Transcript
HP: We love food, so let’s talk about food. In 2050, we’ll have 9 billion people on the planet and we’ll have to feed them three times. We have a challenge on our hands. I’d like to talk about AI in the age of climate change. This is the first talk I’ve had in the last two years where I will not be talking about generative AI. I’ll talk about AI that scales and that is reliable. Today we talk a little bit about climate change.
Why Should We Care About Climate Change?
What is climate change. In simple words, it’s the average global temperature that’s going up, and it’s having an impact. Why should we care? Is it one of these things where you can ignore and go, climate change is not real. People don’t care about it. It’s just something that’s been made up to get people to pay attention to something. Actually, we have scientific evidence that over the last 300, 400 years, the last 100 years especially, we have extreme weather conditions because the global temperature is going up. We have floods, droughts. We have heatwaves here in Europe that we are experiencing the past few years.
The unfortunate thing is it’ll get worse. Don’t worry, I’m not going to do a doom and gloom talk. We’ll talk about how we can possibly get out of this situation. There’s more from climate change. There’s ecosystem disruption. We are looking at a sixth wave of mass extinction right now. There’s a huge loss to our biodiversity, and that has a further impact on our environment. Coming to food, I think COVID-19’s a little behind us, but I can remember the trauma we all had when we went to the supermarket and we couldn’t find pasta. Imagine the situation constantly happening. There’s a large part of the world right now which has food scarcity. They have water shortage problems, and that in turn impacts human health. We should care about climate change. If we don’t want these events to keep happening, we should do something.
What is Contributing to Climate Change?
What is contributing to climate change? How did we get here? Electricity and transportation and heat, all of you would take a guess. The other aspect that contributes a lot to climate change is actually the food system. The food system includes not just the crops that we are growing, it’s the animal husbandry that we have, the transportation, the logistics, the packaging as a whole contributes to 33% of greenhouse gas emissions. Just in one year, we’ve generated 47 billion tons of greenhouse gas emissions, and we’re looking at compounding that to about a trillion over the last decade. That is quite scary. This is a statistical estimate of the temperatures in Svalbard, which is way up north, closer to the Arctic Circle. For those of you who are not fully aware of what’s there in Svalbard, we’ve stored all our grains and seeds in a cold vault there and probably our GitHub repositories as well, just so that we can bring back Earth if there’s a catastrophe, and that plays as looking at increase in temperatures.
Background
I’m Nischal. I’m the Director of Data Science and Data Engineering at Agreena. I’ve been in this space for about 14 years, building AI and ML systems across different kinds of landscapes.
How Did We Get Here?
How did we get here? Why is it worth for us to look back at how did we get here, and how can we use that as a way for us to move forward? In the last century, there’s been a lot of focus in building organizations where the incentivization is maximum profit. You are not thinking about what is the impact of this profit, what is the impact of industrialization? There’s been economic insecurity and climate change denial, so people have been constantly talking about climate change as something that’s never happened.
There’s a lot of economic insecurity that stems from the wars, the geopolitical tensions, and that’s adding to how we take decisions. There’s of course the consumer demand and overconsumption. We’re all looking at prices, and there’s an incentive for us to hit any malls as long as there’s huge discounts that are pushed in. There’s a lack of regulations that don’t really cater to environmental preservation. This is added to where we are. Industrialization and capitalism is not entirely bad. We’ve figured out that if we build a right economic model, it can thrive. We’ve been successful in making this model work.
What’s the Solution?
What’s the solution? I think drinking a pint of good beer during Oktoberfest could be a solution, but have you ever wondered what’s the carbon emission associated with drinking a pint of beer? I don’t think any of us would do that. When we drink beer, we want to enjoy the beer and the company we have, but when you look at the carbon emission, it’s about 205 grams to about 1,483 grams of carbon. That doesn’t mean we all stop drinking beer. That’s one other thing is, if we have to keep things sustainable, we can get wiser. Instead of drinking bottles of beer, it’s good to go to breweries where they’re brewing, and that reduces 40% in carbon emissions already. Enjoy your Oktoberfest beers guilt-free as long as you’re visiting a tent where they’re brewing beer. This shows a little bit about how little we know about carbon emissions. There are different levels of carbon emissions.
There is scope 3, scope 2, and scope 1. Scope 1 is organizations that are generating carbon through their company facilities, through their company vehicles. No organization is doing everything by themselves, they’re dependent on upstream and downstream organizations, and this includes harvesting crops, this includes all the logistic vehicles, packaging, everything that we have. There’s more to greenhouse gas emissions than meets the eye.
There are two ways forward for us. One part is reducing the GHG emissions which talks about all of the transition that we’re looking at in terms of green energy, bringing solar farms and windmills, and hopefully nuclear fusion, and reliable, safe nuclear fission energy. That’s one part of it. That’s a part where it’ll help us in reducing the amount of GHG emissions we have, but we do have to do something about historical emissions that we continue to have in the atmosphere and we need to remove them. There’s a lot of debate that we’ve seen in the past, where we suddenly go from looking at automobiles and saying, we’ve changed our cars and we have an electric car so we have no carbon footprint. The challenge is that even if the entire population ended up using electric cars, we would still be consuming about 99 million barrels of oil per day. Transition to electric cars is good in terms of reducing emissions but that’s not the only answer we have or we should look at in terms of fighting climate change.
How can we remove the GHG emissions? Do we have to build new machines? That means more carbon emissions. Do we have to think about new technologies that is not yet invented? How can we actually do this? The funny thing is, we have around us abundantly the biggest sink for carbon, which is the soil. The soil is the most biodiverse ecosystem on Earth, which nourishes all of us and sustains all of us. The question is, how can we use soil as a terrestrial carbon sink?
We have an answer. This is one of the answers we have is to move from our current agricultural practices to looking at something we call as regenerative agriculture. Regenerative agriculture is a way to look at enhancing soil health and biodiversity. This means we have to change our conventional thinking of agriculture which often leads to soil degradation and loss of biodiversity. We have about 60 more harvest years left where the topsoil is going to hold up, but after that, 60 more harvest cycles, the amount of carbon emissions are going to drastically increase if we continue to do industrial agriculture the way we are.
A little bit in terms of how can we implement regenerative agriculture. Sounds like, ok, we have an answer. What do we have to do? One, we disturb soil as little as possible after every harvest cycle. We have crops that we have in between the harvest cycle so that they’re retaining the soil and they enhance the carbon stocks. We move away from synthetic nitrogen towards more organic manure. We have crop residue. Instead of cleaning up the farms, we just leave the crop residue in there, which becomes manure and further holds the soil better. These are easy things. The question is, if we were to do regenerative agriculture, how much carbon can we actually remove from our atmospheres? There’s an estimate that we can remove 23 gigatons of carbon dioxide from our atmosphere by 2050 if as a whole we start to adapt regenerative agriculture. This means change, and change is very hard.
If you look back at the different ages that we’ve lived in over the last three decades, if we didn’t have an incentive to change, for example, when the internet era came in, if there was no search engine that we could work with where we could search things on the internet, would all of us build these communities around it? If we didn’t have companies investing in the space of AI, would we have what we have right now where we have breakneck innovation? There needs to be an incentive for everyone to change.
Unfortunately, climate change is not a big enough incentive. We’re thinking about, we have this century of industrial agriculture, and the moment you want to change, people look at it and go, ok, moving into regenerative farming is costly and risky because we know very little about it, so there’s no reason for us to go into regenerative farming. There’s potential short-term losses because when you change methods, your yield might get lower, you’ll have to invest in machines, you have to go through a whole new education. This is another challenge that we have. The biggest one is, when people look at their balance sheets and the balance sheets are red and we’re expecting this group of people to start thinking going green, there’s no way that that change will kick in.
Challenges and Mitigations
What do we see as challenges and how can we mitigate them? There’s a lack of financial incentive and aid for farmers. There’s lack of credible and verifiable data for financial institutions to actually give out green loans or build new insurance products. For organizations, every organization, Microsoft, Google, Unilever, Pepsi, they all have mandates to become carbon neutral or offset their carbon emissions, but how will they do this by themselves? They don’t have a credible way of finding or buying commodities that would help them with reduction in their carbon emissions. There isn’t a monetization, there isn’t an economy that will actually facilitate farmers with practices and reduction in carbon itself. There is no economy, there’s no market in there that will support this change. We have to build a new economy. We’ve seen that with industrialization and capitalism, with the right incentives, with the right set of guidelines in place, we can actually build a successful economy.
Let’s start with building credible data. We’re in a tech conference, it’s important to talk about how we can actually solve this problem with technology or one part of it. This is where we do things that don’t scale. What’s the easiest thing to do before we introduce machine learning? We can go to every farm on the planet, we can take photos. We can take samples of the soil. We can annotate those photos, we put them all in a system, we measure the amount of carbon in the soil. We generate these reports. We have to do this two or three times a year so that we’re continuously monitoring the farms, and we only have 600 million hectares to go look at. If we do things in this fashion, we’re probably adding to the carbon emissions and actually helping because you need people to go onto the farms, you need people to travel to different parts of the world.
If you have to do this two or three times, it actually gets very expensive. We can actually avoid things that don’t scale, and this is where AI and machine learning can help us. We have a lot of satellites up in the sky to a point where there’s a startup that’s looking at figuring out how to clear all the debris that we have around our planet because of the amount of satellites and the space debris that we are leaving behind. With the help of satellites, we’re looking at satellites that are spectral and optical, so which means there are satellites that are taking photos of everything as is, and there are satellites that are using spectral waves to look at the same area.
That gives us two different dimensions of looking at things. We can leverage the data from these satellites, and instead of sending humans to every single farm, we can actually get all the farm data that we possibly need by making use of these satellites. European Space Agency has several satellites up in the sky that all of you can play around with, without having to pay a penny because they believe in democratizing and usage of the data.
Once we have images, both spectral and optical, we actually have machine learning models, small machine learning models because we need to run them at scale, and we want to avoid having a big carbon footprint ourselves, and we need higher reliability. We run these small machine learning models and we predict agricultural practices that are being implemented by farmers. These are some of the models that we have where we can look at every single field. I’ll talk a little bit about the number of countries we are monitoring. We can look at what are their tillage practices. You can see from the heatmap that some tillage is conventional tillage and some of the tillage is not conventional, which is regenerative in nature. We can look at every single crop type that is being grown.
This helps us understand and think about food security, and understand a little bit about the risk to supply chain. We also look at if the farmers are growing cover crops between harvest seasons and how they’re actually harvesting their crops. We have layers of information now powered by machine learning models at a granularity level of a field, so every single field. A farmer has a big plot. He has several fields in these plots. At every single field, we know what practices the farmer actually has.
Once we have this, because this is not a one-time thing, we are looking at enabling farmers to do this across 10 years, and monitoring them for at least 10 years to see their practices. We take the output of our machine learning models and we run them through a time series forecasting to make sure that we can identify the differences in the practices across time that the farmer is implementing. These numbers have already changed. We’re monitoring 2.7 million hectares, actually in 23 countries now, mostly in Europe and Asia. Looking at this, we understand that there are differences in practices across different geographical zones. We actually have smaller machine learning models for each of the geographical zones because we want to have more reliability in our machine learning models than generalization, which gives us credibility in our predictions and how we use them.
The question that we would obviously ask next is, how can we translate these agricultural practices into understanding how much carbon that we’ve sequestered into the soil? For this, we’ve built something called as the carbon science programs. The carbon science programs have methodologies where we look at the farming practices, we look at the ground truth data that we’ve collected, we look at the data that the farmers are giving about their farm, and we convert them into verifiable carbon sequestered units. Given that we monitor the farms continuously, we understand the change that they are bringing in so we can almost accurately predict the amount of carbon that is going back into these soils.
Just because I say that our machine learning models are good, our carbon science programs work, how can we make sure that our AI and carbon models are accurate? What you see there as an image is us working with agencies, that’s a soil sampling machine that goes onto a farm, picks up soil for about 10 centimeters below the ground, and we look at every single layer of soil to understand what’s happening with the carbon that’s going back into the soil. We have people going onto the ground, taking photos, generating ground truth data, and looking at what is actually happening on the field. We have a lot of ground truth data to a point where we’ve amassed about a million plus images, amounting to 5 terabytes spread across different seasons, spread across different geographies, so that we have a lot of reliability in our machine learning models.
Of course, we deal with a lot of imbalanced classes, we deal with a lot of imbalanced data, so we have to make sure that we have techniques in place that takes this imbalance into account so we have a higher reliability in our models. Apart from all of this, we have third-party nonprofit organizations who look at our entire process, all the way from collecting our ground truth, how we use it to train our machine learning models, how are we predicting it, how do we use it in our carbon science programs, which is vastly different from a lot of conversations that we have about generative AI and large language models, where reliability and benchmarks is always a perspective rather than the science that’s involved.
Why do we care so much about data credibility? There’s a lot that’s happened over the last decade with something we call as greenwashing. We see a lot of organizations that use PR and marketing to appear environmental-friendly, but when you go and look a little deeper, they have very little data to show that they are actually environmental-friendly. We want to make sure that not only is our data credible, it’s verifiable, so that we avoid falling trap to the idea of greenwashing.
We’ve been able to look at creating verifiable data. We’ve been able to look at, how can we convert that data into something that potentially corporates could buy because there’s credible assets that we can create from this. The way we’ve done this is we have remote sensing data powered by machine learning, data from farmers, data from other public registries as well that help financial institutions to have access to credible data. Using this, they can provide green loans to farmers, they can refinance farmers, and farmers can go through this transition. You have insurance companies that can create new insurances to support the short-term risks that farmers might potentially have. Using this data, we can create carbon credits that organizations can then pay to buy so that they can offset their carbon footprints.
The next part and why capitalism has worked to a large extent is to create independent markets for people to be able to trade commodities. We see a new market that is emerging, which is called as the voluntary carbon markets. We also talk a little bit about scope 3 emissions. It’s a decentralized marketplace. I know I’m not talking about Bitcoin or Ethereum. It’s an actual marketplace where you have individuals such as farmers who can take the carbon that they have sequestered and convert them into carbon credits and put them in this marketplace, and the organizations can buy these carbon credits from there.
The scope 3 emissions reporting helps organizations understand what is their actual carbon footprint, which is very important for them to then figure out a plan to say, if we have this much carbon footprint, what do we have to do to offset that footprint? There is a bit of a challenge in terms of pricing because, who decides pricing for these carbon units? The demand is usually driven by corporates who want to offset their carbon footprint. This is where regulations and education play a huge role. Without strong government regulations to offset the carbon footprints, we’ve had the Paris Climate contract in place, an agreement, we said a lot of companies have to hit their carbon neutral goals by 2030, but there’s an ongoing war, there was the pandemic, and suddenly we’ve forgotten about an agreement we’ve had in place.
There isn’t enough push from the government bodies to regulate these organizations. There’s a lot of work that’s going on, but not at the pace we needed to. We do see that the regulations are picking up. We see demands growing more in terms of organizations wanting to offset their carbon. This is a part where usually I try to not be political in most of my talks, but it’s important for people to vote, because when you vote governments coming to power, more than a lot of the politics that they’re talking about, it’s important to look at the policies they have towards climate change. Because this is something that is going to be impacted beyond their four or five-year term of governance. If you have the power to vote, please go and vote in your respective countries.
Closing the Loop, Farmer Incentives
We now have a market. We have assets that can go into the market, and these assets are verified by data. We’ve connected three different pieces where we have carbon credits that can go into our market. These carbon credits are coming from responsible machine learning models with remote sensing data. We have organizations that can operate in this carbon market. The most important thing amidst all of this, the true agents of change are actually farmers. We have to build an economy around farmers and not so much around organizations as we’ve done in the past. Based on all their agricultural practices, based on everything that they’ve been implementing over the past few years, we need to create incentives for farmers where the selling of credits on the carbon market goes back to the farmers, they get paid out.
It’s a passive revenue stream that is constantly supporting them to bring about this change. Because of the credible and verifiable data, they can now get refinancing from banks, and I’m talking about banks such as the World Bank, which have climate funds that they can use to enable these farmers. As a good byproduct, farmers now can digitally monitor their farms over several seasons, thereby becoming a part of the community, learning from other farmers. Farmers are a very community-driven organization, so when they get to see how good they are faring against their neighbor farmers, it’s like a competitive spirit they have where they want to one-up the other farmer just to get better at what they do. This is closing the loop where we are creating an economy based on verifiable, credible data. We are creating a new economy of sustainable development. This is very important for us to scale.
Conclusion
We are not the only ones here. This is not the only way to do it. I’ve always been interested with AI and agriculture, but even a year ago, I had no idea about voluntary carbon markets. I had no idea about regenerative farming, and how machine learning and data can actually be used to create a new economy. For me, the key interest in sharing this information or knowledge was, there’s a lot going on in this space. If you’re sitting there wondering, are there few organizations? If I want to work in this space, what can I do? There’s a lot of investment coming in. We have about $150 billion invested across 4,000 deals in the last 3 years alone.
In just this year, we’ve seen that $3.5 billion have been invested as of June 2024 in climate tech companies. There’s a whole lot of companies based out of Europe, based out of North America that are looking at this and trying to build this economy. If you are interested to work in this space, of course, at some point, we’ll start hiring again. This is not a hiring pitch, but you can look at what we’re doing. We can actually build a greener tomorrow. There’s a lot of doom and gloom when people talk about climate change, but if we get ourselves to come together and think of a way to get out of this situation by building sustainable economies and new economies, we can actually build a greener tomorrow. We have to save the planet because this is the only planet we have with beer.
See more presentations with transcripts
