Solving for climate: Coasts in the machine
The Earth’s oceans play a crucial role in regulating the planet’s climate by absorbing and storing vast amounts of heat and carbon dioxide from the atmosphere. However, due to human activities such as the burning of fossil fuels and deforestation, the oceans are warming at an alarming rate. This increase in ocean temperature is causing a range of devastating impacts, from more frequent and severe storms to rising sea levels and bleached coral reefs.
To better understand the complex interplay between ocean warming and its impact on land, scientists like Maike Sonnewald are using climate models to simulate and predict changes in ocean currents and temperatures. These models allow researchers to analyze how different factors, such as greenhouse gas emissions, ocean circulation patterns, and solar radiation, influence ocean warming and its effects on the Earth’s climate system. Through these efforts, scientists hope to develop strategies to mitigate the harmful consequences of ocean warming and protect vulnerable communities and ecosystems.
This episode was produced by Anupama Chandrasekaran, and mixed by Collin Warren. Artwork by Jace Steiner. Find a companion to this episode from our friends over at Carry the Two.
And be sure to check out some of Maike’s recent work!
- The Southern Ocean supergyre: a unifying dynamical framework identified by machine learning.
- Bridging theory, simulation, and observations of the global ocean using Machine Learning
- Revealing the impact of global warming on climate modes using transparent machine learning.
Transcript
Shane Hanlon: 00:00 Hi, Vicky.
Vicky Thompson: 00:01 Hi, Shane.
Shane Hanlon: 00:02 Before we even get into this, are you a movie person?
Vicky Thompson: 00:04 Yeah.
Shane Hanlon: 00:05 I know we’ve talked about your favorite movie of all time, but do you-
Vicky Thompson: 00:09 Yeah.
Shane Hanlon: 00:09 Okay.
Vicky Thompson: 00:09 I like movies.
Shane Hanlon: 00:11 Okay. What’s your favorite doomsday or disaster movie?
Vicky Thompson: 00:15 Oh, I don’t like those movies.
Shane Hanlon: 00:17 Oh.
Vicky Thompson: 00:18 No, I don’t-
Shane Hanlon: 00:19 Well, then this is got a real short conversation.
Vicky Thompson: 00:22 No. Okay, so I tried to watch Bird Box.
Shane Hanlon: 00:25 Oh.
Vicky Thompson: 00:25 That counts, right? And I just cried for the whole first five minutes and then insisted that we turn it off and basically had to go calm down.
Shane Hanlon: 00:36 That’s like a horror movie. I don’t even think that’s-
Vicky Thompson: 00:39 It’s like a contagion, isn’t it? Or is it a… I only watched it for five minutes.
Shane Hanlon: 00:44 Oh my goodness. You went much darker with this than I was going, okay.
Vicky Thompson: 00:51 Yeah. I think-
Shane Hanlon: 00:52 Wow, Bird Box. I was thinking along the scale of Armageddon or, did you hear the movie Moonfall that came out last year?
Vicky Thompson: 01:08 Yes, but I didn’t watch it.
Shane Hanlon: 01:09 It’s from, I think the same person who did Independence Day.
Vicky Thompson: 01:14 Okay.
Shane Hanlon: 01:16 Personal opinion, not as good.
Vicky Thompson: 01:18 Oh, sure.
Shane Hanlon: 01:18 The moon is going to crash into the earth and you need to do something about that. That’s the scale I’m talking about. I’m not just super sadding on you.
Vicky Thompson: 01:32 Okay. So I do better with disaster movies where it’s potentially something that’s just completely outside of my control.
Shane Hanlon: 01:32 I see.
Vicky Thompson: 01:41 I can’t do a disaster movie where it’s like, I have to fight for my life. I feel like, yeah.
Shane Hanlon: 01:45 I feel like later this year, we’re probably going to be talking about zombie apocalypses and you’re going to be the one just like, just take me, I’m done. I don’t want to deal with it.
Vicky Thompson: 01:54 Yeah, take me. Yeah, I’ll put some barbecue sauce on myself. Just make me…
Shane Hanlon: 02:07 No, that’s it. You’re just going to get eaten by the zombies.
02:17 Science is fascinating, but don’t just take my word for it. Join us as we hear stories from scientists for everyone. I’m Shane Hanlon.
Vicky Thompson: 02:17 And I’m Vicky Thompson.
Shane Hanlon: 02:28 And this is Third Pod from the Sun. All right, so we’re talking Doomsday or Disaster, well in this case movies, but I guess just overall. Because, Vicky, we have a really light episode today.
Vicky Thompson: 02:44 Really light?
Shane Hanlon: 02:45 Really light. Just like, easygoing topic.
Vicky Thompson: 02:54 Oh, that’s a trap.
Shane Hanlon: 02:54 It’s a trap.
Vicky Thompson: 02:54 You’re trapping us.
Shane Hanlon: 02:54 It’s a trap. Well, I mean, we’re talking serious stuff and we’ll get through it in a informative, and hopefully it won’t be too doomsday. But we’re talking ocean warming, frankly, specifically with someone trying to model climate to understand the churning and the oceans and how that impacts the land. And so to hear more, I’m going to bring in producer Anupama Chandrasekaran. Hi, Anupama.
Anupama Chandra…: 03:21 Hello. Hi, Shane. Hi, Vicky.
Vicky Thompson: 03:23 Hi. Anupama, who did you speak with for this episode?
Anupama Chandra…: 03:27 So I spoke to Maike Sonnewald. She’s an oceanographer, but she is also kind of somebody who studied physics and technology, she’s kind of a polyglot, to learn and understand. So she uses all these sciences, oceanography, physics, and technology to basically create climate models of the oceans.
Shane Hanlon: 03:53 So can technology solve our climate woes?
Anupama Chandra…: 03:57 I wish it was that easy, Shane. I really wish, but I don’t know. I don’t if anybody can solve it. I think it’s beyond solving, but let’s not lose hope. I think Mikey could have some answers because she’s basically trying to understand the impact of currents, water temperatures over the long, long term. So we are talking about our great-great-grandchildren here. Now, that isn’t an easy task and it requires a deep, deep understanding of oceanography, physics, as well as technology.
Vicky Thompson: 04:28 Great. Let’s hear about it.
Maike Sonnewald: 04:34 My name is Maike Sonnewald and I’m an associate research scholar at Princeton University, which means that I do research, but I can teach and have students. I’m also an affiliate assistant professor at the University of Washington in Seattle. I’m also affiliated with the NOAA Geophysical Fluid Dynamics Laboratory. And the NOAA lab is special in the sense that it’s one of those places where we produce a climate model, and I say we in the very general sense in my case because I’m only affiliated with the lab. But it’s one of the really important places in the US where we do things related to climate and understanding the climate and the oceans, as well as long range weather.
Anupama Chandra…: 05:29 Could you talk a little bit about what exactly you do? If you were to explain to somebody what is it exactly that you do, how would you do it?
Maike Sonnewald: 05:38 The main thrust of my work is physics, where what I do is I try to overall understand what’s actually important, what processes are important in the ocean for us to be able to represent the ocean and understand the theoretical foundations of the ocean to be able to both appreciate it and be able to predict it. So if we understand something, that means that we can predict it into the future, which of course in a climate change context is really important.
06:16 What I also do in terms of… So having found these important regions, for example, is build tools and develop methods to be able to predict things using sparse data. We have this paradoxical problem because we have lots of data from the surface. So satellites, for example, send us back these beautiful images, but to actually have data from the subsurface is really expensive and hard because you either have to send a robot or a ship. And that’s one of those grand challenges that I’ve also dedicated a lot of my attention to trying to build, in my case, a machine learning and theory methodology to be able to infer the subsurface. Both initially to help climate model development, but also in general, I’m hoping to port this to also look at long-range weather, so weather on timescales from weeks to months.
Anupama Chandra…: 07:17 You look at nature in association with machine learning, it’s something you mentioned even earlier. What does this mean? What does machine learning mean and how do you actually use it in your work?
Maike Sonnewald: 07:31 I think machine learning is really incredibly good at, is to help us understand and be able to find those underlying patterns that help us understand the system. And this is something where I favor tools from unsupervised learning, and this is just a flavor of machine learning. And there are a number of different ones, ChatGPT, for example, is a deep learning mechanism that’s getting a lot of attention these days. But unsupervised learning is one that is, I think it’s fair to say that it’s harder, but for these contexts it’s also very, very useful in the sense that you can use it to understand patterns. So I don’t need any labels. So for deep learning, for example, you need to have the answers. So you need to say, okay, these are warm regions, that means this is going to happen and vice versa. But with unsupervised learning, you can say, okay, these regions are becoming warm because of these patterns, so you can see the emergence of patterns and then understand them. So that’s one way that I use machine learning.
08:45 The other one is using deep learning. So for example, to look beneath the surface of the ocean using surface fields, what I do is use deep learning. And deep learning is this type of machine learning that, like I mentioned, is getting a lot of attention these days and you can build really beautiful machinery to do a lot of fantastic things. And one of the things that I do with this is blend in a lot of knowledge from geophysical fluid dynamics to try and make sure that the predictions that are made are actually based and rooted in ocean theory, which gives me a really, really nice leverage in terms of making sure that the predictions that the machine learning algorithm is making are actually rooted in something that is correct. So I can effectively demonstrate that the machine learning is reasoning like a physicist.
Anupama Chandra…: 09:48 And if you were to explain this whole idea of deep learning with the help of examples that you have actually kind of learned, interpreted, understood. Could you use that, the regions that you work in, I think which is around the North Atlantic, to actually explain this idea with a little more clarity for somebody who is maybe listening to this podcast without much knowledge of oceanography?
Maike Sonnewald: 10:20 So the fun thing with the North Atlantic is that we’ve studied it for a long time and we have these wonderful observational arrays, and we think we have a reasonable handle on what’s going on in the North Atlantic compared to other oceans where we have much fewer observations. So some of the things we know is that we have this very strong current that runs around the eastern side of the United States, the Gulf Stream, and the Gulf Stream is one of those really important currents that brings a lot of heat north. What I was able to do with my unsupervised machine learning tools is to take one of those really big complicated climate models or the echo state estimates of something, a little bit different, but it’s the same concept, and find out in terms of what sets the ocean in motion, what kind of makes the Gulf Stream separate from the coast.
11:21 So in that sense, you can think of it as the Gulf Stream sort of holding the handrail of the coast of the Eastern United States, and then at some point kind of letting go and just going across the basin. And it’s that letting go process that I thought was really interesting and worth looking into. So just looking at it, we kind of know that these processes are happening, but actually being able to pinpoint, okay, this is where that transition is happening, that’s harder, and that’s what machine learning can help us do. What I was able to do with deep learning was to teach a machine to say, this is where that important thing is happening.
Vicky Thompson: 12:13 So it’s kind of a relief to know that the human touch isn’t completely dispensable in these models. We still need an actual person to look at what makes sense and what doesn’t.
Anupama Chandra…: 12:25 Yeah. Yeah. I mean, everybody’s talking about AI these days. And this is an important point because these models can be super accurate or completely off the hook, and therefore we really need these domain experts who really understand their space, who understand the science so that they can say, well, this doesn’t make sense, or this makes complete sense and it’s totally right.
Shane Hanlon: 12:51 Yeah, I mean, that tracks. I’m glad to hear that that’s kind of how it’s working.
Anupama Chandra…: 12:51 Yeah, it’s a relief.
Shane Hanlon: 12:59 Yeah, and it’s so interesting too. I wonder, how did she get into this field?
Anupama Chandra…: 13:06 Yeah, I mean, with a lot of things it’s also about what your childhood has been about. And for her it’s been a bit like that. She grew up in Norway, she lived by the sea, and that’s really one big part of the answer.
Maike Sonnewald: 13:21 I mean, I’ve always been interested in the ocean. It’s been a constant in my life when a lot of things have changed. As a student, I knew I wanted to do something about the ocean, but it took me a while to realize that oceanography was actually a field. On the other hand, I realized how important computers and computer science was to addressing these problems. And so I decided to do a second master’s in complex system simulation, which was great fun. It was to some extent jumping in the deep end for me, but I learned machine learning math for complex systems, so chaos theory for example, which I think is also one of those really interesting, but also very kind of topics that are hard to mix together, in terms of oceanography and chaos theory. So yeah, I guess I decided to do two masters and then jump into a PhD, and now I’m still around in science. So either something’s going well or I’m lost and confused still.
Anupama Chandra…: 14:26 Why is it important to study climate change via the ocean? Why is it important? Why is it urgent, I should say?
Maike Sonnewald: 14:35 So I mean, as an oceanographer, I guess I would say that the ocean is really one of those key elements of the climate system that has been hard to study because we can’t see it. The atmosphere is there, for example, but the ocean and a lot of systems in the ocean are a little bit less accessible. So really incrementally in the, if I want to call it the field of climate science, the more we’ve been able to add the ocean to the overall system understanding, the more we’ve understood how important coupling the system is. And when I say coupling, I mean having the ocean, for want of a better word, talk to the atmosphere. So if the ocean under here is warm and the atmosphere is cold, like what happens when you get the arctic air coming down and causing a ruckus on the east coast of the United States, is that you have the ocean being able to give heat to the atmosphere when the atmosphere is colder. And vice versa, if the atmosphere is warm it can give heat to the ocean.
15:37 And how this conversation is happening is really important, both on shorter time scales, so weeks to months for example, but also on longer time scales, so climate, decadal, centennial timescales. And it’s been recognized for a long time that the ocean on centennial timescales is very important because it’s taken up over 90% of the heat that we’ve as humans anthropogenically added to the system. In the coming hopefully few years, maybe longer, I’m hoping that we can really start to address some of those big hard problems, like predicting the weather out longer than we can now. Because of course, if you’re a farmer or if you really want to build resilience, knowing more and having better predictions is really important.
Anupama Chandra…: 16:36 So how is climate modeling actually going to help? And what are some of the challenges? You’ve spoken of some of them with regards to the ocean being kind of a relatively unknown space and you’ve spoken about generally the difficulty. But could you be a little more specific with telling us about how climate modeling is going to help and also about the specific challenges that you face as somebody who’s in the field?
Maike Sonnewald: 17:11 What you’re doing in the model or in these types of models is that you’re representing many parts of the earth system. So the ocean for example, and the atmosphere and land and biology on land and in the ocean, not to mention chemistry and the cryosphere, so ice. And this is all very, very complicated. They all talk to each other, like I mentioned, the ocean talks to the atmosphere. And making sure all of this works, one of the challenges is that you have to do what’s called discretization and that effectively means that you have to, if you’re taking the ocean, for example, if you can imagine a globe, what you have to do is sort of chop it up into little bits. So you can think of it as if you have a photograph of a face, what you have to do is you have to pixelate it because numerically it’s just kind of what you have to do to solve these equations.
18:10 And that in itself is a challenge because you can’t see everything. So you can imagine that if you take my face and pixelate it, if you only have 5 pixels to work with, you wouldn’t really be able to see that I’m wearing glasses, for example. But if you were to use, I don’t know, 500 or 5,000 pixels, you could start to see details like, I have freckles. And those types of things are very similar to what’s happening in the ocean, apart from the fact that in the ocean, unfortunately, all of the things matter. So all the things you can’t see matter a lot because of the way the equations work. So if you have small parts of physics that you’re not representing, that’s going to impact everything.
19:07 So in the ocean, for example, one thing that we talk about a lot is eddies or ocean eddies, and they’re like ocean weather systems. You can think of a tropical storm, for example, or a tropical cyclone. The equivalent also exists in the ocean, but it’s on a much, much smaller scale. But you can imagine what one of these things does, it stirs up a lot of water and just does a lot of important stuff, and there’s a lot of really fascinating physics associated with it and we need to capture that, especially on longer time scales. So if we don’t include what’s called parameterizations for these types of processes, our models don’t necessarily do the things we need them to do or they don’t do them for the right reasons.
Anupama Chandra…: 20:06 So how do you create a climate model with all these challenges that is both interpretable and explainable? And what are you seeing? I’m scared to hear that answer really, actually.
Maike Sonnewald: 20:20 Well, there’s a lot to unpack there. I’m going to try. But I guess the short answer to your question, how are we doing this, which is how I interpret the first part of your question, is that we have the equations, which is great, and they’re beautiful. They’re also terribly hard, but we have them. And that’s just at the base of a lot of the progress we’ve made and a lot of the progress that I’m excited for in the future. So I mentioned the Navier Stokes equations, and they’re the equations that describe how fluid moves. And then there’s also what’s called the Equation of State, and that’s an equation that describes how heavy seawater is. So seawater is salty, and it also has a lot of other components to it. And if you can think about some water that has a lot of salt in it, then that’s going to be very dense or heavy.
21:15 So imagine floating in the Dead Sea, for example. There’s pictures of people reading the newspaper just sort of lying on the surface. That actually works, I’ve been there. But dense water is very heavy, while water with less salt is lighter. And there’s an equation, the Equation of State, that is also very complicated that tells us how this is happening. So for example, we talked about the North Atlantic Ocean just now and there, one of the really important processes that happens is that warm water is brought north, warm water is also lighter. It’s brought north, made cold, so denser, but it’s also made denser because salt is being inserted because you’re freezing. So if you can think of ice cubes, ice cubes are very rarely salty. And this is because when you freeze water, the salt sort of rains out effectively. So in that sense, there are a lot of processes there.
22:16 So I talked about the ocean moving heat north and also the brine rejection, which is what happens when ice freezes. And all of these things I can talk to you about, but I also know from my physical intuition, I know from my geophysical fluid dynamics background that these things are going to happen. And indeed I can code them in a model, it doesn’t even have to be a numerical model, it can be an analytical model, and see that these things pop out. And similarly then when I code up the model with my parameterizations and just trying to make up for the things that I know should be there, but I can’t actually resolve. In that sense, having the equations really, and having an understanding and an intuition of the system is one of those things that really helps the community to make progress.
Anupama Chandra…: 23:20 Well, I just wanted to get a sense of something that you have interpreted and explained using these climate models that have told you something about how climate is changing or what’s happening through what you’ve learned from the ocean.
Maike Sonnewald: 23:42 Yeah, I think that’s also a nice segue into the more technical machine learning side. But what I mean when I say interpretable and explainable, I use the terms with a little bit of nuance, I guess. Where for me, interpretable means that you, for example, are working on a closed budget, so the barotropic vorticity budget for example. And you thus, if you apply unsupervised learning, know everything that’s in there and you can have very good tractability of what the machine learning methodology actually did. So in that sense, I can interpret the outcome perfectly. I know everything that went in there. There’s no black box.
24:31 With supervised learning, for example, I call opening the hood there explainability because that’s something that I do after I’ve trained the model. So when you train a neural network model, for example, what you’re doing is you’re giving it the data, so the answers and the labels. And for me, one of the things that I’ve focused on and how I’ve cherrypicked the parts of my background to try and create something that I see as useful is to hardcode the ability to see whether or not the neural network reason like a scientist or reason like someone who knew something about GFD. And that’s something that I could do because I was rooted in this interpretable component of understanding what should be happening.
25:26 So for example, when the neural network is predicting that it’s letting go of the handrail, that the Gulf Stream is letting go of the handrail, is the coast actually where it’s supposed to be? If I look at the gradient of the coastline, is that important? If that’s not important, then that should be a flag for me that the neural network is probably not predicting this for the right reasons. It’s looking at the tree in the background, which might be something completely random that I had no idea was present in all of that data I gave it. Because just fundamentally what a neural network does is it explores this covariant space or this really complicated space given by all the data that I’m feeding it.
Shane Hanlon: 26:19 Vicky, what’s the most complicated work you’ve done using a computer?
Vicky Thompson: 26:25 Oh my gosh, I don’t know if some people would classify this as complicated, but hundreds of thousands of lines in Excel that just freezes everything up and complicated formulas. But I feel like at this point, that’s really it. Doing things to just completely freeze up my computer.
Shane Hanlon: 26:49 Yeah. I used to do, in my previous research life, I did a lot of statistics work. And I was an expert enough to competently do what I needed to do, but that knowledge is completely gone these days. And I look at this stuff once in a while, I go back to old studies and I think, oh gosh, I have no idea how I even knew that.
Vicky Thompson: 26:49 Right.
Shane Hanlon: 27:12 But I’m wondering this because we’re talking about modeling, machine learning, that type of thing. But Maike doesn’t spend all of her time behind a computer, right?
Anupama Chandra…: 27:26 No. No, she doesn’t. I mean, I think she has a pretty cool job. She does have an adventurous job and she’s really gone on quite a few odysseys, so to say.
Maike Sonnewald: 27:38 When you mentioned going to sea, I don’t do that as much as I would like, but I have been on a cruise in the Arctic, which was amazing. I really had a great time. But there I discovered that I don’t become seasick, I become landsick.
Anupama Chandra…: 27:58 What does that even mean?
Maike Sonnewald: 28:00 So when I go from land to being on the boat, I’m totally fine. I adjust and I don’t get seasick. Well, when I go from being on the ship to on land, it’s a different story. So for a number of days, just closing my eyes was just really disorienting, and if I was trying to read something, the world would move. So that was somewhat disorienting. In terms of other moments, I guess more in terms of the mathematical side of things, realizing the power of ensembling and using GFD to rationalize or reason within a neural network was definitely one of those moments where I was like, wow, interdisciplinarity is really, really cool. Geophysical fluid dynamics. So it’s fluid dynamics, the Navier Stokes equations and whatnot, but on a really big scale. So geophysical, so geo being for me the earth, so it’s fluid dynamics on very, very big scales.
Anupama Chandra…: 29:12 How does climate change, global warming look to you as you are kind of sifting through some of this data, which is increasingly getting sharper and sharper?
Maike Sonnewald: 29:25 Well, I think some things have really been fascinating to me where we are able to get more data and put the data together in a way to give a more full representation of the system or the ocean system. And some of the really interesting things for me was just how variable everything is, and me included. When I started studying, I thought of the ocean as being something that moved fairly slowly. But even with the Gulf Stream example that I had just to return to something that we talked about before, that’s part of this really big system of currents where the surface is warming, goes up north, and then it becomes dense and sinks. And this was something that I thought of as a fairly slow system, but it turns out it has huge variability. And this is something that by having records of it, we were able to discover. And that to me was just incredible. It blew my mind.
30:36 And similarly also with other things, although potentially less, I mean, I wouldn’t say the variability of the Atlantic overturning circulation is necessarily positive. But in terms of other things that really surprised me was the way, for example, the ocean around Antarctica can move and how variable that is, and also how much it’s warming. Because for some areas of the ocean, the impacts of climate change are something that we see more quickly, where the higher latitudes are really some of those sensitive areas where, for example, the ocean talks very deeply to the atmosphere. So it’s one of those areas where you get a lot of exchange of say carbon, or also impact on the biochemistry. And there, getting data is really hard because a lot of the time it’s covered by ice around Antarctica, for example. And being able to have new technology, so robots, for example, that can be under the ice and collect data all year round, has shown us that these areas are changing more so than we might necessarily realize or be able to see, which is definitely a cause for concern.
Vicky Thompson: 32:04 How I love it when we end on a high note.
Anupama Chandra…: 32:07 Yeah, I mean this one’s a real high. We can’t really go higher than this. I really wish that things looked brighter, but still, there’s always a balance when you’re talking about anything associated with climate change, hopefully.
Shane Hanlon: 32:25 Yeah. And frankly, the data could be more complete and we definitely don’t want to misrepresent. But the upswing here is that there are folks out there like Maike doing great work to learn more and more and hopefully influence some positive change.
Vicky Thompson: 32:41 And teach us about land sickness.
Anupama Chandra…: 32:43 Land sickness, yeah.
Shane Hanlon: 32:43 Oh my gosh, yeah. It was so funny because I was listening to that and I can relate to that. When my wife and I, we did our honeymoon in Galapagos and we went around on a bunch of islands on a boat. It was a decent size boat, but still, it moved. And by the time the trip was over, being on land was actually rougher than being on water.
Anupama Chandra…: 33:11 Really? Really? Wow.
Vicky Thompson: 33:13 Oh. Wow. Well, that’s good. So we can end on the high note of thinking about you being sick.
Shane Hanlon: 33:18 You know what? I am here to help. And so with that, that is all from Third Pod from the Sun.
Vicky Thompson: 33:26 Thanks so much, Anupama, for bringing us this story and to Maike for sharing her work with us.
Shane Hanlon: 33:31 This episode was produced by Anupama, with audio engineering from Colin Warren and artwork by Jace Steiner. And be sure to head over to the Carry the Two podcast next week for more from Maike on the math and science front.
Vicky Thompson: 33:44 We’d love to hear your thoughts on the podcast, so please rate and review us. And you can find new episodes on your favorite podcasting app or at thirdpodfromthesun.com.
Shane Hanlon: 33:52 Thanks all, and we’ll see you next week.
34:01 Cool. All right. Anupama, I appreciate that you work how we do, like what’s on the page isn’t exactly-
Anupama Chandra…: 34:01 I’m learning.
Shane Hanlon: 34:01 … what you read.
Anupama Chandra…: 34:15 The first time I was really bad, so I’m just like, I’m going to learn to laugh, I’m going to learn to add it on. No, yeah, I’m trying. Yeah.
Shane Hanlon: 34:21 Yeah. I mean, for all of it it’s been like this, but the nice thing with this, I mean, you know because you’ve done this a handful of times with us now. If something doesn’t work, we just rerecord it. Like when Vicky and I do stuff, we’ll do something. We did one yesterday or the day before and read something.
Vicky Thompson: 34:37 We just looked at each other like-
Shane Hanlon: 34:38 It was like, oh no, that doesn’t sound right.
Vicky Thompson: 34:38 That doesn’t work.
Vicky Thompson: 34:40 Right.
Shane Hanlon: 34:42 We’ll just do it again.