Microsoft’s Planetary Computer for a Healthy Earth

While everyone seems to know about AI playing chess and self driving cars, there are many more uses of AI both good and bad. AI based systems can be used to strengthen airport security by recognizing faces and suspicious behaviour. It can help scan medical records and identify problem areas, suggest medication and interventions. At the same time AI can generate fake news, fake videos and imitate human voice. Computer games, apps and websites can be made more addictive and manipulative. Scientists have argued that a goals driven system like AI which is taught to maximize revenue, increase compliance, would increasingly become more and more efficient and therefore less empathetic and more controlling.

But AI’s ability to scale, learn and adapt can have some remarkable impacts on climate change, biodiversity, management of forests and oceans and overall planetary health. Two-and-a-half years ago, Microsoft launched AI for Earth program to put artificial intelligence technology into the hands of the world’s leading ecologists and conservation technologists, and organizations around the world. Continuing on its journey, Microsoft has launched an ambitious program to aggregate environmental data from around the world and put it to work in a new “Planetary Computer.” This project is over and above Microsoft’s ambitious project to be carbon negative by 2030, help suppliers and customers to remove their carbon footprint and to remove all historical carbon generated by Microsoft by 2050. Microsoft has also announced a new $1 billion climate innovation fund to accelerate the global development of carbon reduction, capture, and removal technologies. Microsoft believes that maintaining nature for the benefit of current and future generations is one of humanity’s greatest challenge for which technology can be deployed at scale.

Over half of the global economy is directly linked to natural ecosystems1 but these very ecosystems are threatened or are already in decline2. We are facing a mass extinction of species on the planet as global wildlife is down by 60 percent and reducing further. Global wetlands that purify and store water have been reduced by 87 percent and coral reefs have declined by 50 percent. Insects diversity and abundance has fallen dramatically and this can potentially be catastrophic as more than 75 percent of the world’s food crops are linked to the ecosystem services they provide. The natural ecosystems and wildlife need urgent help.

Further, biodiversity and climate change are large complex problems with direct, indirect and network effects. For instance, Climate change can result in higher potential evapotranspiration from soil and plants. When accompanied by stable or increased precipitation it can help in greater carbon sequestration by forests. However, climate change can also reduce or change patterns of precipitation and increase the strength of hurricanes that make landfall, causing adverse impacts on rainforests.3 Hence a system that can function at scale, is based on data and rooted in science is needed.

The responsibility frameworks of today also look at post facto interventions to define social and environmental responsibility. Today’s world and increasingly in the future we are looking at real time responses to everything. Sensors on machines give real time information for companies to generate so that they can capture breakdowns. Wearable devices such as smart watches capture health data and can predict a heart attack three hour before they occur. Algorithms can take decisions in seconds and reach tremendous scale with big data. Responsibility in the 21st century has to be about creating a better world today and not waiting for harm to happen and then correcting it. Since the world is changing at tremendous speed we can’t really have responsibility models that rest on mere compliance or remediation. We need better decisions NOW!

Further, the connected world magnifies the smallest risk. While the cross-cutting nature of connections provide huge benefits they are also incredibly difficult to implement because of the sheer complexity and scale of the task. The system therefore needs to be changed from the inside. From the areas that drive value and not just the areas that contribute to risk. Assessing the planet’s health must become a more sustained, integrated practice that allows us to understand exactly what is happening in real time to enable smart decision-making. While the Google WWF collaboration would work on open data, Microsoft has announced that the data collected from their project will help enable partners and customers to use the resulting output to enhance environmental decision-making in their organizational activities.

These initiatives highlight the importance of AI for public good and the way our commons need to be governed. Data about the global commons such as forests, oceans, air, corals is increasingly going to gain prominence and critical mass. Machine Learning and AI will need deep science and a nuanced approach to manage and intervene. A planetary computer is a dream – How should it be used and what should be the laws governing it? Further, what should be the moral code to determine use and impact?

The measures to control the pandemic have clearly shown that complex systems need complex answers and that reaching these can be an iterative process. Further, all AI based algorithms failed during the pandemic because they were based on historical data. In a world changing at warp speed, the future can’t be more of the past, hence new frameworks and models are needed. Experts are repeatedly warning that the frameworks governing AI need to be capable of ensuring accountability over the long term. What may seem unbiased and ethical could develop unconscious biases over time and replicate them to scale. Efficiencies and optimisation – is great when you want machines to run in a certain way. But within AI it has different connotations. For instance to improve commercial outcomes the AI system may sack employees or indulge in unethical practices. In the case of AI for the global commons, we need to be aware that nature runs its own course – ecosystems flow and ebb and so do wildlife and plants.

Microsoft, seems to understand these issues and the company’s blog by its President Brad Smith states, “We do not know enough about species, biodiversity and ecosystems that are vital to our health and prosperity. Simply understanding where the world’s forest, fields and waterways are remains a daunting task of environmental accounting. Understanding what species call those ecosystems home or why they thrive or decline is largely unknown. We simply can’t solve a problem we don’t fully understand.” It goes on to add, “It should be as easy for anyone in the world to search the state of the planet as it is to search the internet for driving directions or dining options. We must use the architecture of the information age – data, compute, algorithms, application programming interfaces and end-user applications – to accelerate a more environmentally sustainable future.”

Business responsibility therefore needs a modern context. It needs to bring in the many connections and technology interchanges that really define the 21st century as well as build frameworks that reflect the changing realities. Microsoft’s Planetary Computer project is showcasing that environmental assessment can be done faster, cheaper, and at a truly global scale.