WEBVTT - Reinventing Mining to Power the World 0:00:15.356 --> 0:00:22.516 Pushkin. Digging up metal from out of the ground is 0:00:22.556 --> 0:00:26.116 a business that is literally thousands of years old, but 0:00:26.236 --> 0:00:30.756 mining suddenly has new importance. The energy transition, going from 0:00:30.796 --> 0:00:33.756 fossil fuels to renewable energy is going to take a 0:00:33.836 --> 0:00:38.196 ridiculous amount of metal, metal like copper and lithium. The 0:00:38.316 --> 0:00:40.996 need is so great and so urgent that we're gonna 0:00:40.996 --> 0:00:43.596 have to come up with new ways to find metal 0:00:43.716 --> 0:00:47.116 buried in the earth. And as it happens, a new 0:00:47.236 --> 0:00:50.156 kind of mining company, a mining company you might call 0:00:50.196 --> 0:00:54.556 an AI driven startup, just made the biggest copper discovery 0:00:54.636 --> 0:01:02.196 in over a decade. It's worth tens of billions of dollars. 0:01:04.356 --> 0:01:06.756 I'm Jacob Goldstein, and this is What's Your Problem, the 0:01:06.796 --> 0:01:08.836 show where I talk to people who are trying to 0:01:08.876 --> 0:01:13.236 make technological progress. My guest today is Kurt House, the 0:01:13.316 --> 0:01:17.516 founder and CEO of Cobald Metals. Kurt's problem is this, 0:01:18.276 --> 0:01:22.356 how do you use AI machine learning data science to 0:01:22.436 --> 0:01:26.276 find the metals we need for the energy transition. As 0:01:26.316 --> 0:01:29.076 you'll hear, my conversation with Kurt goes beyond mining and 0:01:29.156 --> 0:01:33.196 AI to cover Kurt's really compelling way of just thinking 0:01:33.196 --> 0:01:37.836 about making decisions in an uncertain world. We started though, 0:01:37.916 --> 0:01:40.636 by talking about how he came up with the idea. 0:01:40.356 --> 0:01:42.516 For his company. 0:01:44.636 --> 0:01:48.916 So if you go back about eight years ago, my 0:01:48.956 --> 0:01:52.756 co founders and I were looking at the trends in 0:01:52.796 --> 0:01:57.116 the energy transition, seeing the electric vehicle and renewable energy 0:01:57.156 --> 0:02:03.916 sort of revolutions coming, and it's quite easy to convince 0:02:03.956 --> 0:02:08.596 yourself that the material requirements for the energy transition will 0:02:08.636 --> 0:02:12.836 be tremendous. The amount of very specific materials that the 0:02:12.876 --> 0:02:18.196 world needs copper, lithium, cobalt, nickel, graphite, others. 0:02:18.636 --> 0:02:22.716 Uh, this is basically stuff to build, like batteries and wires. 0:02:23.116 --> 0:02:23.556 Tony right. 0:02:23.556 --> 0:02:25.996 This is motors electrification. 0:02:26.436 --> 0:02:29.436 Just an electric motor is a bundle of copper wire 0:02:29.556 --> 0:02:34.876 surrounded by surrounding a permanent magnet. Every battery require, every 0:02:34.996 --> 0:02:38.876 mobile battery requires lithium and nickel and cobalt. 0:02:39.316 --> 0:02:40.476 These are all that These are. 0:02:40.516 --> 0:02:43.156 These are the key materials for which which in some cases, 0:02:43.156 --> 0:02:45.116 like the humanity has been using lots of copper for 0:02:45.156 --> 0:02:47.756 a long time, there's a big copper market, but it 0:02:48.316 --> 0:02:51.996 needs to at least double from a very large base lithium. 0:02:52.076 --> 0:02:55.276 Humanity has not been using much lithium for very long 0:02:55.956 --> 0:02:59.596 and now the lithium market needs to grow by well 0:02:59.596 --> 0:03:03.116 more than a factor of ten uh to to UH 0:03:03.236 --> 0:03:09.316 fully electrify just just the transportation sector. So the the 0:03:09.356 --> 0:03:12.996 sort of macro needs were very obvious. So that's observation one. 0:03:13.196 --> 0:03:17.516 Observation two say, okay, well, maybe the incumbents are really 0:03:17.556 --> 0:03:20.636 good at finding new materials, and as prices rise a 0:03:20.676 --> 0:03:22.836 little bit, they'll find new materials and the market will 0:03:22.876 --> 0:03:25.756 just be well supplied. And that turns out to be 0:03:25.796 --> 0:03:28.396 definitely wrong. And it's actually really easy to verify that 0:03:28.436 --> 0:03:33.396 it's wrong because the large, well resourced mining companies basically 0:03:33.436 --> 0:03:37.716 don't even do exploration. Actually, the big mining companies out 0:03:37.756 --> 0:03:40.796 of they spend sixty sixty five billion dollars a year 0:03:40.836 --> 0:03:42.756 on dividends and share buybacks and less than half a 0:03:42.796 --> 0:03:44.356 billion on exploration activities. 0:03:44.556 --> 0:03:47.236 But that half a billion, that's the. 0:03:47.276 --> 0:03:52.116 Deployment of conventional exploration technologies, right, things that would be 0:03:52.276 --> 0:03:57.996 natural to most geologists from the nineteen sixties or seventies, right, almost, 0:03:58.196 --> 0:04:00.356 so you can round to zero how much money they're 0:04:00.396 --> 0:04:03.636 spending on research and development for new techniques and new 0:04:03.636 --> 0:04:05.516 technologies to improve. 0:04:05.196 --> 0:04:06.316 The exploration process. 0:04:06.796 --> 0:04:09.796 So it was basically those sets of observations, those two 0:04:09.796 --> 0:04:10.876 sets of observations. 0:04:11.236 --> 0:04:13.476 We need metals, and nobody's really looking. 0:04:13.516 --> 0:04:15.276 They're not looking for them, but they're certainly not getting 0:04:15.276 --> 0:04:17.636 better at it, they're getting worse at We call this, 0:04:17.956 --> 0:04:21.076 we call that that that that trend and the increasing 0:04:21.156 --> 0:04:25.356 cost of discovery, we call that e Rooms law of mining. 0:04:26.476 --> 0:04:27.196 Moore's law back. 0:04:27.516 --> 0:04:28.516 Good, Yeah, I'm impressed. 0:04:28.596 --> 0:04:33.996 Yeah, they talk about that in biotech as meaning, I'm 0:04:34.156 --> 0:04:38.676 where whereas microchips get cheaper and better every year, mining 0:04:38.756 --> 0:04:41.396 gets worse and slower and more more expensive. 0:04:41.396 --> 0:04:46.076 And specifically exploration, exploit discovery specifically. Okay, yeah, so those 0:04:46.076 --> 0:04:48.316 were the those were the major needs. So then you say, okay, 0:04:48.316 --> 0:04:51.156 what what can we do, how can we do? What 0:04:51.156 --> 0:04:54.836 can we do differently? How can we help? And And 0:04:54.876 --> 0:05:00.996 the answer is that exploration is fundamentally an information problem, 0:05:01.396 --> 0:05:04.196 fundamentally right, if you we know, for deep physical reasons 0:05:04.236 --> 0:05:05.876 which I can explain in a minute, we know there's 0:05:07.236 --> 0:05:12.596 there's gobs and gob of undiscovered uh Rich deposits out there. 0:05:13.036 --> 0:05:15.716 We don't know where they are. So so the gap 0:05:15.756 --> 0:05:18.596 is the knowledge of where they are. Right If if 0:05:19.076 --> 0:05:21.876 God gave you a perfect model of the Earth's crust, 0:05:21.956 --> 0:05:24.396 right the location and form of every atom. 0:05:24.596 --> 0:05:25.636 You'd be a perfect explorer. 0:05:25.676 --> 0:05:28.676 You'd know where all the where all the high grade 0:05:29.036 --> 0:05:30.836 high concentration anomalies were. 0:05:30.956 --> 0:05:32.236 You'd also be a perfect miner. 0:05:32.356 --> 0:05:36.236 The miner's religious vision, right, is the gift from God 0:05:36.276 --> 0:05:37.316 of perfect information. 0:05:37.476 --> 0:05:40.636 Yes, perfect exactly, But it's not that it's a so 0:05:40.676 --> 0:05:41.796 we don't have that. So we have we have a 0:05:41.876 --> 0:05:46.676 huge amount of uncertainty. But the sort of managing the 0:05:46.756 --> 0:05:50.236 data that you have and then making probabilistic inferences on 0:05:50.396 --> 0:05:55.476 that data, uh is fundamentally an information problem. We look 0:05:55.516 --> 0:05:57.996 at it as this is this is kind of a 0:05:58.436 --> 0:06:03.516 perfect tailored application for data science and modern scientific computing. 0:06:04.036 --> 0:06:06.636 It has it's it's it's a little different, it has 0:06:06.676 --> 0:06:09.396 some some sort of unique, really cool attributes to it. 0:06:09.436 --> 0:06:12.516 But it is fundamentally an information problem and fundamentally a 0:06:12.556 --> 0:06:16.556 search problem. And so the thing that could be massively 0:06:16.636 --> 0:06:19.916 different would be a company built from the ground up, 0:06:20.876 --> 0:06:24.156 a sort of a Silicon Valley company built from the 0:06:24.156 --> 0:06:29.636 ground up that combines the best existing knowledge of geoscientists 0:06:29.356 --> 0:06:34.556 with world class data scientists and software engineers coming out 0:06:34.596 --> 0:06:39.276 of the major tech monopolies Google, Apple, Facebook, you name it, 0:06:40.316 --> 0:06:43.116 who have never worked in the metals and mining business before, right. 0:06:43.236 --> 0:06:47.756 So it's fundamentally sort of bringing the tools of data science, 0:06:48.316 --> 0:06:52.036 machine learning AI to bear on geoscience. 0:06:52.076 --> 0:06:54.236 Absolutely, if I'm going to reduce totally totally. 0:06:54.316 --> 0:06:57.716 Yeah, Yeah, it's amazing that nobody got to it before 0:06:57.756 --> 0:07:01.076 you did. There are these giant billion dollar mining companies 0:07:01.076 --> 0:07:02.796 and it was right there for them, but they didn't 0:07:02.796 --> 0:07:05.876 do it. I mean, why didn't somebody do it before you? 0:07:06.476 --> 0:07:08.716 What you will definitely hear is, oh, we use data 0:07:08.716 --> 0:07:11.796 science like we we we use scientist, right, And it's 0:07:11.796 --> 0:07:14.396 like and it's like not totally wrong. But what is 0:07:14.436 --> 0:07:19.676 what is definitely unambiguously different, if not unique to Cobold, 0:07:19.836 --> 0:07:22.196 is that we're we're a full stack explorer. 0:07:22.196 --> 0:07:25.196 We were started, we were started and built. 0:07:26.476 --> 0:07:32.436 On the concept that that applying uh vanguard scientific competing 0:07:32.476 --> 0:07:36.836 techniques to these problems would improve the efficacy and efficiency 0:07:36.836 --> 0:07:38.836 of exploration, right, that that is the goal. So we 0:07:38.916 --> 0:07:43.556 have our technical staff is about sixty percent data scientists 0:07:43.596 --> 0:07:46.556 or software engineers, about forty percent geo scientists, right, so 0:07:46.596 --> 0:07:49.476 we're roughly equal equal numbers across the three disciplines. 0:07:50.156 --> 0:07:53.596 And that's that's completely Uh, that's that's unique. 0:07:53.996 --> 0:07:57.876 Let's let's talk about data, right. I feel like, uh, 0:07:58.316 --> 0:08:01.716 discussions about AI, for me tend to get more interesting 0:08:01.796 --> 0:08:04.436 when we get into data and and it seems like 0:08:04.476 --> 0:08:06.316 that's where a lot of the action is and and 0:08:06.716 --> 0:08:08.796 from what I understand about the story of your company, 0:08:08.876 --> 0:08:12.156 kind of building the data set is a big part 0:08:12.156 --> 0:08:13.716 of the story and a big part of what has 0:08:13.756 --> 0:08:16.996 differentiated you. So you have all these data scientists, what 0:08:17.036 --> 0:08:19.916 they need is data. How do you go about building 0:08:19.916 --> 0:08:22.076 this data set to find these metals? 0:08:22.116 --> 0:08:23.436 Yeah, it's incredibly good question. 0:08:23.916 --> 0:08:28.236 So most of the data we use was collected by 0:08:28.236 --> 0:08:33.236 other people at other times. Humans have been collecting information 0:08:33.396 --> 0:08:35.716 about the physics and the chemistry of the Ear's crust 0:08:35.756 --> 0:08:38.836 for a very very long time. Right, They've been doing 0:08:38.836 --> 0:08:41.116 it for well, I mean, in some sense for millennia, 0:08:41.196 --> 0:08:44.076 but certainly over the last century they've been doing it 0:08:44.196 --> 0:08:49.716 in ever more sophisticated ways, and for reasons I can explain. 0:08:50.076 --> 0:08:52.156 Most almost all of that data is actually in the 0:08:52.196 --> 0:08:58.316 public domain. The problem is it is a utter mess 0:08:58.796 --> 0:09:02.716 It is like an end member hard messy data problem. 0:09:03.076 --> 0:09:08.316 Think of different humans in different decades, speaking different languages 0:09:08.356 --> 0:09:09.956 and different places of the world. 0:09:10.876 --> 0:09:12.316 Collecting different types of. 0:09:12.356 --> 0:09:13.836 Data, and I'll get into the types of data in 0:09:13.876 --> 0:09:18.076 a moment, with different measurement techniques based on the vintage 0:09:18.076 --> 0:09:21.196 of their of the era, and then storing it in 0:09:21.316 --> 0:09:25.476 all manner of storage media, everything from literally hand written 0:09:26.476 --> 0:09:29.596 geologic notes or handwritten drilling notes all the way to 0:09:30.036 --> 0:09:34.436 cloud native data data structures right and everything in between. 0:09:35.196 --> 0:09:37.436 And so it is this incredible mess of data. 0:09:37.556 --> 0:09:41.076 Give me some specific examples. What are specific like did 0:09:41.076 --> 0:09:43.796 you find stuff in a drawer or something sort like, 0:09:43.876 --> 0:09:45.396 give me some specific examples. 0:09:45.556 --> 0:09:50.396 So I'll give you examples of there's geologic libraries archives 0:09:50.756 --> 0:09:59.036 right with carefully carefully constructed geologic maps that were that 0:09:59.156 --> 0:10:02.156 might be one hundred years old, and there they were 0:10:02.156 --> 0:10:07.876 a smart skilled geologist just make it a doing field mapping, 0:10:07.916 --> 0:10:12.116 which basically means deserving and recording the observations of outcrops 0:10:12.156 --> 0:10:15.196 and describing the rock, the rock types and those outcrops 0:10:15.516 --> 0:10:19.276 right and locating them in space. And the Earth's crust 0:10:19.356 --> 0:10:22.916 changes very slowly. So provided that was provided that was 0:10:23.076 --> 0:10:27.116 like a well done one hundred years ago, it's still valid. 0:10:27.356 --> 0:10:30.116 It's just that it's it's it's you know, literally in 0:10:30.236 --> 0:10:33.076 drawers piled on top of each other in you know, 0:10:33.236 --> 0:10:36.276 and uh and. 0:10:35.516 --> 0:10:36.836 And basically not used. 0:10:36.876 --> 0:10:39.116 It would only be used by a very industrious human 0:10:39.156 --> 0:10:43.076 being who spent who spent countless hours sort. 0:10:42.916 --> 0:10:44.516 Of searching through the old archives. 0:10:44.596 --> 0:10:47.756 Right, so we go, so we go to various archives, 0:10:47.956 --> 0:10:51.396 and we make an arrangement to digitize the information at 0:10:51.396 --> 0:10:53.676 our expense, and we give the. 0:10:53.476 --> 0:10:56.516 Owners all full digital copy. 0:10:56.956 --> 0:10:59.996 It's almost always public domain data, and so we we 0:11:00.036 --> 0:11:03.076 have a right to use it, or we negotiate a 0:11:03.116 --> 0:11:08.396 specific use right. So digitizing a geologic map is like 0:11:08.676 --> 0:11:10.076 is the very very beginning. 0:11:10.436 --> 0:11:13.556 Then then you need to extract the information from from the. 0:11:13.516 --> 0:11:16.396 Digital copy of the map, uh and, and you have 0:11:16.596 --> 0:11:18.716 many different types of information there. 0:11:18.956 --> 0:11:20.636 You can have in the paper records. 0:11:20.676 --> 0:11:23.876 You might have you might have chemical assays, so measurements 0:11:23.916 --> 0:11:28.876 of the concentrations of the elemental concentrations of samples taken 0:11:28.916 --> 0:11:32.116 from different locations on the map, and that could be a. 0:11:32.076 --> 0:11:33.436 Part of a part of the record. 0:11:33.796 --> 0:11:36.036 And so that's tabular data because it'll say, well, this 0:11:36.116 --> 0:11:39.996 sample sample whatever had had x percent calcium and y 0:11:40.036 --> 0:11:43.356 percent magnesium and z percent silica, and et cetera, et cetera, 0:11:43.396 --> 0:11:46.396 et cetera. That's all valuable information. So that's tabular information 0:11:46.436 --> 0:11:50.876 that then gets extracted by by our systems and populated 0:11:50.916 --> 0:11:55.356 into into what we call our universal schema, which just 0:11:55.436 --> 0:11:58.476 means that every data type is stored in a in 0:11:58.516 --> 0:11:59.596 a consistent format. 0:11:59.916 --> 0:12:03.996 You're standardizing this wildly messy heterogeneous test. 0:12:05.716 --> 0:12:07.396 That's exactly right, And we should talk about more about 0:12:07.396 --> 0:12:09.676 what the data is because it's really fascinating, right, So 0:12:09.876 --> 0:12:11.596 I gave you, I give you two examples. I give 0:12:11.636 --> 0:12:14.956 you the sort of qualitative almost like drawn geologic map, 0:12:14.996 --> 0:12:19.236 which is incredibly useful information, but qualitative and continuous in nature. 0:12:19.556 --> 0:12:21.636 Then there's the sort of tabular data that would be 0:12:21.636 --> 0:12:25.276 any kind of any kind of assay data, measurements of composition. 0:12:25.556 --> 0:12:27.556 But then you have a whole different classes of data, 0:12:27.716 --> 0:12:30.076 like geophysical data, which tells you something. 0:12:29.756 --> 0:12:31.956 About the physics of the Earth's crust. 0:12:32.036 --> 0:12:35.796 So, for example, you probably know that the Earth's gravitational 0:12:35.836 --> 0:12:38.596 field changes from place to place as you move around. 0:12:38.916 --> 0:12:42.076 It changes because you can go up or down in elevation. Well, 0:12:42.076 --> 0:12:44.156 that's easy to adjust for because you know the elevation. 0:12:44.276 --> 0:12:46.916 It also changes because the density of the rocks below 0:12:46.996 --> 0:12:50.836 you change. And so if you're standing over over a 0:12:51.436 --> 0:12:56.116 or body that has twice the density of whose rocks 0:12:56.116 --> 0:12:58.516 are twice as dense as the surrounding rocks that'll pull 0:12:58.556 --> 0:12:59.796 on you slightly more. 0:13:00.196 --> 0:13:02.716 Okay, and you can measure that that I did not know. 0:13:02.796 --> 0:13:05.436 And this is let's go down this rabbit hole actually 0:13:05.436 --> 0:13:09.236 because it's super interesting. Okay, because imagine, so imagine you 0:13:09.276 --> 0:13:12.276 make this measurement. What are you actually measuring. You're measuring 0:13:12.276 --> 0:13:14.956 the force of gravity in a particular location, and you 0:13:14.996 --> 0:13:17.276 can measure Okay, I've adjusted for elevation and the force 0:13:17.276 --> 0:13:19.956 of gravity is a little bit higher here. Okay, that's 0:13:19.996 --> 0:13:22.556 all you actually know at this moment. So what is 0:13:22.556 --> 0:13:24.636 that telling you? Is it telling you you have a 0:13:24.796 --> 0:13:28.436 modestly more dense object like just below the surface, or 0:13:28.476 --> 0:13:30.756 is it telling you you have a massively more dense 0:13:30.796 --> 0:13:33.556 object deeper. Well, it turns out that this is a 0:13:33.636 --> 0:13:37.836 fundamentally degenerate problem or non unique problem would be the 0:13:37.876 --> 0:13:38.756 way to describe it. 0:13:38.796 --> 0:13:39.476 Mathematically. 0:13:39.676 --> 0:13:41.836 There are many ways to solve that problem. There are 0:13:41.836 --> 0:13:43.716 many different configurations. 0:13:43.716 --> 0:13:46.076 You don't know the answer, you know the ant, but. 0:13:46.036 --> 0:13:48.356 You know, what you do know is that there's a 0:13:48.836 --> 0:13:52.956 there's a very large class of in of invalid solutions, 0:13:53.396 --> 0:13:56.956 and then there's a smaller, uh but still very large 0:13:56.996 --> 0:13:58.636 class of valid solutions. 0:13:58.836 --> 0:14:00.676 Ok there's a lot of things that it is not, 0:14:00.916 --> 0:14:02.916 and there's some things that it could be. 0:14:02.796 --> 0:14:06.156 Could exactly incredibly well said, that's that's that's exactly right. 0:14:06.356 --> 0:14:08.796 So here's a there's a really cool application of our 0:14:08.876 --> 0:14:12.196 of our our technology and our approach. So the industry 0:14:12.236 --> 0:14:14.836 standard approach to this is basically, what what does a 0:14:15.156 --> 0:14:18.636 does a normal conventional company do with this gravitational noma. Well, 0:14:18.636 --> 0:14:20.556 they do one of two things. Most of the time, 0:14:20.636 --> 0:14:23.396 maybe ninety percent of the time. They'll just look at 0:14:23.396 --> 0:14:25.676 it and say, okay, here, here is a gravitational anomaly. 0:14:25.876 --> 0:14:28.756 This is hot, it's higher here than here. That's interesting, 0:14:28.796 --> 0:14:30.316 and they just see it on a two D map. 0:14:30.356 --> 0:14:33.956 Okay. So that doesn't do anything for the non uniqueness. 0:14:33.996 --> 0:14:35.276 It just tells you. It just tells you what the 0:14:35.316 --> 0:14:37.956 measurement is, okay. So that now they would just use it. 0:14:37.956 --> 0:14:41.156 They would just say that, well, it's interesting because it's higher, Okay. 0:14:41.156 --> 0:14:46.236 So what Cobold does is very very different and sort 0:14:46.276 --> 0:14:49.436 of impossible to have done even ten years ago, maybe 0:14:49.436 --> 0:14:52.316 even five years ago. What we do is we solve 0:14:52.716 --> 0:14:55.676 a conditioned on a set of geologic hypotheses that we 0:14:55.716 --> 0:15:01.916 find interesting. We solve for the full set of possible subsurfaces. 0:15:02.236 --> 0:15:06.316 So we might actually test like a billion subsurfaces and 0:15:06.396 --> 0:15:12.356 say nine hundred and nine d nine nine nine million, 0:15:12.476 --> 0:15:16.116 nine hundred thousand of those don't match the data we tried. 0:15:16.156 --> 0:15:16.836 They don't match it. 0:15:16.876 --> 0:15:19.236 So there they've been rejected, But we still have one 0:15:19.316 --> 0:15:21.476 hundred thousand now that do. So now we have all 0:15:21.556 --> 0:15:25.076 of the good we've rejected the bad possibilities and we've 0:15:25.156 --> 0:15:27.236 narrowed on the good possibilities. But it's still an incredibly 0:15:27.316 --> 0:15:28.036 large search space. 0:15:28.516 --> 0:15:32.236 Yes, impractical, impractically, you've got to get a lot smaller 0:15:32.276 --> 0:15:33.236 for you to do anything right. 0:15:33.436 --> 0:15:36.276 But there's there's a lot of information in the uncertainty 0:15:36.356 --> 0:15:39.716 that we've now quantified. We've now quantified the uncertainty, and 0:15:39.756 --> 0:15:42.596 so we apply something that we call efficacy of information, 0:15:42.996 --> 0:15:45.556 which is is a phrase that we coined and you 0:15:45.556 --> 0:15:47.796 can read scientific papers about it. It's it's a very 0:15:47.836 --> 0:15:50.636 general and really cool concept, and it's also kind of 0:15:50.676 --> 0:15:52.836 obvious actually, although it's hard to formalize. 0:15:52.956 --> 0:15:55.996 But you look very happy right now. This is an 0:15:56.476 --> 0:15:59.596 audio medium, Like your face is full of delight right now. 0:15:59.596 --> 0:16:01.276 I don't want to interrupt to you. Keep going good. 0:16:01.596 --> 0:16:03.876 Yeah, I'm excited. I love talking about this stuff because 0:16:03.876 --> 0:16:07.116 it's super cool. So we say, okay, the basic idea 0:16:07.156 --> 0:16:11.436 behind EOI is you're going to collect some information next, okay, 0:16:11.476 --> 0:16:14.796 And that's really what every exploration process always is. 0:16:15.636 --> 0:16:17.556 You're going to go try and get more information to 0:16:17.596 --> 0:16:19.876 figure out what is going on in the rocks of 0:16:19.916 --> 0:16:20.356 the surface. 0:16:20.396 --> 0:16:24.876 So the question you obviously want to want to ask 0:16:25.076 --> 0:16:28.716 is what next piece of information will I will tell 0:16:28.716 --> 0:16:33.316 me the most given sort of uh per unit, per 0:16:33.396 --> 0:16:35.636 unit of dollar that I expend, what am I going 0:16:35.716 --> 0:16:36.556 to learn the most from? 0:16:36.996 --> 0:16:39.516 What has what has the highest return on investments? 0:16:39.556 --> 0:16:39.716 Yeah? 0:16:39.756 --> 0:16:42.276 And from a from a knowledge an information perspective, what 0:16:42.316 --> 0:16:44.876 am I going to learn the most from? And so 0:16:44.916 --> 0:16:46.436 here's a way to think about that. It is the 0:16:46.436 --> 0:16:49.436 piece of information that decreases your uncertainty the most. And 0:16:49.476 --> 0:16:52.076 because we solved, we solved the we have the one 0:16:52.116 --> 0:16:56.316 hundred thousand possible answers, right. Yeah, The piece of information 0:16:56.396 --> 0:16:59.756 that will decrease our uncertainty the most is in fact, 0:17:00.276 --> 0:17:04.556 the piece of information that tests the most number that 0:17:04.636 --> 0:17:08.236 falsifies the highest number of those of those one hundred thousand. So, 0:17:08.516 --> 0:17:11.236 for instance, say we're gonna we're gonna drill in different 0:17:11.276 --> 0:17:13.956 in different directions. Okay, yeah, So we're gonna drill and 0:17:13.956 --> 0:17:17.516 we're gonna intersect the sort of various predictions of concentration. 0:17:17.796 --> 0:17:19.596 If we can do one drill hole and it would 0:17:19.596 --> 0:17:22.756 test one hundred of the one hundred thousands, yeah yeah. 0:17:22.596 --> 0:17:23.516 Okay, that's only. 0:17:23.596 --> 0:17:26.316 That's only, we're only testing zero point one percent of 0:17:26.356 --> 0:17:27.316 the possible answers. 0:17:27.476 --> 0:17:29.796 Don't don't drill there, Yeah, don't do that. 0:17:29.836 --> 0:17:30.836 We're gonna learn very little. 0:17:30.876 --> 0:17:33.396 We're gonna end up with basically the same amount of uncertainty. 0:17:33.676 --> 0:17:35.876 But if we could drill a different hole, a different 0:17:35.916 --> 0:17:39.316 core that tests fifty thousand, say half. 0:17:39.076 --> 0:17:41.556 Of them, we massively reduce our search space. 0:17:41.876 --> 0:17:46.716 We we we falsify fifty fifty percent of the possible answers, right, 0:17:46.956 --> 0:17:49.596 and sometimes we can actually falsify like eighty and ninety 0:17:49.636 --> 0:17:52.396 percent of the possible answers. So we massively reduce our 0:17:52.436 --> 0:17:56.876 search space. And we fundamentally it's the most most possible 0:17:56.876 --> 0:17:59.676 information you can get per unit dollar. So every time 0:18:00.036 --> 0:18:03.276 we go to collect any information, we try we try 0:18:03.316 --> 0:18:07.956 to form a formally quantify the uncertainty, and then calculate 0:18:07.996 --> 0:18:10.676 this EOI cont of ter, which is which is the 0:18:11.196 --> 0:18:13.636 piece of information that we have the greatest expectation will 0:18:13.676 --> 0:18:17.036 reduce our uncertainty the most compelling. 0:18:17.636 --> 0:18:18.756 I'm glad you think so. 0:18:18.836 --> 0:18:20.396 It would be nice to be able to do that 0:18:20.476 --> 0:18:22.596 in life. More generally, it's. 0:18:22.476 --> 0:18:25.396 Super super hard, and it comes. The really hard part 0:18:25.596 --> 0:18:30.556 is quantifying the uncertainty correctly. Once you have that quantified correctly, 0:18:30.876 --> 0:18:33.796 then calculating the EOI, like if you know for sure 0:18:34.116 --> 0:18:38.916 you correctly quantified the uncertainty, optimize the calculating the EI 0:18:38.996 --> 0:18:40.396 is kind of an engineering optimization. 0:18:40.436 --> 0:18:41.836 It's kind of it's kind of straightforward. 0:18:42.236 --> 0:18:45.756 In this instance, quantifying the uncertainty is basically how many 0:18:46.036 --> 0:18:49.076 how many ways could the rock under the surface. 0:18:48.796 --> 0:18:50.916 Be exactly Yeah, that's that's exactly right. 0:18:51.076 --> 0:18:53.596 So you've been talking about sort of gathering this very 0:18:53.716 --> 0:18:57.796 old school data and making it useful to you and 0:18:57.836 --> 0:19:00.156 then what you do with it. There is another piece 0:19:00.316 --> 0:19:03.756 of your data gathering operation, or another set of pieces 0:19:04.276 --> 0:19:07.316 that are more high tech and that involve going out 0:19:07.316 --> 0:19:09.276 into the world and getting new data that the then 0:19:09.396 --> 0:19:12.076 already exist. And some of those are kind of fun 0:19:12.476 --> 0:19:15.116 and so I want to talk about that a little bit. 0:19:15.396 --> 0:19:17.076 Tell me about detecting muons. 0:19:17.916 --> 0:19:19.836 Yeah, so this is this is this is kind of 0:19:19.836 --> 0:19:24.436 one of our frontier of R and D projects within 0:19:24.476 --> 0:19:24.956 the company. 0:19:25.036 --> 0:19:27.796 Something we would the opposite of a hundred year old map. 0:19:27.876 --> 0:19:31.596 Yeah, exactly exactly, And we have a lot of a 0:19:31.636 --> 0:19:33.716 lot of physicists at the company, so they love the stuff. 0:19:33.876 --> 0:19:36.356 So what is a muan. Let's start with that and 0:19:36.356 --> 0:19:38.676 then I'll get to the why why they can be useful. 0:19:39.116 --> 0:19:41.836 So in the so cosmic grays are hitting are hitting 0:19:42.276 --> 0:19:45.796 air molecules in the in the upper atmosphere all the time, 0:19:46.236 --> 0:19:49.876 and when when they collide, sometimes they produce they produce 0:19:50.036 --> 0:19:52.316 muons in the in the in the reaction. So it's 0:19:52.356 --> 0:19:55.916 a it's a it's a sub atomic uh particle, a muon, 0:19:56.236 --> 0:19:59.716 and it's it travels very very very fast. It's a 0:19:59.796 --> 0:20:02.916 sort of relativistic particle, travels near the near the speed 0:20:02.916 --> 0:20:03.156 of light. 0:20:03.476 --> 0:20:05.316 So right now muons. 0:20:04.916 --> 0:20:08.316 Are showering through us, you and you in your studio 0:20:08.356 --> 0:20:10.476 and me and my home. Newons are coming through us, 0:20:10.516 --> 0:20:12.276 and they I think about like if you put your 0:20:12.276 --> 0:20:15.516 hand flat, you can expect about one muon per second 0:20:15.716 --> 0:20:17.756 to be going through you through your hand. You don't notice. 0:20:17.836 --> 0:20:20.556 They mostly go right through you. Uh, it doesn't cause 0:20:20.556 --> 0:20:25.036 you any harm. They do interact with electrons, and it 0:20:25.076 --> 0:20:28.356 turns out that when they every time they interact with electron, 0:20:28.476 --> 0:20:29.436 they they. 0:20:29.276 --> 0:20:30.316 Slow down a little bit. 0:20:31.196 --> 0:20:31.476 Uh. 0:20:31.516 --> 0:20:33.836 And when they eventually when when they slow down, eventually 0:20:33.876 --> 0:20:36.236 they slow down enough that they decay into other things. 0:20:36.236 --> 0:20:38.236 Okay, so they they disappear. 0:20:38.556 --> 0:20:41.276 So if you if you are measured, Let's say you 0:20:41.276 --> 0:20:43.156 have a muon detector and it's sitting at the surface 0:20:43.396 --> 0:20:46.276 and you're listening. You know, you like listen to its detection. 0:20:46.396 --> 0:20:50.116 So it's like click click, click click. That's just telling you, okay, 0:20:50.156 --> 0:20:52.116 mwan's going through means going through me's going through right. 0:20:52.156 --> 0:20:52.436 Okay. 0:20:53.076 --> 0:20:56.756 Now I drill a borehole and I start lowering the 0:20:56.836 --> 0:20:59.916 muon detector into the borehole, and you know the rate 0:20:59.956 --> 0:21:00.716 it was at the surface. 0:21:00.756 --> 0:21:02.316 Then as I get to say one hundred meters, now 0:21:02.356 --> 0:21:02.876 it'll be like. 0:21:02.996 --> 0:21:07.396 Click, click click, And then as I go to like 0:21:07.396 --> 0:21:11.476 five hundred meters, it'll be like click. 0:21:12.636 --> 0:21:12.916 Okay. 0:21:13.196 --> 0:21:15.196