NARRATOR Welcome to the DNV GL Talks Energy podcast series. Electrification, rise of renewables and new technologies supported by more data and IT systems are transforming the power system. Join us each week as we discuss these changes with guests from around the industry.
MATHIAS STECK Welcome to a new episode of DNV GL Talks Energy. My guest today is Dr Kirk Borne, principal data scientist in the Strategic Innovation Group at Booz Allen Hamilton, former professor of astrophysics and computational science at the George Mason University, as well as former scientist and researcher at NASA. Welcome, Kirk.
KIRK BORNE Thank you, Mathias.
MATHIAS STECK What we want to discuss today, Kirk, is the power of data and how big data is revealing new ways to drive energy efficiency, and therefore reduce the impact of climate change. But before we do this, it would be great if you could introduce yourself and tell us a bit about your interesting career.
KIRK BORNE Thank you, very glad to be here. As you mentioned, I’m working now with Booz Allen Hamilton, which is a technology and management consulting business, and people think what is that? Management consulting, to me, especially in this modern, data-enriched world, is all about three things. Data to action, taking data to get actionable insights, and taking those actions. So, this actually matches up, in my mind, in a continuum with my prior life as an astrophysics researcher, astrophysics professor, research scientist. Some people say, that’s a really big change, going from that to management consulting, but it’s not really, for me. It’s really all about data.
So, I worked for 20 years in the NASA system, with astronomy space data, space mission data, managing large data systems for scientists to use. I fell in love with data because it was my day job and my night job as an astrophysicist, and then I moved from there to teaching data science at the university, as you mentioned. So, I was a professor of astrophysics but I never taught astrophysics, I taught data science, and I did that for 12 years, and this company called me up and asked me to be their principal data scientist, which means I get to now talk about the value of data and analytics across all kinds of different industries and different domains and different applications.
MATHIAS STECK So, you mentioned your work in NASA and the Hubble Space Telescope or Programme had, I think, hundreds or even thousands of research projects around it. Among others, I read also, regarding climate change, not only on our planet but also on others. So this is actually the fantastic thing about data: analysing is giving us deep insights into very complex phenomena. So, I would like to maybe draw on this climate discussion which is very relevant these days. So, what can data actually tell us about climate change, what it will do to us when we get it wrong, and how we can possibly mitigate it?
KIRK BORNE Well, the interesting thing about what you were just saying is the similarity between looking at climate to my prior life, studying the universe,;and another analogy or another metaphor for that is the human body. These are all complex systems, they have many moving parts, many different fields of study around each one of those things. There are many different ways people can look at how the thing operates, what are the causal factors when you see things happen, and it’s all about collecting data.
And so, in the climate change sense, we’re collecting data for many different sensors, those sensors can be satellite images, they can be just in-situ sensors in an environment. For example maybe people have weather sensors or they have moistures sensors in their soil, or even just data that could be considered not normal data, like a report, a report talking about development in a given country. And so, development that is changing how land is used, so, basically removing some of the rainforest to develop that bit of land is changing the dynamic of how the earth can process oxygen, because we no longer have quite the same capacity for processing the carbon dioxide as the rainforest gets switched out for human development.
And so, all these different things, whether they’re reports or satellite images or sensors, are providing input to our understanding of our world and hence, ultimately, predictive models will show us what is likely to happen if we continue doing the things we do; but these models can also say if we tweak this or tweak that, or change this or change that, we can see, oh, we can achieve a different or better outcome if we make some adjustments in the way we’re doing things.
And so, that’s the difference between predictive modelling and prescriptive modelling. Predictive, you just say, here’s what’s going to happen, people; prescriptive says, here’s what’s going to happen if we change something. So, we try to achieve a better outcome, and that’s what prescriptive modelling is about. And so, the only way you can achieve prescriptive modelling with any accuracy is if you have many, many different sources of information, many data points, many types of data, that is you need to know what kinds of things will change the outcome.
Hence, I like to use the analogy, in this case, with the stock market. So, you can invest in a stock and you can hope that it will go up or down, depending on whether you’re shorting the stock or whatever you’re doing. So, if you just looked at the time series of that stock alone, you could make some kind of prediction of its behaviour, but it’s only based on the prior history of that stock, which can be very dangerous because if some conditions change in the world in some economic or social or political way, then that stock can go in some direction you didn’t expect, based upon its own unique time series.
So, what you need is this additional information from all these other sources to say, what are the conditions under which this thing will move in a certain direction, and then try to understand not only those conditions themselves but are some of those conditions things that you can have control over, that you can change? And that’s definitely true in the climate sense, that is: are there conditions that will lead to outcomes we don’t like, and are some of those conditions things which we have control over that we can change to achieve a better outcome?
And again, it comes from having all these different types of sensor data, not just a single time series. If all we had was the temperature of the earth for the last 10,000 years and we predict it’s going up in the next century, that doesn’t give us any information about what we can do to change it, but all these other sensors, which we now call the big data world… I hate to use that expression but basically, for me, when people say big data, I tell them it’s not about the big or the data, it’s about the insight you get from all these different sensors, all these different contextual ways of looking at a complex system.
MATHIAS STECK Actually, on that point about data and how we use it, and I tie that back to this example of climate change again, there are people with very different opinions about the impact of climate change. They base these opinions on some data, but that’s a general problem: how, from a data science point of view, do we ensure, first of all, that the data is correct, that we are looking at the correct data, and that we then arrive at the correct findings out of this data?
KIRK BORNE It’s a complex scientific process that’s taking place here, one of which is the actual collection of the data, and the next thing that we’re tying these observations to are simulations of this complex system. So, we build models of what the earth will look like 10, 50, 100 years from now. So, those simulations, those complex, numerical, high-performance computing models, by themselves produce a lot of data. Just the output from these computer models, which we might just call synthetic data, is enormous in volume. So, scientists are collecting those data, so then we try to combine the two, and that’s called data assimilation.
So, assimilation is marrying the actual observed data to the numerical model, and when you see there’s congruence, you know you’re building the right model; when there’s incongruence, you try to adjust your computer model to take into account this new data information that you have about the environment. And as you assimilate, that is the data becomes part of the process of improving the computer model, then you achieve better and better predictive modelling, and also better prescriptive modelling, that is you see what things can cause certain outcomes and which things can produce better or worse results. And again, as I say, some of those are things that you have control over, then we ought to take charge and do something about them.
And so, there’s a lot of political discussion around this, and I, as a data scientist, prefer to take the objective, scientific approach, what does the data tell me to do, and I’ll let other people have the conversations in the other realm of human discussion which is less scientific; I’m not saying that is good or bad, it’s just not focused on the objective data processes which I am focused on.
MATHIAS STECK Talking about the power of data, and especially also that computational power has become so accessible and affordable, we hear visions about things people call artificial intelligence, decisions being made for us, but I wonder about things like human sense, instinct, judgement or maybe even moral and ethical aspects.
KIRK BORNE Yes, so, this community, working in data science and AI these days, is receiving a lot of attention within their own discipline and now from outside, the world in general looking at what we’re doing, and it’s all around the ethical and moral questions of what we’re doing and how we’re making judgements. Are we empowering machines to make judgments? How do we know the algorithms are not biased? We’ve discovered that they are biased. And so, what most people in this field now like to say is AI is not really artificial intelligence, because there’s nothing artificial about it; we like to say it’s augmented intelligence or assisted intelligence, maybe amplified intelligence or actionable intelligence. S,o all these different ways of thinking about how is it that we’re using this information to be more intelligent?
And so, I like to think of the assisted intelligence as a good model, or working model, or how AI is changing now; and that is it’s a human assisting a machine and a machine assisting a human. One of the ways this happens is something that humans are good at and machines are not so good at, and that’s separating the noise from the signal. So, for example, today we’re sitting in a car having this conversation, in a very pleasant little environment surrounded by birds and trees but also surrounded by other cars and trucks, so if people hear the noise in the background, that’s what they’re hearing. A human is good picking that out and filtering that out, so you and I can have a conversation and I can not pay attention to the bird that’s singing or the car that’s driving by.
But when the machine is analysing the data, all that audio signal is part of the input to an algorithm, and the algorithm has to figure out how to sort out which part of that is the right part of the signal, which part of that is the noise. And so, that’s why you need assisted intelligence; that is a human can assist the machine, say, oh no, this is not part of the signal, that’s part of the background noise. And so, it makes it faster and easier for the machine to do its thing.
And so, AI being assisted intelligence, especially when we’re looking at climate change, where there are all kinds of moral and human judgements that are taking place, we don’t want to empower the machines in those ways; we, as humans, have a hard time empowering ourselves to make all the right moral judgements. We certainly can’t encode something into a machine if we don’t know how to tell ourselves what is the right or wrong thing to do most of the time. So, the assisted intelligence helps us to find the signal in the midst of all the signals and sensors and noise that’s coming at us.
MATHIAS STECK As you know, Kirk, we were at the Global Smart Energy Summit in Dubai, both of us, and I want to also look into that a little bit. Two major themes which we had here at the summit were the integration of large renewal energy systems, as well as energy efficiency. And this is an important topic for utilities and for large consumers, for different reasons, regulations incentives, being sustainable or just having a good business case, but how do you think data can help us improve on the clean energy and energy consumption side, and therefore also help us to reduce emissions?
KIRK BORNE Well, I love the title of the summit: smart energy, because I talk a lot about smart data. And, for me, smart data comes back to something I said earlier, which is when you have many different contextual pieces of information, you can make a smarter decision, as a human. It’s basically being cognitive, that is you’re taking in all the information to make a better decision. So, smart energy is not just about looking at usage and challenging people to do better; it’s giving people other pieces of information that will encourage them to do better.
And so, in the economics world, they call this nudge. The nudge is just the gentle motivation that you give someone to do better, to make a different decision to maybe reduce energy use, to be more efficient, to be more sustainable. And sometimes, that nudge is as simple as just a simple number. So, this happens in my community: we receive a letter, every family receives a letter, how is our energy usage this past month compared to the previous month, and then they compare us against our most energy-efficient neighbours. There are no names attached to this but you just see where you rank with respect to your neighbourhood.
And there’s an incentive: I want to do better next month, I want to get better than my neighbour’s X amount, and s,o people are incentivized through this little piece of information, that’s just like a single number, really – actually, it’s a sequence of number but it’s a very small amount of data – and that data itself is in a sense smart data, because it’s my data compared to other people’s data compared to the surrounding region’s use of energy. So, this data provides information and insight and, therefore, motivation for people to be more efficient.
So, it’s not just about the individual consumer but also about energy usage by large companies or energy production or policy within countries, looking at how things can be improved based upon the numbers, and I think that’s where we get to this place where we can talk about smart energy, it’s decisions based upon good knowledge and good insights from all these different sensors that are in the environment, that are in our systems, and even in our homes as well as our businesses.
MATHIAS STECK So, smart energy, when we look at this in the future, this will more and more also mean that, based on assisted intelligence, we can run large, complex systems, like an electricity system for a city or a district, to optimize an entire system level. Now, having said this, that clearly also means that the way we are organized today, industries are organized, might change. So, what is your view on how will the insights we get from data change the way we are structured as a society, as an economy, or maybe even move the centres of gravity of geopolitical powers?
KIRK BORNE Well, these are big questions. I’m not geopolitically knowledgeable enough to answer those kinds of questions per se, but from a data science perspective, again I’d say that optimization, which you mentioned, is a statistical process – in the mathematical definition of optimization – and it’s, again, based upon looking at data and the functional dependence on data, if you want to say it that way. That is, what causes something to go up or down, and then once you learn the causal factors for something to go up or down, then you ask yourselves the question: am I trying to make this outcome go up or down? Energy usage I want to go down; sustainability I want to go up; efficiency I want to go up; black-outs I want to reduce.
So, depending on what you’re looking at, you want it to go up or down. And so, once you understand what can I do to make it go in the direction that I want it to go, then that’s about optimization, trying to achieve that best performance point, the performance point that you’re trying to achieve. And again, this comes from data, looking at data usage, maybe by time of day or by industry or region. And so, there are things that can be done to move energy usage to more optimal times of day or more optimal times of week or more optimal locations in your region, for whatever reasons.
You may have some motivation to do that, maybe there will be some change in the way we operate as a society, the way we enable or encourage people to use public transportation, for example. All kinds of things we can start doing if we start being more objective about what are the things we’re trying to optimize and what are the things that we can measure that tell us whether we’re performing in an optimal way or not. So, it really comes down to standard business practice. First, you say, what are your goals, and how am I going to measure? What can I measure to see if I’m achieving my goals? And then set about measuring those things.
MATHIAS STECK Kirk, unfortunately we are slowly coming to the end of this episode now already; but relating back and reflecting on what we have just discussed in the past 15, 20 minutes or so, what would you expect would be the bigger advancement in this part of the world?
KIRK BORNE The biggest advancement to me is the integration of all these data sources. I like to talk in analogies, so I’ll use another analogy for this question, and that is this famous cartoon of the blind man feeling the elephant. Maybe you’ve seen this cartoon where there are three to four blind men, they’ve got blinders on, they’re not seeing what they’re looking at. One of them is feeling the leg of the elephant, one is feeling the tail, one is feeling the trunk of the elephant, the other is feeling the body of the elephant, and so they have a completely different description of what this complex thing is that they’re feeling.
And so, until they break down the silos of information, which most businesses and organizations deal with, until you take the blinders off and you can now integrate all these different sources of information about this complex system, you’ll never really understand it and deal with it in the right way. And so, I think this integration of data from many different sources, it’s breaking down the silos, sharing information across not just boundaries within a single organization but across organisations and across communities, and even across countries maybe, we’re not going to really be able to get the best insights on how to manage this complex system which we call earth or climate or whatever you want to call it.
And so, it’s putting back that humanized piece of the work we’re talking about here. We’re talking about we’re cognitively aware and seeing change in our world and we know we need to do something about it, and instead of being frustrated as to what to do, we now have enough sensor information that says, if you do this, you can improve that. It goes back to that optimization discussion. If I can improve efficiency or reduce energy usage, if I want efficiency to up and usage to go down, I can understand what things can I do to move those functions in those optimal directions. And again, it comes from data and it comes from many different sources of data, so break down the silos, take the blinders off, let’s share data, let’s share information, let’s share insights so we can do that other form of AI, which is actionable intelligence, that is we’ve got intelligent enough information to take action. And I think that’s the best AI there is.
MATHIAS STECK Very good. Thank you very much, Kirk, for these invaluable insights, and thank you very much also to the audience for tuning in. That was Dr Kirk Borne, principal data scientist in the Strategic Innovation Group at Booz Allen Hamilton, about the power of data and how big data is revealing new ways to drive energy efficiency and therefore reduce the impact of climate change.
NARRATOR Thank you for listening to this DNV GL Talks Energy podcast. To hear more podcasts in the series, please visit dnvgl.com/talksenergy.