Skip to content
InTechnology Podcast

What That Means with Camille: Mineral Exploration with AI (131)

In this episode of What That Means, Camille gets into batteries and AI with Jef Caers. The conversation covers the necessity of lithium-ion batteries and the metals they require, the AI being used to speed up the discovery and mining process, and the human impact of mining for those metals.

To find the transcription of this podcast, scroll to the bottom of the page.

The views and opinions expressed are those of the guests and author and do not necessarily reflect the official policy or position of Intel Corporation.

Follow our hosts Tom Garrison @tommgarrison and Camille @morhardt.

Learn more about Intel Cybersecurity and the Intel Compute Life Cycle (CLA).

Lithium-Ion Batteries and Metals

Jef Caers explains how lithium-ion batteries are the best possible batteries the tech industry can use right now. As we need more batteries to power our technology, especially as we move to renewable energy, we also need all the metals that go into making batteries. These metals include cobalt, nickel, copper, and—of course—lithium.

Batteries are essential to the process of harnessing, storing, and transporting energy. Because renewable energy like solar, wind, and hydro are intermittent in their production, we need the batteries and the materials they require in order to offset the looming environmental danger caused by CO2 emissions.

Using AI to Find More Metals and Meet Demand

In order to meet this high demand for more batteries and the metals needed to make the batteries, scientists have to find more metals in the Earth. Unfortunately, this is a very time-intensive and logistics-intensive process. With the help of AI, Jef Caers shares, the discovery of new metal reserves can be sped up as much as from a 10-year process to a 2-year process.

The AI used for finding these metals is different than usual. Currently, AI in this case is used to determine where and how to acquire the data through sequential decisions, but it doesn’t use the data itself. While this is helpful, it also hinders speeding up the discovery process even further. However, there is hope for the future of using this type of AI for the discovery of resources beyond our planet, such as on the Moon and Mars.

The Human Impact

Finally, Jef emphasizes the importance of the human impact of mining for these metals. There is currently no standard for involving the communities where metal discoveries have been made and deciding how to go about mining in the area—or if the mining should be done at all. From countries like Greenland to indigenous communities around the world, this is an angle scientists and mining companies cannot ignore.

Jef shares how Stanford University’s Doerr School of Sustainability is focusing more on environmental justice and the human impact of mining and all technological progress.

Jef Caers, Professor at Stanford University

Jef Caers lithium-ion batteries AI renewable energy

Jef Caers is a Professor of Earth and Planetary Sciences at Stanford University’s new Doerr School of Sustainability. Additionally, he is the director of the Stanford Center for Earth Resources Forecasting. Jef has earned a Ph.D. in engineering as well as an M.S. in Mining Engineering & Geophysics from Katholieke Universiteit Leuven in Belgium. He was awarded both the Andrei Borisovich Vistelius Research Award and the William Christian Krumbein Medal by the IAMG, and his work has been widely published throughout the scientific community.

Share on social:


[00:00:00] Camille Morhardt: Hi, I’m Camille Morhardt. And today we have a prequel to our podcast. At the very end of the conversation I have with Jef Caers, he got contemplative and he said something that I think was really interesting; just wanna let this play before we go into the interview.

[00:00:17] Jef Caers: Our field is starving for talent and lots of talent goes into cybersecurity or Facebook media. And then when I give talks to engineering departments, and people are like, “Whoa, these are really cool problems. I didn’t know they existed.” And I’m like, “Well, these are the problems that are gonna save the planet.” Net Zero, it has to save the planet, right? If it doesn’t happen, we know what’s gonna happen. We see it happening today. You live in California, it’s 116 degrees in the Bay Area, that’s never happened before. So things like that, things are changing and Net Zero is the solution to that, but the plan to implement it is there either, right? And that’s where we need people to come and help us because they have the talent.


[00:01:07] Intro: Welcome to What That Means with Camille, companion episodes to the Cyber Security Inside podcast. In this series, Camille asks top technical experts to explain, in plain English, commonly used terms in their field, then dives deeper giving you insights into the hottest topics and arguments they face. Get the definition directly from those who are defining it. Now, here is Camille Morhardt.

[00:01:39] Camille Morhardt: Hi. And welcome to this episode of What That Means. I have with me today, Jef Caers, who is a Professor of Earth and Planetary Sciences at Stanford University’s brand new Doerr School of Sustainability. He is gonna be talking with us about how we’re using artificial intelligence to seek resources that we need to move into this new world of renewable resources. We’ll get more into it. Welcome to the podcast Jef.

[00:02:10] Jef Caers: Glad to be here.

[00:02:11] Camille Morhardt: It’s really good to have you. Okay. I have to ask: earth and planetary sciences? Does this restrict you now from looking into resources on the moon?

[00:02:21] Jef Caers: No we’re, in our department we’re looking at the moon, we are looking at Mars. Everything’s very exciting because studying other planets and planetoids is very interesting also understanding our own earth and how it’s formed.

[00:02:34] Camille Morhardt:  Yeah. I forgot to mention also you’re Director of the Stanford Center for Earth Resources Forecasting. What exactly does that mean?

[00:02:42] Jef Caers: Yeah, we are a group of students and post-docs research staff, about 20 of us and we are working on the general problem of exploration, exploitation of the earth resources, whether that’s minerals or water or energy, and even subsurface storage, for example, carbon. Of course, CO2 is gonna be very important in the future.

[00:03:02] Camille Morhardt: Okay. Here’s my first question, cobalt, what is it used for and what is the forecasted demand for cobalt? How is that increasing; and what is the known supply of cobalt in the world, and what are we gonna do about all those things? [chuckle]

[00:03:23] Jef Caers: Yeah. Well, that’s a lot of questions in one, but it’s a very good question to start with, I think. So cobalt is used in the lithium-ion batteries and particularly in the cathode, and it’s particular element that is ideally suited for that, in that combination. I’m not gonna go into the numbers, but safest to say that we have significant shortfalls–if you think about dollars, it’ll be like two to trillion dollars from now to 2050 if we’re thinking about changing over our vehicle fleet into electrical vehicles. That’s posing a significant challenge particularly on the supply side.

And since cobalt is found only in few countries so far in large quantities–in fact, about 60% of the cobalt comes out of the Congo– that potentially lead to conflicts like we have had with oil and gas. And so part of my research is to help increase that supply and increase that supply relatively fast, so that we can avoid all these problems that we’ve had with oil and gas. And of course, that we don’t wanna get with transitioning to renewable energies.

[00:04:29] Camille Morhardt: I do just wanna back up briefly, I think probably everybody gets it, but I just wanna make sure we’re clear on why batteries are important as we move to renewables.

[00:04:39] Jef Caers: Yeah. The problem with renewable energy, particularly solar and wind is their intermittency. And so living now in California, yesterday was a big heat wave, so electricity demand goes up, and after four o’clock, the sun goes down. And so then everyone is supposed to turn off their air conditioning after four or five o’clock for that reason. So this intermittency is something that we need to be able to deal with and one of the ways to deal with is storing the energy of these renewables into batteries, and thereby being able to use that energy for our transportation.  Because even right now, our electricity grid is not quite clean, it’s still dirty. Natural gas is still being used, of course, particularly in California. And so if you’re charging your vehicle in California, that will be about 30% supply of renewables, but 70% is still non-renewable energy. And so that’s something that needs to change and batteries are a key component for that for storing energy.

[00:05:39] Camille Morhardt:  Okay. So batteries for fixed usages like in buildings and also for mobile usages like vehicles.

[00:05:48] Jef Caers: If we’re talking about buildings, then yeah, we will be installing batteries. For example, I’m going to get solar panels combination with that, I’m gonna have a battery in my garage. I have an electrical vehicle. The appliances that I have are all electric, and so except for heating–I can live off the grid so to speak with only solar energy. That is something that more and more is being encouraged.

Of course, the new bill that was just passed is also very much encouraged, there’s also tax credits that are coming, certainly incentivizes a person like me to go fully solar. And in San Francisco, this is certainly possible. I think it’s a great trend that we’re going into.

[00:06:24] Camille Morhardt: I was shocked to learn that we’re somewhere on the order of 10 pounds of cobalt going into an electric vehicle.

[00:06:36] Jef Caers: Yeah, it’s not just cobalt; I think that often the discussion goes on cobalt because of course, we hear about the Congo and the children mining the minerals. But there’s also other metals that are very important. So for example, nickel is a good second to cobalt in terms of a cut out material. Then we also need copper if we’re going to build more wires and chargers and things like that. And of course, we also need more lithium.

And it seems to me that the real supply crunch is gonna hit us around 2030, 2035, because we know that a lot of countries and corporations have promised us that by 2030 or 2035, we should be going EV. For example, California, you will now be able to buy an electrical vehicle in 2035. Most European countries will be fully EV by 2035. So that is not just a supply crunch on the cobalt side, but also on the nickel and the lithium and the copper side. So it’s not just about that one particular element.

[00:07:36] Camille Morhardt: Okay. So now we come to the, what are you gonna do about it?

[00:07:39] Jef Caers: Yes, there is a simple solution and the simple solution is to find more. So what we want to do is to speed up the discovery of these particular minerals. And in over the last, I would say, 10 to 20 years, there’s actually been a decline in the discovery of these minerals. And that has a lot to do with the fact that the easy deposits have been found–the one that you can see on the surface. So we’re now looking underground. So it becomes much harder to do, we’re also gonna have to do it faster. And so that’s where AI comes in, because a traditional way of mineral exploration is still very manual and expert driven. Geologist goes on the field, studies rocks, comes back, analyzes those rocks, makes models, and usually there’s one or two people. And of course, if you’re going to go to meeting the supply over the long term, that’s not gonna work. So we need to increase this discovery rate, which also is very important in mitigating conflicts or other things.

[00:08:39] Camille Morhardt: Interesting because one of my best friend’s fathers when I was growing up was a geologist for a major oil company, and flew all over the world and made these predictions. He was that, what you just described.

[00:08:50] Jef Caers: Right. Exploration geologist. Yeah.

[00:08:52] Camille Morhardt: Yeah. Why is AI better? How is it going to work?

[00:08:57] Jef Caers: Well, for one, it’s not used today at all by major mining companies. I collaborate with a startup company called KoBold Metals, and that company is founded with the idea of using artificial intelligence to speed up discovery. There are a number of application areas that we have to talk about, I think, with regard to what this AI does. One of the major problems with mineral exploration and gas, exploration is that you need to try a lot to find something, because if you’re looking for an ore body, for example, in the subsurface, well, lots of things look like an ore body, they’re mimicking an ore body. And so you’re thinking you have something, but in reality there’s nothing there; we call that the false positive problem.

Right now, the false positive rate is about… the discovery is about one in 200 times trying. So we need to create that and go down to maybe one in 50 times trying. That means, number one, we have to look at more data faster and also use artificial intelligence decision science in order to make better decisions. Rather than have an expert person making one decision at a time, now we need to leverage this massive amount of data that we have over the entire planet and make better decisions related to how that data can be used in mineral exploration.

[00:10:18] Camille Morhardt: What are you worried about in this conversion to AI? Are there any major concerns and do we have enough data to even feed into the model at this point?

[00:10:28] Jef Caers: Yeah. So that’s one of the major questions. We don’t have enough data to feed into the models. Our AI that we are developing is not an AI that uses data, but determines where and how to acquire data. This is an AI that can decide where in the world or what technique should be used in order for that to do that. And that’s not a single decision; we call that a sequential decision, because it’s not gonna be in one step. You’re not gonna drill one borehole and say, “Yay, it’s great or it’s bad.” No, it’s gonna be multiple of these steps. And so artificial intelligence is really great at solving problems like that, like self-driving car problems, these are similar problems or chess playing problems.  These are problems where you need to make sequential decision, decision after decision in an uncertain world, uncertain environment. And so that’s the techniques that we use to do that.

So in a way it’s a very data sparse problem rather than what we are used to hearing about machine learning and deep learning as a very data dense problem. I use the opposite, we need AI to decide what the data should be and that is where the acceleration takes place, because right now it’s just one person deciding that.

[00:11:42] Camille Morhardt: And how do you get there?

[00:11:44 ] Jef Caers: It’s a lot of computer programs and a lot of computer modeling. What we’re basically doing is we’re modeling the future, right? Is like saying, if I would be taking that data, what would that effect be? How would that, for example, reduce uncertainty? How much grade there is or how much volume of ore there is?  And so, predicting how data will affect our uncertainties is really key to addressing this false positive problem. If you don’t understand uncertainty really well, then you’re thinking there’s something there. And so you will go out and collect the data. Well, it turns out, well, that’s a waste of time, this is data you shouldn’t have collected, time you shouldn’t have spent. And so, that’s where we can improve, we save money and we make it faster.

[00:12:31] Camille Morhardt: So with the AI exploration techniques that we’re using, are they going to come with incremental advances or are you expecting some breakthrough advance after gathering a bunch of data?

[00:12:44 ] Jef Caers: We’re tuning these AI to the current experiments that are running. We’re exploring in Australia and Zambia and Greenland and all these places. And I think once we hit the way to do it, then it explodes and you can use it everywhere else because the way it’s set up is general. We don’t hope it’s incremental, we hope that it does explode out and discovers much more in a short amount of time.

[00:13:11] Camille Morhardt: Jef, are you optimistic?

[00:13:14 ] Jef Caers: I’m pretty optimistic. Yeah. I’ve seen already the first results. The company has been achieving in two years what a normal mining company would achieve in 10 years. Right? They haven’t necessarily discovered deposits but they have found what we call vectors towards deposits. That means indications to where to go next. But there’s still a slow process, right? You have to fly to Northern part of Canada, Quebec, and you go only in the summer and there’s helicopters involved and there’s fuel involved and there’s so much involved. It’s not just sitting on your desk and programming in it. There’s a lot of people stuff and moving stuff involved as well.

[00:13:51] Camille Morhardt: Okay. I have so many questions I don’t really even know where to start.  I guess one of the questions is you’re talking about, okay, we need more of these resources, and surely, hopefully there are more on the earth and we just have to find them. But I’m wondering about the other way to solve the problem which could be design batteries differently or synthesize material as opposed to finding raw material.

[00:14:19] Jef Caers: Yeah.

[00:14:20] Camille Morhardt: What’s going on in those areas?

[00:14:21] Jef Caers: Yeah. That’s a good question. There’s always future technology that’s gonna be better. And there are certainly for batteries, there are potential other ways of doing batteries. Lithium seems to be always within the area, but we have to also realize that the CO2 problem has to start to be addressed today. We can’t wait for 10 years in the future to design a better battery. We are emitting CO2 in the atmosphere today at large quantities; we are not leveraging enough solar and wind power and so on. So we have to build that out right now and the best technology right now is the lithium-ion battery.

There will probably be better batteries down the future, talk about solid-state batteries and other types of batteries that require less of these materials. But at the same time, we are going to use copper, we’re gonna use nickel, and we are going to use lithium, maybe we can mitigate somewhat the cobalt; but still we have to do what we can today and then work toward the future to hopefully design things that require less of these very sparsely distributed materials.

[00:15:30] Camille Morhardt: I guess the only other approach I could think of offhand would be having some way to, say the problem is intermittent energy generation when it comes to renewables like solar, wind. I guess the only other thing would be to switch to ones that aren’t intermittent like ocean tides.

[00:15:48] Jef Caers: Yeah.

[00:15:49] Camille Morhardt: Or at least they’re predictable or to switch to an ability to shift power sources depending. So you have multiple inputs coming to one household or something, and I’m gonna use solar, wind, or wave or whatever there is.

[00:16:02] Jef Caers: Yes. There are other types. Wave energy is very expensive. One of the great things about solar and wind is very cheap. In fact, solar is beating almost other types of renewable energy. The other that are less intermittent are hydro power as well as geothermal. So also work on the area of geothermal energy, and there you’re thinking about geothermal is almost constant, right? And if we could drill deep enough to say 20 kilometers deep, which hasn’t been done by the way, then we would’ve access to an infinite source of energy.

[00:16:36] Camille Morhardt: Did you say hasn’t been done or has?

[00:16:38] Jef Caers: No, it hasn’t been done. I think the deepest well goes to about three to four kilometers, but people are working on new technologies for drilling using laser-based drilling rather than rotary bit drilling. And so again, those are things that are probably gonna be here in ten years. But in those ten years, we’re going to emit a lot of CO2 if we don’t do anything. And that’s why I think the current technology needs to be used right away.

Another option for the intermittency is of course, hydrogen, right? Is to use solar and wind to create hydrogen.  Now, hydrogen is created using, for example, electrolysis and that requires platinum and palladium, where you need more metals, right? It’s not free. Hydrogen is still very expensive and again will be something that will probably be in the future maybe in a decade from now. It depends on how fast we develop that. There’s also issues in transportation with hydrogen, but it’s definitely a fuel that’s in the future.

I think if we start picking winners and losers, we may actually lose out on a winner. And so betting on all the horses right now and seeing what develops is the best option. And we are going to have to use energy in various forms. If you’re think about heating your house, that’s not gonna work with batteries. That’s not gonna work with hydrogen. Right? There are the people thinking about geothermal, most people in Europe are starting to look into that, and the United States isn’t as developed yet, but these are also almost three forms of energy that are available to you.

[00:18:05] Camille Morhardt: Talking about betting on the horses, are people in different parts of the world emphasizing different kinds of renewables?

[00:18:14] Jef Caers: Yeah, because it really depends on your geography. United States is a very big, widely spread country, right? So having one energy system for the whole country may not make a lot of sense. Climatic, of course, is also very important. Are you in a solar area in the west of United States where solar is just so available, wind is so available since this is great for us, but if you live in Norway, that would be less so; but Norway has hydro, pumped hydro, right–fjords and things like that. We have to develop, I think, strategies that work for different areas of the world and that would be different.

[00:18:49] Camille Morhardt: Do you think we’ll be moving to more distributed grids like individual grids?

[00:18:55] Jef Caers: I’m not too familiar with the electricity grid. I do know from colleagues that our electricity grid is woefully inadequate at this point to deal with this renewable revolution, simply because we’re going to rely much, much more on electricity for our energy consumption. And so that requires building a very different grid, smarter grid and things like that. So that’s what I… I’m not expert at that, but I know it’s inadequate.

[00:19:22] Camille Morhardt: You mentioned some of your colleagues are looking at the moon and Mars, what are they looking for? Do they have a specific target in mind?

[00:19:29] Jef Caers: Yeah. Of course, we’re looking whether there was life, right? It’s also an exploration problem. And I sometimes talk with my colleagues and just say, “Hey, I hear you’re working on mineral exploration using AI, what about Mars life exploration using AI?” Why not, right? So, yes, for example, I’ve had one student in my class once said, “well, I have this rover on the moon and that rovers can take three samples to discover water, right? Ice.”  And so knowing where to go and where to take the samples and what order to take the samples and what data to use in order for you to better take the next move, again, it’s a problem where probably you have to drill a lot of times in order to find the water or evidence of life. Again, I think AI is a natural solution there to improve the exploration endeavor.

[00:20:20] Camille Morhardt: Very interesting. In that case, you would be telling the AI that you want to find evidence of water.

[00:20:26] Jef Caers: Right, exactly.

[00:20:27] Camille Morhardt: Not life, because we’ve got a whole bunch of negatives on that data labeling for looking for life. [laughter]

[00:20:33] Jef Caers: Yeah. [chuckle] There’s no data, there’s no labels yet, so it’s an unsupervised problem in a way, right? Is that, again, it’s a problem of what data to collect and there’s a lot of data already, but it’s remote sensing data is from circling around the moon and having taken the temperature of the moon, and of course, you would look in a crater where there’s a shadow, that’s where you look for water.

But again, it’s the same like in mineral resources, we have geo-physical data, which is also data acquired by flying around the planet, but we don’t have any subsurface data, right? When you do exploration, the problem is there’s no labels, there’s only unsupervised indirect information.

[00:21:08] Camille Morhardt: Mm-hmm. What else do we need to know? What else is important in your field that people are arguing over, talking about, or worrying about?

[00:21:17] Jef Caers: I think of certain point is to start thinking about the human aspect and impact of what this  mining a revolution or renewable revolution is going to be. Although people talk about green energy, most energy does require materials. Solar, wind requires steel and all kind of other materials, copper. And so those materials are going to be excavated in certain parts of the planet. And I think right now the process of how to include the communities that would live around these future mines, that process almost doesn’t exist. And we’re looking specifically in the continent of Greenland, because in Greenland ice is melting, it’s exposing a lot of interesting outcrops for rare earth’s settlements, and so nickel, cobalt, copper. But of course, people live there. A nd even though it’s only 50,000 people, we can’t just say, “Hey, move and we’ll come and dig up and destroy the whole continent.” That’s not what we wanna do either.

One of the things that I want to work on and it needs to be worked on is how you can start engaging the communities that live there into the development of any mines or deciding, well, maybe this is not a good area to be mining for these and these reasons. Right now, those communities are often largely ignored or what would happen is that mining companies would visit them and they will have an info session, a 45-minute info session and then walk away and say, “yeah, we did our work info session, right?” That’s something I want to change too, is that one of my group, we’re starting to develop with anthropologists, social anthropologists to develop ways of communicating, developing strategies for better including communities in the development process. Because once something is found, the pressure to mine is very hard, right?

[00:23:08] Camille Morhardt: Mm-hmm.

[00:23:11] Jef Caers: Because, okay, it takes a lot of money to find it and then now we have this potential $10 billion profit lying in front of us. There’s a lot of pressure to do that. So we wanna also look at, “hey, if you’re going to explore in these and those areas like Greenland, what does that imply for the population if you would be starting mining over there?” That may actually help exploration companies in thinking about how and where they would be exploring.

Because we don’t wanna build our new economies on the back of native communities or other disadvantaged communities. And that’s something that needs to be addressed, as well. And that’s something that actually in the School of Sustainability is something that we wanna build more and more in that this is environmental justice, these environmental issues to put that in with the science as well. It’s not just about the science and the AI, it’s also about the human impact and how we do understand that.

[00:24:05] Camille Morhardt: Wow. Thank you Jef Caers, who is Professor of Earth and Planetary Sciences in Stanford University’s new Doerr School of Sustainability. Thank you for talking with us today.[00:24:18] Jef Caers: Pleasure to be here.

More From