Artificial Intelligence can lead to greater impact on sustainability than ever before: Microsoft's Lucas Joppa
Sonal Khetarpal May 3, 2019
As climate change becomes a hard-hit reality corporates around the world are wearing their green thinking hats to increase sustainability initiatives. Microsoft recently announced that it will double its internal carbon fee to $15 per metric ton on all carbon emissions. The idea is to use the funds to maintain neutrality and help it take a tech-first approach to sustainability. BusinessToday.In spoke to Microsoft's Chief Environmental Officer Lucas Joppa on Microsoft's initiative that is closest to his heart, AI for Earth -- a five-year, $50 million initiative that offers grants to organisations who are using AI to solve the most pressing environmental challenges.
Business Today: You have been speaking about doubling down the action on AI for Earth. Please elaborate on what it entails - increasing the grant size or technological support or more?
Lucas Joppa: The AI for Earth program is all about deploying machine learning technologies to problems in the areas of agriculture, water, biodiversity and climate change and the thing about the machine learning, AI is that it's an extremely data-intensive approach. It is in this line, we will be making two new investments in the AI for Earth perspective, and in particular invest in the hosting of some of the world's largest and most important environmental data sets for our growing portfolio of grantee organisations over 50 countries to leverage in their work.
As we worked with these grantees we realised that the datasets they care about or want to use are not just large but are derived from increasingly sophisticated sensors on satellites and they require those datasets right next to the computer they use. And so we've taken that feedback and we're committing to investing in hosting those large environmental data sets to further accelerate our grantees' success.
We're also really focusing on the way that we can help our grantees deploy their models so it is not locked up in a code base but are accessible on the internet through application programming interfaces.
BT: How has the AI for Earth program evolved over the years? What where some of the learnings?
Joppa: One of the most important thing is we're constantly updating our program based on feedback and interactions with our grantees. We started with a small $2 million grant and based on the user demand grew it to $50 million.
We also created a data science team inside Microsoft to engage directly with our grantees because we heard from them that access to expertise was extremely important. So now we don't just provide things like cash grants or grants to our cloud infrastructure, we also provide grants for machine learning specific tasks like data labelling and we heard from our grantees that they would like to see us host data. So now we're committing to that.
Everything we're doing is driven by the requirements of our portfolio of grantees because what's really important for an organisation like Microsoft to recognise is that while we may be a significant technology company, we also have to recognise that most of the expertise in the environmental science space lies outside of our four walls. It lies within our grantee portfolio. So, it's really by working together that we help them and they help us understand the problems and the promise of working in the environmental sustainability space. Together we identify new ways that we can achieve greater sustainability.
BT: Please share about some of the AI for Earth grantees that are close to your heart.
Joppa: One of the applications that I find just incredible, both in application and as well as in journey to build that application, is an organisation called iNaturalist. It was originally designed to connect environmental experts with one another and allow them to share data that they collected with photos. Principally, photos of plants and animals and describe them both by species, name, and place and time and contribute an unprecedented amount of data to the global knowledge base on where the species are and how their populations are faring.
It started out as a small master's project at Berkeley and the number of people has not significantly grown. Recently, they have created a new smartphone app called Seek, it tells which species are near you or likely to be near you based on the location. As one turns on their camera and scrolls around the environment, it immediately detects and classifies the species of trees, plants, animals that you're seeing in real time.
What's incredible is how just a handful of people could start with nothing and build a community of millions of users, collect unprecedented amounts of data, train an incredibly large and complex machine learning model in Microsoft's cloud and then deploy the results of that model onto a smartphone in everybody's hands that allows you to classify your natural surroundings immediately in live video. Ten years ago, it was impossible to do that. Five years ago, you would have needed a team of hundreds of people to do that. This year a team of about five people did it.
BT: Beyond AI for Earth, what are the new initiatives, which Microsoft is looking at to reduce carbon footprint?
Joppa: Last July, I became the chief environmental officer for the company. It's the first time the company has had somebody in that position and the remit there really is to look left and right or 360 degrees across or around the company and understand not just how we can do things like reducing our operational carbon footprint.
We have significant renewable energy goals and we've been surpassing early every goal that we've set. We already have an outstanding commitment to reducing our carbon emissions by 2030 by 75 per cent. We're really focusing on these kinds of core areas of operational sustainability.
Then we have the circular server initiative, where we're putting in place the technology to track, monitor and then ensure the resiliency of our hardware assets as they move through the supply chain. One example of that is a new water replenishment commitment to replenish the water that we use in our data centres in water-stressed regions of the world.
But I'm also tasked with looking across the entirety of the rest of the company; our products, our services, the way that we engage with customers and partners to think about how we can use those aspects of our business to extend well beyond our own four walls and create impact for environmental sustainability.
To ensure that we leave no stone unturned. One example that I like to talk about is the AI for Earth grantee SilviaTerra. They use machine learning algorithms on high-resolution satellite imagery to build our data sets or maps down to the individual tree and so they can identify the trees and can tell you how many more might be in there, the carbon sequestration potential of that tree, what species it is etc. etc. We've been investing in them through our AI for Earth program to build up their capabilities so that they can produce the first map of its kind in the United States -- a tree level inventory of all of the forests here in the U.S. It's quite exciting to see how a small scrappy start-up can take advantage of our machine learning technologies to produce this incredible data asset and can be successful as a small business themselves.