Ai for the PLANET

IA et nature

When people talk about ‘artificial intelligence’, what springs to mind is usually humanoid robots as depicted in science fiction, or perhaps conversational assistants such as Siri and Alexa, that are able to launch a music playlist, give you the weather forecast or read out your diary entries to you. Or some people might immediately think of the AlphaGo software package from Google’s DeepMind, which soundly defeated the world’s best Go players in some highly media-hyped matches. However, we tend to think less about how we can use AI technology to protect the environment. Nevertheless, there is huge potential here, as the recent Microsoft initiatives in this field have shown. Last July, the computer giant launched a new programme called AI for Earth, whose purpose is to put Artificial Intelligence to work to help protect the planet. This means, for example, making it easier for researchers and organisations working to preserve our environment to use the technology. Four areas have been targeted: water, agriculture, biodiversity and climate change.


PROTEcting BIODIVERSITy with ALGORITHMS?

Mountain moon

PicsArt

AI in the role of geographer

"Our objective is to train computers so that the task no longer takes years, but just a few minutes.”

Microsoft quickly put its words into action by starting to work on three concrete projects. The Chesapeake Conservancy Project, which works to promote what is known as ‘precision conservation’ in Chesapeake Bay on the east coast of the United States, is the first of these. It uses AI for geographical mapping of the area. “If you want to protect an ecosystem, you need to know what it contains. Trees, rivers, plains, ditches; everything has to be listed. Today, this is done laboriously by hand. Our objective is to train computers so that the task no longer takes years, but just a few minutes,” explained Robert Bernard, Chief Environmental Strategist at Microsoft, when L’Atelier caught up with him at the Web Summit in Lisbon last November. 

With the recent advances in deep learning – a branch of Artificial Intelligence which enables an algorithm to train itself up on the basis of a large volume of data – computers have become capable of deciphering the contents of a given image, identifying the image’s various components as well as, or often even better than, a human being. One example is Facebook’s algorithms which automatically identify faces shown in a photo. To quote Jeff Dean, who is Leader on the Google Brain Project: “computers have opened their eyes”.

CHESAPEAKE bay

bay

Once you can identify a person, it is not that much more difficult to recognise an individual tree or a water course. Microsoft has teamed up with California mapping and spatial data analytics technology company Esri, and the Chesapeake Conservancy, an NGO, to apply the latest advances in image analysis to help protect Chesapeake Bay. By feeding an algorithm with the bank of images available for the area, they have managed to draw a map of the area that is more precise and up-to-date than previous versions, and far more quickly than if it had been done by hand. This new classification provides 900 times the amount of information as conventional 30-meter resolution land cover data; it is now at one-metre resolution land cover. Using the latest technology allows the NGO to spend less time on sorting out technical problems and more on actually protecting the environment. 

Ai for listing fauna and flora

IA nature
Shutterstock

Cataloguing species and preventing epidemics

Rob Bernard
Regard d'expert

Robert Bernard

Chief Environmental Strategist at Microsoft

Today we think that there are around ten million different animal species on the planet, and most of those have not been catalogued. In order to preserve these species, we need to know what they are. The technologies we use to spot them have changed very little in the last fifty years.

Mapping an ecosystem is a good start, but knowing how it works and identifying the different species that live there is just as important. This is what the second Microsoft project is all about. The basic idea is to use mosquitoes to help list the animals living in a given geographical zone. “Today we think that there are around ten million species on the planet, and most of those have not been catalogued,” revealed Robert Bernard, adding: “We think that they are going extinct at one thousand times the natural rate of extinction, and one hundred times the rate that’s required to sustain human life. Yet we don’t understand the majority of species on the planet, and with today’s technology – this is how we track animals with collars, we hope they run in front of cameras that we’ve placed on trees and in areas around the world – it’ll take potentially 500 years to catalogue all the species on the planet.” With this in mind, Microsoft is trying to come up with a more effective solution, based, says Robert Bernard on the question: “What if I can turn nature into a sensor for nature?” Working along these lines, Microsoft is using drones to find mosquito nests in a given area. Then a trap with audio and video sensors is set up. The trap draws on machine learning to recognise the species of mosquito the experts wish to catch.

PROTEcting people from dangerous species

tiger

If, for example, you want to check whether there are tigers in the area, the trap can be trained to recognise the varieties of mosquito that are known to bite tigers. The trap also uses infrared light to identify those mosquitos which have blood in their abdomen and so will provide crucial information. Once an interesting specimen has been spotted, the trap emits small quantities of carbon dioxide and blue light to lure the insect inside and then closes on it automatically. “When the trap is full, it will send out a signal to alert us. We then just have to take it to the lab. In the space of three hours we’ll be in a position to know which species are present in this ecosystem and in what number,” explained Bernard. By analysing the blood sucked in by the mosquitos, the researchers can count the species they already know and identify new ones. The system is currently trialling in Florida and Texas.

The second aim of this project is to catalogue and prevent the spread of infectious diseases, up to 75% of which are believed to be of animal origin. The system enables specialists to identify upstream diseases affecting animals and take appropriate action before they start to affect human beings.

Ai for precision AGRICULTURE

Agriculture

Shutterstock

Precision agriculture for all

The aim of the third project is to bring agricultural activities into greater harmony with natural ecosystems. With this in mind, Microsoft plans to make precision agriculture accessible to everyone, including farmers in the poorest countries or working in remote areas that have no Internet connection. “Not many farms around the world have WiFi or a 4G network. Now, installing sensors for precision agriculture serves no purpose if you can’t get the data out. To do that we use TV White Spaces, unused broadband spectrum. This means we can transmit high resolution data in ten-kilometer hops, using a little bit of a solar cell,” Robert Bernard told us. Several pilot projects have been run in Africa and India where, according to Robert Bernard, farmers have seen their yields increase by 20%. This idea is not only suitable for developing countries but also for some of the poorer areas of rich countries. In fact, 23 million people living in the United States today have no Internet connection. Microsoft has therefore set up the Rural Airband Initiative in the US. The target of this five-year plan is to provide, for a modest sum, broadband access to two million people living in remote areas.

DEVELOPing the poorest rural AREAS

Agriculture

Microsoft has also set up a system that enables farmers to obtain aerial images of their farms without bankrupting themselves. Bernard underlines: “Today, if you want an aerial view of your farm you need a plane or a drone, which not everyone can have. This is why we’ve developed a system using a smartphone attached to a helium balloon, with an algorithm to stitch [the pictures] together. Then software is used to process the images and extract the key information. Additionally, we install low-cost sensors to measure – for instance – soil humidity or PH, as well as the weather forecast obtained using a mobile phone. We combine all this data together and with a layer of artificial intelligence on top we get a real intelligent scanner for the farm. Thus we can determine what sort of crops suit the environment best. Long term, with this type of initiative, our aim is to enable any farm anywhere in the world to make use of connected agriculture for a hundred dollars of investment.”

THE bigGEST names IN AI commit to WORKING FOR the planet

PLANET EARTH

Kurzgesagt 

A race among the big names in AI

Microsoft has a double interest in committing to help safeguard the planet. Firstly, protecting the environment is a basic challenge facing our civilisation, which nowadays is increasingly finding resonance among the general public. Secondly, of course, for enterprises such as Microsoft, Apple and Amazon, which constantly come into direct contact with consumers, this kind of initiative is not just a philanthropic exercise, it is also a business issue. Being seen to support pro-environment initiatives is good for their brand image, and so by definition for their revenue as well.

With this in mind, Microsoft has also committed to limiting its own impact on the environment. Robert Bernard reveals that the computer giant is now carbon neutral and is also planning to obtain half of its energy needs from renewable sources by next year. This will also involve taking a ‘greener’ approach to its data centres, as it is no secret that these facilities consume huge amounts of energy. In the long term, as well as presenting the company in a good light vis-à-vis the consumer, this energy policy switch will have the advantage of reducing both its dependence on fossil fuels and its operating costs.

real commitment or MARKETING POSE?

clouds

By no means least, these environmental projects will allow Microsoft to find ways of applying the results of its research into Artificial Intelligence. This technology is today becoming crucial in the rivalry between the US tech giants and each of them is trying to get the best out of it: IBM with Watson, Amazon with Alexa, Google with AlphaGo and Waymo, and Apple with Siri. Microsoft clearly has no intention of being left behind and testing out technology enables the firm to keep improving.

This strategy also helps Microsoft to reassure the general public about the growing use of Artificial Intelligence, which tends to make people anxious – whether we are talking about the much-feared ‘technological singularity’, killer robots, or just the prospect of mass unemployment as machines become capable of taking over most jobs from human workers. By showing that AI can also be put to work for the common good, Microsoft is cultivating a more positive image in the eyes of both the general public and the regulators.

By Guillaume Renouard