AI is having a moment. The media is full of it, tech bros have made squillions out of it, and politicians are falling over themselves to pretend they understand it – and know what to do about it. Scary stories abound, but there’s a brighter side to the AI coin, too. It holds some practical promise of tackling a number of stubborn challenges.
Here are four ways it’s doing just that.
1) Futuring the forests
The rainforests of South America may be under threat, but they’re still vast and remote enough to make monitoring that threat a real challenge. Satellite surveillance is one answer, but it has its limitations, and foot patrols can only cover so much ground. Cue drones. Equipped with sophisticated cameras and motion sensors, they can soar almost silently through the canopy, spotting the slightest sign of illegal logging or other disturbances. AI is able to decode and analyze their findings within seconds, giving real-time reports of emerging threats – such as signs of a new logging road – to allow prompt action to nip them in the bud.
AI can be at its most effective when it helps match high-tech observation with the eyes and ears of people on the ground. In the Peruvian Amazon, Indigenous activists are using a customized version of MapBuilder software to build up a detailed picture of the state of their forest lands, combining satellite data with observations from their own patrols. As campaign group Global Forest Watch puts it, this “provides them with undeniable evidence of the challenges they face, transforming abstract threats into tangible realities that can be addressed and acted upon”. Or in the words of local Indigenous leader Ranin Koshi: “Technology allows us to reaffirm and guarantee our rights for our future generations.”
2) Where the Wild Things AI
Some of the world’s most precious and endangered species are, by definition, the most elusive – as are some of the most insidious threats facing them. AI can help track down both. By analyzing photos from camera traps, it’s helped Indian conservationists build up an unprecedented database of individual tigers – so that if a pelt appears on the international market, it’s possible to check whether it’s from a legitimate (if unedifying) tiger farm, or has been illegally poached.
Faced with the challenge of comparing and analyzing a vast amount of wildlife data from numerous sources – camera traps, individual observations and the work of citizen scientists – AI comes into its own. It’s helping identify and track individual creatures in remote environments, from the snow leopards of the high Himalayas to whales and dolphins far out at sea, giving vital information for conservation strategies.
In Uganda and South Africa, meanwhile, wildlife rangers are using a combination of drone cameras and AI tools developed by researchers at the University of Southern California to anticipate the movements of poachers intent on killing rhinos or elephants, allowing the rangers to move in swiftly before the fatal shots are fired.
3) An AI-grarian revolution
One of the strengths of AI is its ability to scan a vast amount of data from a hugely disparate array of sources, compare and analyze it, and present a customized result to the screen of a basic smartphone in seconds – or quicker. And one of the benefits of such an ability is to provide low-income farmers with the sort of detailed expert advice that would previously have been far out of their reach, or their pocket.
Take the Khammam district of Telangana state in southern India. Here, 7,000 chili farmers took part in a pilot run jointly by the World Economic Forum India’s Centre for the Fourth Industrial Revolution and the Indian and state governments. It used AI to help with everything from soil analysis at field level, to nutrient management and pest prediction, all the way to quality control and marketing, and developed a chatbot in the local Telugu to give farmers direct access to the findings in a language they could understand.
At the end, the average farmer taking part had increased chili production per acre by a fifth, cut pesticide and fertilizer use, improved quality and marketability – and saw a doubling in their net income. Unsurprisingly, state authorities are now adopting the program, expanding it to 500,000 farmers growing five different crops.
4) Soundscapes at sea
Heard of ecoacoustics? It sounds like an option on the decks of a recording studio, but in fact it’s one of the latest, and most promising, ways of monitoring the health of coral reefs. And it does so by listening to them. Super-sensitive underwater mics – known as hydrophones – can pick up the slightest sound. Now an AI system called SurfPerch, developed by scientists at the University of Bristol and University College London, in association with Google and its AI division, DeepMind, is able to decode the sound data to identify and trace the movement of fish and other marine species, and so build up a picture of the ecological health of the reef and its surroundings.
Its developers have helped ‘train’ the model by drawing on libraries of birdsong, after discovering that, despite sounding completely different, there are sufficient common patterns with fish sounds for the model to learn from one to improve results for the other. It’s a classic example of how AI can in effect teach itself, at speed, to boost its performance. And it holds out hope for a fragile ecosystem that covers just 0.1% of the ocean’s surface, yet is home to 25% of its species.
Remember Dr. Dolittle and his passionate wish to “talk to the animals”? AI now holds out the dizzying if distant prospect of being able to do just that, by analyzing everything from a bat’s squeak to a whale’s song. In Project CETI (Cetacean Translation Initiative), researchers are using AI’s pattern recognition capabilities to analyze a vast range of recordings from sources such as whale-mounted tags and underwater robots, and so start to map out how individual creatures are communicating with each other, under all sorts of different conditions.
It’s a ways off from being able to interpret whale song – let alone sing back to them – and when that time arrives, there will be plenty of ethical questions to navigate. But the potential benefits in terms of conservation insights, not to mention intra-species respect and understanding, could be off the scale …
The energy-hungry elephant in the room…
No serious consideration of AI should duck the question of its thumping levels of energy consumption. Run a query through ChatGPT, and it uses 10 times as much electricity as a Google search. Some estimates suggest this could mean doubling carbon emissions from data centers by 2030, and that AI as a whole could lead to a surge in electricity consumption, just at a time when we desperately need to curb it.
It’s a serious issue, but again, there’s another side to the story. Some of the biggest players in AI – Google, Microsoft, and others – are investing heavily in solar and other renewables to power their work. But the most promising aspect lies in the potential of AI itself to cut consumption. When Google used the AI expertise of its newly acquired DeepMind to analyze the electricity used for cooling its data centers, they discovered they could cut it by a staggering 40% – and this was after they thought they’d already made them as efficient as possible. At a stroke, that knocked 15% off the whole group’s data center consumption.
AI is still in its infancy, but already experts are predicting that similar levels of savings are possible thanks to its ability to analyze energy demand and consumption at the most granular level, and optimize it – whether it’s a building adjusting minutely to changes in the weather or occupancy patterns, or a fridge providing just the right amount of cooling for what’s inside it.
That elephantine obstacle to AI’s adoption may not be such a chunky beast after all …
Originally published byMartin Wright, Positive.News
| < Previous Article | Next Article > |
|---|