A clever start-up is using AI combined with old smartphones to help combat illegal logging in rainforests, which is contributing to deforestation and carbon emissions.
So much of the focus about AI is on automation and what socioeconomic impact it will have, but all of this pales into insignificance when you consider the broader ramifications of climate change.
Therefore, one day in Amsterdam late last year at a Google Making AI event, it was an enormous thrill to see AI being combined with old mobile phones to fight deforestation.
‘We are currently covering around 2,500 sq km of rainforest in 10 countries. We intend to scale this to another 4,000 sq km in 20 countries’
– TOPHER WHITE
According to Topher White, co-founder of Rainforest Connection, a San Francisco-based start-up, deforestation caused by illegal logging is the second largest contributor to carbon emissions.
When White was visiting a gibbon reserve in Indonesia, off in the distance he heard the distinctive sound of chainsaws. Roads were being cut through the rainforest to make way for illegal loggers to transport lumber.
White’s instincts as an engineer kicked in and he set about devising a system whereby the microphones on old smartphones could connect to solar panels and, using AI to distinguish chainsaw noise from bird and animal noises, they could alert the authorities to illegal logging.
“One-fifth of carbon emissions comes from illegal logging. And I felt if we want to offer support for the tribes, one of the best ways is to use technology.
“The moment a chainsaw goes off, our mic in the trees picks it up and we can then alert local rangers to stop people in the act.
“We basically take an old cellphone and put it in a box and, using the powerful microphone in most smartphones, listen to the sounds of the forest and detect chainsaws and alert local people.”
Up until now, satellites have been used to monitor for signs of deforestation but this doesn’t happen fast enough to halt the destruction. “It is usually after the crime has been committed and there are more dangerous people on the ground. This real-time response has been helpful to people in the field.”
AI against climate change
White said that the system used by Rainforest Connection was built using TensorFlow, Google’s powerful open source computing resource. “We can stream all of the audio in real time into the cloud, and even 20km from the nearest cellphone tower we can use TensorFlow to pick out the sounds of chainsaws.
“This enables us to process massive amounts of data in real time – all because it’s in the cloud.”
White built the system initially in his parents’ garage in Silicon Valley.
Today, according to Rainforest Connection’s website, the system monitors 26,000 hectares of forest, has gathered 4,629 days worth of data and has helped to sequester more than 6.5m metric tonnes of CO2, equal to taking 1.3m cars off the road.
“We spent the first few years building tech partnerships and getting to know the tribes, the governments and the NGOs. One of our first alerts was responsible for allowing two dozen indigenous rangers to get to the scene, seize trucks and send loggers away. These people [the rangers] have more impact on climate change than engineers. They don’t need to know all about the technology – just that it works and the faster they get there, the safer it will be for all concerned.
“We are currently covering around 2,500 sq km of rainforest in 10 countries. We intend to scale this to another 4,000 sq km in 20 countries, but we can make use of the AI for so much more. In five years’ time, we hope to use old phones to protect 25,000km of forest.”
White estimated that one device in a tree can help protect three sq km of rainforest. “That equals taking 3,000 cars off the road and 15,000 tons of CO2.
“Climate change is something we just cannot wait around to solve. The challenge we have is finding data scientists and resources. There are people that love solving these problems and we want to be available to them, but finding them is hard.”
In terms of AI and data science, there is a lot that can be done with Rainforest Connection’s platform.
“Imagine if we had hundreds of years of biology in audio recordings and very soon, with the help of data scientists, pick up all the species in a recording and make it available to people to do biological studies. Imagine if you could make ecology into a big-data problem – like how the sounds of birds change if a jaguar walks through.”
Rainforest Connection has already created a streaming app for people to listen to the sounds of rainforests. However, the project has not been without its share of problems. “A rainforest can be an oppressive place for electronics.”
He continued: “Our technology works hard 24/7 and some of the boxes have lasted more than two years and the solar panels keep them going.” Despite the struggle, White said old phones are proving durable enough to survive even cold winters in Siberia. He added: “Funnily enough, we have had no instances of orangutans tearing the boxes apart but insects will find a way to eat into anything.”
While the big-data problems are tantalising, from a machine-learning sense, making the network scalable and useful for rangers to react is a bonus in itself.
“We have done a lot of work in South Africa and Cameroon and, as more and more old phones become available, the costs are getting cheaper and cheaper. We want to expand this across the whole world and the more people who can help us on the data science and software side, the more scalable it becomes.”
Originally, White and his team were using Arduino boards before they cottoned on to repurposing old mobile phones that still contain a wealth of technology. “We came to a point where it was obvious that the future of data analytics was going to be based on machine learning and we had to get on that bandwagon.
“There are other people working on other effective ways to save the rainforest and we are just one them,” White concluded.