision to identify individual animals from jaguars to dolphins by their unique markings and is revolutionising wildlife tracking and anti-poaching efforts, said Amirhosseini, with other tech even predicting the likely location of poachers.
AI is powering cheap air pollution monitoring devices so smoggy hotspots around the world can be tackled. A food bin camera system backed by AI monitors waste and tells kitchen managers what to buy less of. AI will also help England’s National Parks make detailed maps (pdf) of their fragmented ecosystems at affordable cost to enable the best use of conservation funds.
Concerns over the rapid rise in the energy needs of data centres due to AI are valid, said Amirhosseini. “But we need to look at the broader picture. If AI is used to [cut carbon emissions], then the net environmental benefit can far outweigh the energy cost of training the models.” Furthermore, the increased demand from electric vehicles, air conditioning, heat pumps, and industry will far surpass that from data centres. And work on making the training of AI models more efficient may make the energy concern “a temporary blip”, said Rogoyski.
There is also the serious issue of biases in AI models. “They are as biased as the data used to train the underpinning software,” said Patelli. Rich western countries tend to have far more data than poorer, developing countries, meaning people in the latter could be ill-served if AI tools are used without careful thought.
This points to a deeper cultural issue, said Amirhosseini, which is “the risk of overrelying on automated systems and sidelining local knowledge and lived experience. We need to treat AI as a partner, not a substitute, for human judgment, community engagement, and scientific integrity.”
If we do that, AI can continue to make ever-larger differences in the real world to how we understand and protect our planet.
Read more:
