Rapid advances in artificial intelligence (AI) are helping the energy industry accelerate the transition to a low-carbon future. The Energy Podcast explores how AI is being used today and discusses how to unleash its potential.
Presented by Eno Alfred-Adeogun. Featuring Kate Kallot, founder and CEO of Amini, Bob Flint, CEO of Mirico and Amy Challen, Shell’s global head of AI.
Additional reporting by Claire François and Berry Mulder.
The Energy Podcast is a Fresh Air Production for Shell, produced by Annie Day and Sarah Moore and edited by Eno Alfred-Adeogun.
00:00:00
Eno Alfred-Adeogun: Today on The Energy Podcast.
00:00:07
Audio: I will be working alongside humans to provide assistance and support and will not be replacing any existing jobs.
You sure about that, Grace?
Yes, I am sure.
00:00:20
Eno Alfred-Adeogun: That’s Nurse Grace speaking at the world's first robot press conference last year. And yes, she’s a robot. Powered by artificial intelligence this humanoid can diagnose illness, deliver treatments, and even offer patients emotional support. Impressive, right? Well, yes, but she’s just one of many examples of AI-enabled machines designed to address some of the world's biggest challenges; social care, disease, hunger, and probably sooner than you think. Consider how deeply AI is already entwined in so much of our daily lives. From work commutes …
00:01:03
Audio: You’ve arrived at your destination.
00:01:04
Eno Alfred-Adeogun: ... to virtual learning …
00:01:07
Audio: (foreign language).
00:01:07
Eno Alfred-Adeogun: ... to, " Alexa, what’s on my to- do list?"
00:01:11
Audio: Subscribe to The Energy Podcast.
00:01:14
Eno Alfred-Adeogun: A global AI revolution isn’t coming, it’s already here. So, could this rapidly advancing technology also tackle the pressing challenge of lowering emissions? Hello, I’m Eno Alfred-Adeogun, and today on The Energy Podcast we ask, can AI get the world to net- zero faster?
Joining me to discuss this is Kate Kallot, founder and CEO of the African tech startup Amini. Bob Flint, CEO of methane emission monitoring company Mirico. And Shell's global head of artificial intelligence, Amy Challen. It’s really great to have you all on the episode today.
Now, before we delve into the world of AI, a really helpful place to begin is defining what it actually is. Because by the number of definitions I found when researching this episode, that's actually harder to do than it sounds. So let’s briefly see if we can reach a consensus of what it actually is. Kate, coming to you first.
00:02:20
Kate Kallot: For me, I have one simple definition of AI, which is going to literally take one sentence. It is the science to make computers think and take actions like humans.
00:02:30
Eno Alfred-Adeogun: Love it. Brilliant. Brilliant. Amy, what about you? Can you add to that?
00:02:34
Amy Challen: I think I’m going to give a more boring answer. I often think about it compared to software. In software, we write the rules. We say, " If this happens, then that happens," and we define what that rule is. But AI works differently. We give AI a load of historical data, and we say, " You tell us what happens based on the patterns you've observed in the past." And so it can be a bit of surprise what it comes up with. The other thing to watch out for is that if our historical data is biased, if the world has changed, then we're going to see that in the model. So we have to be quite careful.
00:03:10
Eno Alfred-Adeogun: Okay. Bob, no pressure. We’ve got two great answers. Do you have anything to add onto the definition?
00:03:16
Bob Flint: Yeah, I get to go last. I think all of the above, plus extending into areas where humans aren't necessarily good, which is looking at huge volumes of data. So being able to process all the bits and bytes that come from sensors from the real world and floods of information like you would find in an oil and gas company and process that in super high speed.
00:03:40
Eno Alfred-Adeogun: Okay, so now that’s clear, we can turn our attention to how the energy sector is actually harnessing the power of AI. So, Bob, your company, Mirico, it monitors emissions, and then companies can pair the data that you gather with AI, which can then help to combat the emissions found. Can you share some good examples of this pairing in action?
00:04:06
Bob Flint: What we do is primarily we address the issue of methane emissions from energy. And methane is about 30 times worse than carbon dioxide for global warming, so it’s something that we all should be concerned about. The good news is if you stop emitting methane, the world starts to cool pretty much immediately because methane just decomposes in the atmosphere. So that’s why it’s so important. So we scan an area, say, an oil and gas facility, for those emissions. Measurement leads to action, leads to reduction, and we can then put that data into one of Shell's systems of record, say a digital twin.
00:04:48
Eno Alfred-Adeogun: On the digital twin, because that's another concept I feel like I've read a lot about as well. Could you just expand on what that is?
00:04:54
Bob Flint: A digital twin would be a replica of something in the real world, a refinery, a rig and it sits in silico, in a computer rather than in the physical world. So that would enable you to, for example, start making predictions or what-if type of questions on what might happen in the real world in circumstances which you wouldn't want to see. So you might predict what happens if I change temperatures and pressures in a way that might be dangerous. Well, I can do that in a computer. I wouldn’t want to do that in the real world.
00:05:30
Eno Alfred-Adeogun: Now, considering the ambitious climate targets that have been set worldwide, it's really no wonder that people are looking to AI for solutions. Amy, what would you say are the main ways that AI is enabling a low-carbon energy system?
00:05:47
Amy Challen: I think there are four main ways it’s already doing this and is going to continue in the future. Firstly, it can improve the efficiency of everything we're already doing. Whether that’s renewables or hydrocarbons, a couple of percentage points of efficiency can make a difference to the productivity of our energy system and can reduce CO2 emissions and methane emissions. The second one is I don't think we're going to enable a widespread renewable system without having really excellent forecasting and optimisation, two things that AI is incredibly good at. Because renewables like wind and solar, they're intermittent. Sometimes the wind blows, sometimes the sun shines, sometimes it doesn't, and it's not entirely predictable. So, you use AI to predict when that will happen and to optimise how you use your battery, how you buy and sell energy to make sure that we always have the energy supply when we need it.
The third way is through everything we can do in lab sciences, in research and development there, and in design of a new energy components and systems. So for example, with the lab sciences, you have a whole process of research you go through. You look at papers and patents to have an idea of what to research on. AI can really speed up your search there and find the right things to be looking at to make suggestions. And finally, it’s got a big role to play in monitoring. If we don’t know what the current situation is in terms of working out where methane emissions is, then we don't have a baseline, and we can't track. And that’s hugely important to be able to know what's effective, what's not, and where are we going.
00:07:25
Eno Alfred-Adeogun: Kate, you’re based in Africa, so it'd be really, really good if you could paint a picture of the role AI is playing on the continent, if at all, in different energy systems.
00:07:37
Kate Kallot: Africa is still the most data- scarce continent. We think about the data scarcity affecting many different places, whether it's having billions of people that are still unconnected, whether it's having data scarcity when it comes to satellite imagery, when it comes to meteorological data. So when we think about the transformative role of AI in Africa a little bit more broadly, there's still step one that needs to be fixed, which is the data scarcity.
Now, when it comes to where AI can be applied, it's important to remind everyone that AI is just a tool, and it's not going to come and transform everything and make everything better – it’s a means to an end. It’s not actually the end. So when we think about the solutions that AI will enable to build, we always have to go back to the foundation, which is what problem we're trying to solve.
00:08:28
Amy Challen: And could I build on something Kate just said as well, which is there's no AI without data, absolutely foundational. But also, AI and data, they're not the answer to the energy transition. They’re both mega-enablers of it, but fundamentally, you have to have the physical and chemical technologies and a huge change in the way we work and the way we live in order to make that happen.
00:08:50
Eno Alfred-Adeogun: There’s an old adage, but don't ask me to quote where this came from. But I bet you might have heard of this, that you can't manage what you can't measure. So, with that in mind, Bob, how important is it to not only track greenhouse gas emissions, but to do so accurately?
00:09:11
Bob Flint: As you say, you stand no chance of understanding what to do if you don't know what the baseline is. And actually, maybe people think emissions often come from leaks, a flange that isn't quite tightened properly. In many cases, emissions like methane just come from the way that an asset is being managed. And actually, you can start to manage the asset in a different way if you know why emissions are happening. You can start to maybe change the way in which it's loaded and unloaded or even cleaned. And then the last piece where AI could really help in emissions is helping to optimise. So in that digital world, if you have got a digital twin and you're able to run some sophisticated algorithms, you can start to say, " Well, I'm not going to tolerate last year's emissions. This year, I’m going to make less. And I’m going to set myself a target, and I'm going to meet or exceed that target.
00:10:04
Eno Alfred-Adeogun: Now, with the speed at which technology is advancing, the potential solutions that will arise through AI are vast. And when it comes to the energy sector, we're already seeing that potential turned into reality. Take for instance the drone parked in a box at Shell's Energy and Chemicals Park Rotterdam in the Netherlands, where crude oil is processed. Claire François is with Shell’s robotics expert, Berry Mulder, to find out how the drone uses AI to monitor the site and detect leaks.
00:10:36
Berry Mulder: So, there’s actually no one on site to fly the drones. They’re sitting remotely, and the drones go around a few times a week to scan for emissions, to check on rooftops, and all the other things that operators typically go around for.
00:10:53
Claire François: Well, we’ve just seen the drone fly off. Can you tell us what it's doing now?
00:10:56
Berry Mulder: Looking at the tanks from above, so basically saving us many hours a day to, well, not walk up to the tanks every time. It just brings us the pictures. So, the people, the operations, they look at the pictures as if they would walk around and then interpret what needs to be done to the top of the tanks.
00:11:13
Claire François: And so how is AI part of this process?
00:11:17
Berry Mulder: Right now, people are still looking at the pictures because it's a learning process for all of it. We expect that AI machine vision analytics will help us to analyse our pictures faster and better in the near term, but just spotting the differences, the anomalies, corrosion or things that broke or emissions leaks similar to what we do with the ground robots right now.
00:11:38
Claire François: Can you tell us how a drone like this and your other robots in the field are helping to detect methane emissions or broader emissions?
00:11:46
Berry Mulder: So really having additional sensors on the ground and in the sky to detect leaks while they're still small, and so we can fix them while they're still small and not big. And that’s where having additional noses fly and drive around really helps us if you take it to the AI part. It’s also eventually sort of making correlations. Where are the sources coming from? Where is it building up faster? We expect it will even help us in emergency responses, try to detect the plumes, the concentrations faster than we do today.
00:12:23
Eno Alfred-Adeogun: As Berry just demonstrated, technical solutions to help reduce emissions already exist. But despite that, according to the United Nations, the world is falling short of its climate goals. How can AI be a better companion to existing digital technologies, Bob, to drive progress more quickly?
00:12:44
Bob Flint: Most operating companies would have vast data sets already. AI can play a role in just mining that data for information and then surfacing that so that it can be acted upon. The other thing I just want to point out is that the emissions challenge, and we heard something of that drone solution in that clip, is really about deploying lots of different solutions.
There’s no single silver bullet in that measurement journey. So, you’ll probably have some satellite systems, some drone systems, some ground sensors, maybe even some robots. And bringing all those diverse data sets together is a big challenge, and AI can really help us do that. So, build one picture of what's happening in the real world from a set of very disparate data sets and then using that to drive decision- making. That will be a wonderful application.
00:13:43
Amy Challen: And that’s essentially what a digital twin is at the end of the day. It’s a combination of all these different data sources. So, like you say, Bob, your laser- based detection of methane using machine vision techniques to do that, and then all the other collection of data about, say, the operation of an asset, and you pull that together. And as Berry in the extract just said, you're going from quick detection of leaks, which you can shut down as quickly as possible, to prediction of leaks. So, you’re much less likely to happen because you can fix things in advance, you can work out where there's risk, and you can do preventative maintenance, so it doesn't have to happen. And that’s the aim at the end of the day, is how you can act in advance to reduce emissions rather than just react quickly.
00:14:23
Eno Alfred-Adeogun: Now, we haven't touched on an important group I think that we need to consider in this conversation, and that's energy customers, both businesses and the public. What benefits could AI offer that will encourage them to reduce their own emissions? Bob can I ask you that?
00:14:41
Bob Flint: Yeah, I think whether it's consumers or businesses, actually, energy is unfortunately a topic that's quite hard to get people to engage with. And one of the methods of doing that is price. We all feel energy is something that is visible when we get the bill. But in order to be energy efficient, maybe change our behaviours takes quite a lot of effort, and I think AI can really help us with taking away some of that friction by giving us maybe some price signals, “Here's a discount if you do something." Like maybe in the US, there are demand reduction programs, which are run by businesses for businesses, encouraging people to switch off their refrigerators and air conditioning systems for an hour during peak times. And it could even start to automate some of those interventions, turn things off automatically. So, taking away some of the pain of acting responsibly for people who want to do the right thing but who are busy and need to find the time to do it, I think AI will play a huge role in that.
00:15:48
Eno Alfred-Adeogun: AI may be helping to find solutions to reduce emissions, but we also need to consider the vast amount of energy that's used to develop and run it, which risks increasing the world's emissions. I’m thinking particularly about the huge amount of energy that's used to run data centers. What do you think is the most important action that the tech industry must take to prevent AI from contributing itself to global warming, Kate?
00:16:23
Kate Kallot: The technology has made so much advancements today that you don't actually need to build large scale data centres everywhere to be able to run the amount of compute that you need. There are new ways to do it. They’re decentralised ways, power- efficient ways. The industry is heading towards the right direction, but I think there's still a lot of work to do when it comes to understanding what that really means on the ground.
For example, understanding that most of the compute that you need, you need it for the training. When it comes to deploying AI systems, it doesn't require that amount of compute. So, can you find innovative ways to share the computes between countries, between industry, between companies? Can you find innovative ways to build decentralised data centres or decentralised computing infrastructure where you have smaller amount of compute in different regions that are linked together, and different actors can get access to that compute at different times or different days?
00:17:21
Eno Alfred-Adeogun: Amy, is there any way we can stop AI becoming a monster in itself?
00:17:27
Amy Challen: Absolutely agree with Kate. Even more fundamentally, though, you can also ask yourselves, "Do you need to train this model?" So I think what I see externally is the technology companies in particular, they're in a kind of arms race to get the best generic large language model. Do we need 10 of them or do we just need a couple? Within Shell, certainly, I really challenge everyone, do we need to do this? And to Kate’s point, it's a lot more effort to train a model than just to deploy it. So if you’ve already got a really good core model, which you can just adjust slightly, whether through fine-tuning or prompt engineering, let's do the minimum we have to because that will save a lot of energy.
00:18:06
Eno Alfred-Adeogun: Now, before we close, I'm going to put each of you on the hot seat. I don’t know if you've already been feeling like you were on one, but it's about to get hotter! Because I want each of you in no more than 20 seconds to say what you think the most exciting development in the energy sector you expect or want to see AI transform in the future. Bob, let’s start with you.
00:18:30
Bob Flint: Ooh, okay. Great question. I think the bit we haven't really got to yet is thinking about energy as a system. How do we interconnect the traditional bits of that system, like the fossil parts with the new parts, like renewables and storage and hydrogen, to create a whole new ways of working? I think that’s really one of the areas where AI is going to have to be deployed if we're going to break down some of those barriers. Look at complex new areas like hydrogen, and use them most effectively.
00:19:06
Eno Alfred-Adeogun: Okay. Kate, same question to you.
00:19:09
Kate Kallot: I’m going to be a bit biased because I run a climate tech company, but for me is around exploration. So, making sure that using AI actually to better understand and assess the reserves, but also avoiding the destruction of more biodiversity.
00:19:26
Eno Alfred-Adeogun: And, Amy, what do you think?
00:19:28
Amy Challen: I think I’ll say it’s this use of AI in research and development and in design that I think we're only scratching the surface in how quickly we could speed up the advancement of development of materials, advancement of development of new chemical solutions, and in designs for better, more efficient assets.
00:19:51
Eno Alfred-Adeogun: Kate Kallot, Bob Flint, Amy Challen, thank you all so much for your incredible, insightful, and thought-provoking contributions today. Really great. You’ve been listening to The Energy Podcast brought to you by Shell. Be sure to follow and subscribe for free wherever you get your podcasts, so you don't miss a single episode. The Energy Podcast is a Fresh Air production, and the views you've heard today from individuals not affiliated with Shell are their own and not Shell PLC or its affiliates. I’m Eno Alfred- Adeogun. Thank you for listening. Goodbye.