1 00:00:00,000 --> 00:00:02,500 In the financial services industry. 2 00:00:03,300 --> 00:00:08,300 It is not so much your IP or your technological Edge, but it 3 00:00:08,300 --> 00:00:11,900 is the ability to day after day year after year. 4 00:00:12,300 --> 00:00:16,500 Deliver consistent, good quality service to your best customers. 5 00:00:16,700 --> 00:00:19,900 Yes. The focus is always on solving 6 00:00:19,900 --> 00:00:23,700 the problem as opposed to dumping or bringing in the most 7 00:00:23,700 --> 00:00:27,400 sophisticated method. Let's start by saying that 8 00:00:27,900 --> 00:00:32,800 typically, when Allah Large company and established form 9 00:00:33,100 --> 00:00:38,900 sets up a data science team. Write the objective is not to 10 00:00:38,900 --> 00:00:41,900 solve a problem per se hit. The objective is to have a data 11 00:00:41,900 --> 00:00:56,400 scientist. Hello and welcome to data 12 00:00:56,400 --> 00:00:59,600 chatter the podcast, on all things data. 13 00:01:00,300 --> 00:01:03,700 This podcast is a series of conversations with experts and 14 00:01:03,700 --> 00:01:06,300 Industry leaders in data. And each week. 15 00:01:06,400 --> 00:01:09,300 We aim to unpack a different compartment of the data 16 00:01:09,300 --> 00:01:11,900 suitcase. I am your host got the chassis 17 00:01:11,900 --> 00:01:15,400 that I want. Logger newspaper columnist book, 18 00:01:15,400 --> 00:01:19,200 author and a former data and sadly consultant at currently 19 00:01:19,200 --> 00:01:21,900 head, analytics, and business intelligence for delivery. 20 00:01:22,300 --> 00:01:23,800 One of India's largest logistics. 21 00:01:23,800 --> 00:01:26,700 Companies. You can follow me on Twitter at 22 00:01:26,700 --> 00:01:32,800 Karthik s that is Kar Phi. K s and read my blog at no 23 00:01:32,800 --> 00:01:35,800 Intruder.com. That is n 0e. 24 00:01:36,200 --> 00:01:42,100 N th you be a.com or opinions expressed in this podcast. 25 00:01:42,100 --> 00:01:44,900 Belong to me and my God. His and they do not reflect the 26 00:01:44,908 --> 00:01:47,500 views of any organizations. We might be Associated. 27 00:01:47,800 --> 00:01:49,600 Nothing discussing. This podcast, will be taken as 28 00:01:49,600 --> 00:01:57,400 fine, a cheap, but England rice. One of the first industries that 29 00:01:57,400 --> 00:02:03,400 started using data in a big way was Financial Services sometime 30 00:02:03,400 --> 00:02:06,800 in the 1970s soon after the black-scholes model got 31 00:02:06,800 --> 00:02:12,000 developed wall, State started hiring phds in maths physics and 32 00:02:12,000 --> 00:02:16,200 Engineering in a large way in order to build models to price 33 00:02:16,200 --> 00:02:18,500 derivatives and help trade better. 34 00:02:19,900 --> 00:02:23,000 However, over the last 10 years, especially after the global 35 00:02:23,000 --> 00:02:25,800 financial crisis. It is appeared that Wall Street 36 00:02:25,800 --> 00:02:29,100 doesn't have the same pole position as it did in terms of 37 00:02:29,100 --> 00:02:31,300 the use of Maths for making money. 38 00:02:33,300 --> 00:02:36,600 So how did this happen? Why is it that the financial 39 00:02:36,600 --> 00:02:40,700 services industry is not in the same poll position as it was, in 40 00:02:40,700 --> 00:02:43,800 terms of use of maths, what has changed with the Financial 41 00:02:43,800 --> 00:02:46,700 Services Industries and what are the incentives that are 42 00:02:46,700 --> 00:02:50,700 different here from Silicon Valley in order to discuss all 43 00:02:50,700 --> 00:02:53,800 of this today. We have hurry balloting. 44 00:02:54,600 --> 00:02:57,000 Honey. Is a co-founder of Romulus and 45 00:02:57,000 --> 00:03:00,300 award-winning unstructured data automation platform for 46 00:03:00,300 --> 00:03:02,900 financial services firms, right? Found. 47 00:03:03,100 --> 00:03:05,000 Tell me this. How do you spent a decade in 48 00:03:05,000 --> 00:03:07,000 Quantum data rooms at Goldman Sachs? 49 00:03:07,600 --> 00:03:16,400 Come back, Singapore. What is data science and what is 50 00:03:16,400 --> 00:03:19,400 artificial, intelligence? How would you describe these? 51 00:03:19,400 --> 00:03:21,200 Two? Sure. 52 00:03:21,300 --> 00:03:25,100 So let me start actually with the, with the more contentious 53 00:03:25,100 --> 00:03:28,600 of the two terms I had, which is basically a, I, you know, I 54 00:03:28,600 --> 00:03:31,000 think this is. So I was actually listening to a 55 00:03:31,000 --> 00:03:36,100 podcast a couple of weeks ago on on cryptography. 56 00:03:37,000 --> 00:03:39,600 Okay. So, the interesting thing there 57 00:03:39,600 --> 00:03:43,600 was that the word crypto was being sort of, obviously, you, 58 00:03:43,800 --> 00:03:46,700 Just in the context of crypto, meaning, cryptography. 59 00:03:47,000 --> 00:03:50,500 And, but every time someone said crypto in my head, it was 60 00:03:50,500 --> 00:03:54,400 actually cryptocurrency. This is a, this is a common 61 00:03:54,400 --> 00:03:57,800 problem where and I think somewhere in the middle of the 62 00:03:57,800 --> 00:04:01,100 podcast, you know, the host talked about how he owned a 63 00:04:01,108 --> 00:04:04,200 shirt, which says crypto means cryptography. 64 00:04:04,500 --> 00:04:07,100 Okay? Okay, and I think the similar 65 00:04:07,100 --> 00:04:13,500 problem exists in the world of AI today where, you know, while 66 00:04:14,100 --> 00:04:17,300 They might have been, let's call it a computer science or 67 00:04:17,300 --> 00:04:21,100 mathematics related discussion around what is artificial 68 00:04:21,100 --> 00:04:24,400 intelligence and all the concepts around strong and weak, 69 00:04:24,400 --> 00:04:27,900 Ai, and sort of, you know, what is really intelligence. 70 00:04:28,500 --> 00:04:31,600 What is happened, is that because of the overuse of this 71 00:04:31,600 --> 00:04:36,700 term, it is now come into common parlance to mean, what we would 72 00:04:36,700 --> 00:04:41,300 probably call we Ka. Okay, which means, you know, 73 00:04:41,300 --> 00:04:44,800 let's actually sort of, you know, junk On a couple of levels 74 00:04:44,800 --> 00:04:47,900 and just start by talking about intelligence, right? 75 00:04:48,000 --> 00:04:51,500 Yeah, the goodness. So, you know, I think there are 76 00:04:51,500 --> 00:04:53,500 many different definitions to this. 77 00:04:53,500 --> 00:04:56,600 And in, in my book, it's basically at some level. 78 00:04:56,600 --> 00:04:58,200 It's the ability to learn, right? 79 00:04:58,200 --> 00:05:02,900 And it's the ability to learn not just from instructions. 80 00:05:02,900 --> 00:05:06,900 And I think that is, that is probably one type of learning, 81 00:05:07,100 --> 00:05:10,200 and you can obviously have a discussion about, you know, is 82 00:05:10,200 --> 00:05:12,700 following instructions really learning or is it really sort of 83 00:05:12,707 --> 00:05:15,800 intelligence? But basically something a bit 84 00:05:15,800 --> 00:05:22,200 deeper where you're able to take past situations, past experience 85 00:05:22,200 --> 00:05:25,900 and map, the down to new situations and come up with some 86 00:05:26,000 --> 00:05:29,800 useful hypotheses or attempts to solve new problems. 87 00:05:29,800 --> 00:05:33,700 And I think at a very sort of high level, this is what I would 88 00:05:33,700 --> 00:05:35,000 call intelligence. Right? 89 00:05:35,000 --> 00:05:37,300 Yep. And if you're able to sort of 90 00:05:37,300 --> 00:05:41,400 demonstrate that machine is able to do this, then you can call 91 00:05:41,400 --> 00:05:43,500 that artificial intelligence, right? 92 00:05:43,800 --> 00:05:49,100 Now I think this is really sort of the core sort of idea that 93 00:05:49,100 --> 00:05:53,800 differentiates. Let's say, you know, AI from 94 00:05:53,800 --> 00:05:56,400 something like bubble sort, right? 95 00:05:56,500 --> 00:05:58,000 Yeah. Where you're able to sort of 96 00:05:58,000 --> 00:06:02,600 write a bunch of code, which are basically an explicit set of 97 00:06:02,600 --> 00:06:05,200 instructions, which can be followed without any 98 00:06:05,200 --> 00:06:08,900 understanding of the context of the underlying problem and still 99 00:06:08,900 --> 00:06:12,700 achieve the result and we want to sort of differentiate. 100 00:06:12,700 --> 00:06:16,100 So I think at the very Basic level in common parlance today. 101 00:06:16,400 --> 00:06:19,800 When people say I they mean something which is not, this is 102 00:06:19,800 --> 00:06:23,600 something which has some semblance of being able to 103 00:06:23,600 --> 00:06:27,600 autonomously solve a problem based on experience rather than 104 00:06:27,600 --> 00:06:29,700 based on instructions. Right? 105 00:06:29,800 --> 00:06:33,400 I think that's sort of the core idea. 106 00:06:33,700 --> 00:06:37,100 Now where that experience comes from is an open question. 107 00:06:37,300 --> 00:06:41,400 You can simulate the experience by, for example, in the case of 108 00:06:41,400 --> 00:06:44,700 alphago, a machine playing. Itself, right? 109 00:06:44,700 --> 00:06:48,300 So it can argue that there is no real world experience here or 110 00:06:48,300 --> 00:06:51,800 you can sort of show. Let's say a machine thousand 111 00:06:51,800 --> 00:06:55,900 images of a cat or a million images of a cat and a militant, 112 00:06:55,900 --> 00:06:59,600 which is a dog and and build a classifier which distinguishes 113 00:06:59,600 --> 00:07:01,600 between what a cat and a dog look like. 114 00:07:02,200 --> 00:07:06,500 So the, but the basic thumb rule is that you need to be creating 115 00:07:06,500 --> 00:07:11,200 some form of experience for the system to learn from as opposed 116 00:07:11,200 --> 00:07:15,700 to co defying rules. And and sort of creating if 117 00:07:15,700 --> 00:07:19,400 something is closer to bubble sort, then it is to sort of, you 118 00:07:19,400 --> 00:07:22,600 know, showing examples to distinguish between a cat and a 119 00:07:22,608 --> 00:07:24,900 dog. Then I think that is probably 120 00:07:24,900 --> 00:07:27,800 not a right. But once again, the term has 121 00:07:27,800 --> 00:07:31,600 become very murky because we are sort of it's become sort of this 122 00:07:31,600 --> 00:07:35,300 lowest common denominator term because it's being bandied about 123 00:07:35,300 --> 00:07:39,000 so often. And which is I mean, you can be 124 00:07:39,500 --> 00:07:42,500 I'm not much of a silly saying it's a bad thing, but 125 00:07:42,500 --> 00:07:46,300 essentially it is It is now driven by what a large portion 126 00:07:46,300 --> 00:07:50,100 of the population understands this to mean, right? 127 00:07:50,100 --> 00:07:55,500 As opposed to what you know, it can the discussions around what 128 00:07:55,500 --> 00:08:01,400 AI is in the you know in the halls of Academia and sort of 129 00:08:01,400 --> 00:08:04,400 all these ideas around strong which is weak, AI think these 130 00:08:04,400 --> 00:08:07,500 things are all sort of now sidelined and it's become this 131 00:08:07,500 --> 00:08:10,600 catch-all term, which is I mean, I'm not going to comment on 132 00:08:10,600 --> 00:08:12,500 whether that's a good or a bad thing, but that's what it is 133 00:08:12,500 --> 00:08:15,100 today now. Now data science, which is the 134 00:08:15,100 --> 00:08:17,200 other term. You brought up is perhaps a bit 135 00:08:17,200 --> 00:08:20,500 less contentious. And once again, I think in my 136 00:08:20,500 --> 00:08:24,700 book the way I think about data science is it's basically the 137 00:08:24,700 --> 00:08:31,000 ability of ability of someone to build a mathematical model that 138 00:08:31,000 --> 00:08:36,500 can interpret real world data right now, is that definition 139 00:08:36,500 --> 00:08:41,100 very close to statistics. Perhaps I don't have a better 140 00:08:41,100 --> 00:08:46,200 answer other than possibly data. So that first two, you know, 141 00:08:46,600 --> 00:08:51,600 performing this activity in a practical situation as opposed 142 00:08:51,600 --> 00:08:54,100 to in a more theoretical situation, right? 143 00:08:54,200 --> 00:08:57,400 And I think that also brings to data science somewhat of an 144 00:08:57,400 --> 00:09:03,500 artisanal quality wherein, you know that you need to have, the 145 00:09:03,500 --> 00:09:07,900 more you do it probably the more intuition, you develop around it 146 00:09:08,200 --> 00:09:12,900 and the better you get. Yes, and also there are many 147 00:09:12,900 --> 00:09:15,600 many instances. Has where, you know, there is no 148 00:09:15,600 --> 00:09:20,900 one answer, right? And I think that's what makes 149 00:09:22,100 --> 00:09:25,000 that I think that's what gives our data Sciences artisanal 150 00:09:25,000 --> 00:09:29,100 quality in that, you know, every person I have met who professes 151 00:09:29,100 --> 00:09:33,500 to be a data scientist, has their own favorite approach or 152 00:09:33,500 --> 00:09:37,300 their own favorite model or their own internal mental sort 153 00:09:37,300 --> 00:09:41,000 of prioritization of what they would like to try first given a 154 00:09:41,000 --> 00:09:43,200 problem. So I think that's what makes it 155 00:09:43,200 --> 00:09:44,600 very interesting. I think I'm different. 156 00:09:45,800 --> 00:09:49,900 And yeah, so I think that's probably how I would you know, 157 00:09:49,900 --> 00:09:52,500 let's say summarized data science. 158 00:09:52,800 --> 00:09:53,400 Okay. Thanks. 159 00:09:53,400 --> 00:09:56,300 I mean, so while you were describing data since I got 160 00:09:56,300 --> 00:09:59,200 reminded of some old Jokes which was like data Sciences 161 00:09:59,200 --> 00:10:04,300 statistics done on a Mac or data Sciences statistics done in 162 00:10:04,300 --> 00:10:09,400 California or a data scientist knows more statistics than a 163 00:10:09,400 --> 00:10:12,100 software engineer and more software engineering than a 164 00:10:12,108 --> 00:10:14,600 statistician and so on. This is a whole bunch of jokes 165 00:10:14,600 --> 00:10:16,800 that it came out back in 2013. 14. 166 00:10:16,800 --> 00:10:20,800 I think when data science was just taking notes, but I think I 167 00:10:20,800 --> 00:10:23,200 like your, I like your answer which says that like, I mean, 168 00:10:23,300 --> 00:10:26,600 that it is like more of a artisanal quality that each 169 00:10:26,600 --> 00:10:29,400 person did has their own pet methods, of approaching a 170 00:10:29,408 --> 00:10:33,000 problem and things like that. I think I completely, I sort of 171 00:10:33,200 --> 00:10:37,200 subscribe to that as well. So, now, now that we talked 172 00:10:37,200 --> 00:10:38,900 about, so what do you were, what do you do? 173 00:10:38,900 --> 00:10:41,900 Whatever we discussed AI, we discussed data science. 174 00:10:41,900 --> 00:10:43,400 Do you put yourself in any of it? 175 00:10:43,700 --> 00:10:46,400 All about you and your firm. Do you put yourself in any of 176 00:10:46,500 --> 00:10:49,100 these buckets? Or like how do you describe the 177 00:10:50,900 --> 00:10:53,800 what you do? Sure. 178 00:10:53,800 --> 00:10:58,600 So let me, I think, I think that what should have, I think 179 00:10:58,600 --> 00:11:01,200 requires a lot of background, but yeah, but basically, a very 180 00:11:01,200 --> 00:11:05,900 high level, right? So we are basically, I would do 181 00:11:05,900 --> 00:11:07,000 if I had to summarize, what we do. 182 00:11:07,000 --> 00:11:11,300 We are a software firm that's building software products for 183 00:11:11,300 --> 00:11:13,500 financial services and at the highest level. 184 00:11:13,700 --> 00:11:17,900 Yes, right. And I do not want to bring in 185 00:11:18,100 --> 00:11:22,600 words like Ai and data science into the picture because I think 186 00:11:22,600 --> 00:11:28,500 those Our Legacy tools, or those are expressions of what we do as 187 00:11:28,500 --> 00:11:32,800 opposed to defining qualities of our form, right? 188 00:11:32,800 --> 00:11:35,000 Okay. Exist in. 189 00:11:35,000 --> 00:11:39,300 So, the company that I founded in 2017, and now run is called 190 00:11:39,300 --> 00:11:44,700 egregore labs and we built a wide range of software products. 191 00:11:45,000 --> 00:11:50,600 And I think the core idea really is to solve a set of financial 192 00:11:50,600 --> 00:11:55,000 services problems. Each we think married solving 193 00:11:55,000 --> 00:12:02,300 and which are probably we think, you know, not so sexy or hot but 194 00:12:02,300 --> 00:12:07,100 at the same time are incredibly valuable to solve, right? 195 00:12:07,100 --> 00:12:11,800 And I think that is sort of been let's say the driving force or 196 00:12:11,800 --> 00:12:15,800 the principle behind what we have tried to build and what we 197 00:12:15,808 --> 00:12:18,500 have, what we continue building within the company. 198 00:12:20,500 --> 00:12:22,800 So that's like a very sort of high level that. 199 00:12:22,900 --> 00:12:24,800 I would say, that's what we do. Now. 200 00:12:24,800 --> 00:12:29,200 We do end up using elements of, you know, natural language 201 00:12:29,200 --> 00:12:31,800 processing. Some machine learning and deep 202 00:12:31,800 --> 00:12:33,800 learning within the company, as well. 203 00:12:33,800 --> 00:12:37,200 As, you know, a lot of traditional programming and 204 00:12:37,200 --> 00:12:40,700 traditional software building. So all of that comes together, 205 00:12:40,700 --> 00:12:45,800 but I think we are where I think what we have learned the hard 206 00:12:45,800 --> 00:12:52,700 way and when I say we I mean me as a Founder is that the focus 207 00:12:52,700 --> 00:12:57,800 is is always on solving the problem as opposed to dumping or 208 00:12:57,800 --> 00:13:00,800 bringing in the most sophisticated method, right? 209 00:13:01,100 --> 00:13:06,300 If so, you're not ashamed of your humble enough to admit that 210 00:13:06,300 --> 00:13:10,500 there are situations where addicts will do and complex LP 211 00:13:10,500 --> 00:13:14,500 is not required and we are quite happy and humbled to sort of 212 00:13:14,800 --> 00:13:16,900 take that approach. Okay, right. 213 00:13:17,700 --> 00:13:20,000 And I think that's, that's really sort of been the big 214 00:13:20,000 --> 00:13:22,800 learning from the last couple of years for me. 215 00:13:23,400 --> 00:13:28,700 Which is that the problem comes first and the method is really 216 00:13:28,700 --> 00:13:30,900 whatever solves the problem the best. 217 00:13:32,000 --> 00:13:36,700 And you must have a Swiss army knife mentality, where you're 218 00:13:36,700 --> 00:13:41,500 willing to learn, pick up things as required, to solve the 219 00:13:41,500 --> 00:13:47,600 problem in the most, you know, economical / complete, you know, 220 00:13:47,600 --> 00:13:53,700 customer satisfying fashion as opposed to Sort of you know 221 00:13:54,100 --> 00:13:57,700 approach the problem in a more of a screwdriver, a hammer and 222 00:13:57,700 --> 00:14:00,800 nail fashion, where you have this particular tool you're 223 00:14:00,800 --> 00:14:03,500 excited about and now you are going around applying it to 224 00:14:03,500 --> 00:14:06,000 every context. I agree with a lot of it. 225 00:14:06,000 --> 00:14:08,900 I mean, I strongly believe believe that you need the 226 00:14:08,900 --> 00:14:11,300 scissor mean I for a toolbox kind of an approach to other 227 00:14:11,300 --> 00:14:15,700 than a hammer and nail kind of an approach which again. 228 00:14:16,000 --> 00:14:19,700 So what are you? And I completely says subscribe 229 00:14:19,700 --> 00:14:22,100 to that. So but how do you let's say your 230 00:14:22,100 --> 00:14:25,200 customers take. How do you are the sort of 231 00:14:26,100 --> 00:14:28,500 competitors? Take it and so on, like what is 232 00:14:28,500 --> 00:14:30,400 there? How do they approach the problem 233 00:14:30,600 --> 00:14:32,200 in the financial industry. What is it? 234 00:14:32,200 --> 00:14:36,200 Like now in terms of like use of use of data? 235 00:14:36,200 --> 00:14:39,300 If I were to call it that is in terms of solving the non-sexy 236 00:14:39,300 --> 00:14:42,500 problems that you that you mention, right? 237 00:14:42,800 --> 00:14:46,500 So I think, right? So to answer that question. 238 00:14:46,500 --> 00:14:52,400 I think we kind of need to sort of spend some time going down 239 00:14:52,600 --> 00:14:54,700 the The the roads of History. A little bit. 240 00:14:54,700 --> 00:14:55,300 Right? Right. 241 00:14:56,400 --> 00:15:02,000 So if you sort of looked at what you know, let's give are. 242 00:15:02,000 --> 00:15:04,900 Let's call it interesting. Math for the lack of a better 243 00:15:04,900 --> 00:15:09,900 word was being applied in the financial services industry for 244 00:15:09,900 --> 00:15:12,100 for let's say from the 80s or probably earlier. 245 00:15:12,200 --> 00:15:16,200 So from then to let's say, you know, the first decade of the 246 00:15:16,200 --> 00:15:22,700 2000s, I think it was largely around being able to model. 247 00:15:22,900 --> 00:15:26,500 Six places better. Yep, and and so, when we say 248 00:15:26,500 --> 00:15:29,700 asset prices here, we mean, you know, things that are real, 249 00:15:30,400 --> 00:15:34,800 which could be, let's say the price of an insurance policy, 250 00:15:34,800 --> 00:15:37,500 which is, which is what, some whatever tangible nature to it 251 00:15:37,500 --> 00:15:39,800 to something. That's a derivative, to 252 00:15:39,800 --> 00:15:41,600 something that's a derivative derivative and so on. 253 00:15:41,600 --> 00:15:46,700 So there is basically a world of underlying assets and then there 254 00:15:46,700 --> 00:15:49,300 are derivatives piled onto that and sort of add infinite right? 255 00:15:49,300 --> 00:15:52,800 There are sort of as you say tortoise is all the way. 256 00:15:52,900 --> 00:15:56,200 Down in many ways. I think this is also 257 00:15:56,500 --> 00:15:58,900 interestingly enough. I think some more over Zero Sum 258 00:15:58,900 --> 00:16:01,400 game, right? Where there's a very competitive 259 00:16:01,400 --> 00:16:05,900 nature of this particular to this particular industry, which 260 00:16:05,900 --> 00:16:10,400 is you know, and remember that when we are talking about asset 261 00:16:10,400 --> 00:16:12,300 prices and we are talking about derivatives. 262 00:16:13,000 --> 00:16:17,600 We are not talking about what we would call the primary markets 263 00:16:17,600 --> 00:16:20,300 as yet, right? So, you know, we're not talking 264 00:16:20,300 --> 00:16:24,500 about pricing a stock into IPO. Which one? 265 00:16:24,600 --> 00:16:26,800 What would fall? Very neatly into sort of the 266 00:16:27,000 --> 00:16:29,600 investment banking. Origination Mna words, all of 267 00:16:29,600 --> 00:16:32,200 that. I'm kind of removing from this 268 00:16:32,200 --> 00:16:35,900 conversation because, you know, I so we're talking really about 269 00:16:35,900 --> 00:16:39,100 secondary markets and there is a bit of a competitive nature to 270 00:16:39,100 --> 00:16:44,200 it because you know, you are competing with every other bank 271 00:16:44,200 --> 00:16:48,600 on the street in order to be able to sort of price and asset 272 00:16:48,600 --> 00:16:52,500 better or more accurately portrayed an asset better and 273 00:16:52,500 --> 00:16:54,200 effort. There is a certain, you know, 274 00:16:54,200 --> 00:16:59,900 darwinian nature to this. What would, what we could call a 275 00:16:59,900 --> 00:17:02,900 zero-sum game because there is just so much money to be made in 276 00:17:02,900 --> 00:17:06,099 the market through trading, and you have all these competing 277 00:17:06,200 --> 00:17:10,000 players. And the interesting nature of 278 00:17:10,000 --> 00:17:12,400 this problem. Is that anyone? 279 00:17:12,400 --> 00:17:17,700 Someone I met very early in the journey, in my journey in the in 280 00:17:17,700 --> 00:17:20,800 the financial services industry said that, the the one of the 281 00:17:20,800 --> 00:17:22,700 things that makes the financial services industry. 282 00:17:23,099 --> 00:17:25,200 Very interesting is the proximity to money rate. 283 00:17:25,200 --> 00:17:31,100 Where, you know, if you, if you make, let's say $100 trading 284 00:17:31,100 --> 00:17:35,600 today, right? You know, there is some slippage 285 00:17:35,600 --> 00:17:37,900 to let's say, you know, some brokerage house some 286 00:17:37,900 --> 00:17:41,400 commissions, this possibly some taxes to be paid. 287 00:17:42,100 --> 00:17:44,500 If you're, if you're working within large form. 288 00:17:44,600 --> 00:17:46,900 Then, you know, there is some element where the form dig some 289 00:17:46,900 --> 00:17:50,300 of that away, but essentially, there is a very sort of simple 290 00:17:50,300 --> 00:17:52,300 straightforward equation to how much of that lands in your 291 00:17:52,300 --> 00:17:53,300 pocket. That right. 292 00:17:53,700 --> 00:17:56,900 There is no, there's no complexity in terms of how do 293 00:17:56,900 --> 00:18:03,000 you draw the translation of your skill or effort or your result 294 00:18:03,400 --> 00:18:07,300 into a financial gain? For you is very linear in many 295 00:18:07,300 --> 00:18:08,100 ways. Right? 296 00:18:08,900 --> 00:18:11,000 And there is no marketing element. 297 00:18:11,300 --> 00:18:15,000 There is no sort of, let's say there is no fuzziness and then 298 00:18:15,000 --> 00:18:17,200 attracts a lot of people to the industry. 299 00:18:17,200 --> 00:18:21,700 Because, you know, very quickly you can sort of make a lot of 300 00:18:21,700 --> 00:18:25,400 money. Simply by performing by having 301 00:18:25,400 --> 00:18:28,000 one skill value, very good at or like a set of skills, which are 302 00:18:28,000 --> 00:18:33,200 very good at and you're very tangible results to show and 303 00:18:33,200 --> 00:18:38,900 therefore, you know, argue that you should basically be rewarded 304 00:18:38,900 --> 00:18:41,600 in a certain fashion. So in many ways, I think it's a 305 00:18:41,600 --> 00:18:45,300 sort of very short distance between result and instead of 306 00:18:45,300 --> 00:18:48,200 personal outcome for you. And the other thing which makes 307 00:18:48,200 --> 00:18:50,500 the industry. Very interesting at least like, 308 00:18:50,500 --> 00:18:55,300 you know, from the 70s. To the Greenspan years is that 309 00:18:56,000 --> 00:19:02,500 we were we were in a world where I would say possibly cost of 310 00:19:02,500 --> 00:19:06,800 capital was sort of going up at least through the Greenspan has 311 00:19:06,800 --> 00:19:11,000 always expected to go. And basically, how much you 312 00:19:11,008 --> 00:19:13,900 could get paid for an incremental amount of risk was 313 00:19:13,900 --> 00:19:16,700 quite huge, right? And that manifested itself in 314 00:19:16,700 --> 00:19:20,000 various ways. Starting from, you know, what is 315 00:19:20,000 --> 00:19:22,700 the rate differential or what is the sort of? 316 00:19:22,900 --> 00:19:25,900 Referencing even between an emerging market and develop 317 00:19:25,900 --> 00:19:29,300 Market asset, right? Quite quite massive, right. 318 00:19:29,300 --> 00:19:33,200 And to a whole host of things like carrots, and generally 319 00:19:33,200 --> 00:19:35,200 Market, volatility was also very high. 320 00:19:35,200 --> 00:19:37,800 And you could say that we're going through a period where 321 00:19:37,800 --> 00:19:40,900 there was a lot of action. There was a lot of willingness 322 00:19:40,900 --> 00:19:45,000 to participate, that created, liquidity markets, interesting. 323 00:19:45,000 --> 00:19:48,200 And broadly, there was money to be made from Financial 324 00:19:48,300 --> 00:19:52,100 engineering, right? And you know, we sort of got 325 00:19:52,100 --> 00:19:56,100 into this arm Arms raised for yield especially in the private 326 00:19:56,100 --> 00:20:00,700 wealth and fixed-income world. And also, you know, you had gone 327 00:20:00,700 --> 00:20:04,200 through this very interesting fails in the early late 90s 328 00:20:04,200 --> 00:20:09,400 early 2000s where there was there was sort of this one 329 00:20:09,400 --> 00:20:13,500 opportunity when you know, the the equity markets. 330 00:20:13,500 --> 00:20:18,900 Specifically, let's say that the tech world you had this had this 331 00:20:18,900 --> 00:20:22,500 sort of ability to suck away all this liquidity. 332 00:20:22,900 --> 00:20:25,900 Um, into sort of tech Investments. 333 00:20:26,200 --> 00:20:28,900 And for whatever reason, that had not really sort of panned 334 00:20:28,900 --> 00:20:31,100 out the way people expected and the tech bubble had burst, 335 00:20:31,100 --> 00:20:33,300 right? So there was sort of a let's 336 00:20:33,300 --> 00:20:35,500 read employee Capital into other assets. 337 00:20:36,300 --> 00:20:39,000 Let's sort of find a new sort of a thing. 338 00:20:39,000 --> 00:20:41,700 Fixed income is very interesting at that point, for some reason, 339 00:20:41,700 --> 00:20:43,400 which is that people wanted safety. 340 00:20:43,400 --> 00:20:44,500 People want to guaranteed returns. 341 00:20:44,700 --> 00:20:48,700 So, for for a bunch of various different reasons, essentially, 342 00:20:48,700 --> 00:20:53,000 you kind of ended up in this interesting decade where Are 343 00:20:53,400 --> 00:20:58,000 there was a progressive complexification of financial 344 00:20:58,000 --> 00:21:00,500 products. Yeah, and people were quite 345 00:21:00,500 --> 00:21:06,000 happy sort of, you know, buying trading complexity for yield. 346 00:21:06,400 --> 00:21:08,500 They were willing to sort of purchase things that they didn't 347 00:21:08,500 --> 00:21:11,300 completely understand because they were yielding of better, 348 00:21:13,400 --> 00:21:17,700 giving them better income and through the world was going in 349 00:21:17,700 --> 00:21:19,500 One Direction. And you had sort of a series of 350 00:21:19,500 --> 00:21:22,600 rate hikes things seem very predictable housing market. 351 00:21:22,800 --> 00:21:27,600 Booming and you know, frankly regulation was not that intense, 352 00:21:27,700 --> 00:21:28,200 right? Yeah. 353 00:21:28,200 --> 00:21:32,400 And so you had everyone sort of a very sort of large intense 354 00:21:32,600 --> 00:21:35,800 population of folks who are letter broadly, good at math, 355 00:21:36,800 --> 00:21:40,800 kind of accumulate in this one industry and sort of, you know, 356 00:21:40,800 --> 00:21:43,500 competed each other find excitement happiness, in sort of 357 00:21:43,500 --> 00:21:46,900 building more and more complex models and coming up with new 358 00:21:47,000 --> 00:21:50,300 products and sort of essentially, the, the going was 359 00:21:50,300 --> 00:21:51,500 good. Great. 360 00:21:52,100 --> 00:21:55,100 And then we Had, of course, the GFC and you can sort of, 361 00:21:55,100 --> 00:21:57,200 obviously argue. And I don't think too many 362 00:21:57,200 --> 00:21:59,300 people disagree with you that in some ways. 363 00:21:59,300 --> 00:22:03,800 It was pretty, it was precipitated by, you know, these 364 00:22:03,800 --> 00:22:08,900 complex products and and sort of, let us say the way they 365 00:22:08,900 --> 00:22:12,500 behaved and the way they influenced the underlying assets 366 00:22:13,200 --> 00:22:17,200 instead of a market where conditions are quite different 367 00:22:17,200 --> 00:22:21,100 from what they had been before, you know, JFC commenced. 368 00:22:21,500 --> 00:22:25,600 Yeah, and, you know, People sort of had to sort of then agree and 369 00:22:25,600 --> 00:22:28,900 eat Humble Pie and agree that they had misplaced assets. 370 00:22:28,900 --> 00:22:31,900 And not really model things correctly and gotten too caught 371 00:22:31,900 --> 00:22:35,500 up in the math and let the whole bunch of things happened, right? 372 00:22:35,500 --> 00:22:38,900 Which is that, you know, the first thing that happened was 373 00:22:38,900 --> 00:22:45,600 that you had in the short run. You had basically this flight 374 00:22:45,600 --> 00:22:49,100 away from risk and complexity, where people were unwilling to 375 00:22:49,100 --> 00:22:51,400 touch things that were complex, which they didn't understand. 376 00:22:51,800 --> 00:22:54,000 And they sort of Went in the opposite direction. 377 00:22:54,700 --> 00:22:57,800 The second thing was that the financial Regulators came in and 378 00:22:57,900 --> 00:23:00,900 said, you know what like, you know, we need you to make sure 379 00:23:00,900 --> 00:23:04,800 that we don't have to bail you out again and you need to sort 380 00:23:04,800 --> 00:23:10,100 of show your BMO Prudential in the way, you manage risk and the 381 00:23:10,100 --> 00:23:14,400 kind of products you sell and you know, a lot of the Caveat 382 00:23:14,400 --> 00:23:19,300 Emptor approach where, you know, the person who's buying knows 383 00:23:19,300 --> 00:23:21,000 what they are buying and it's all the risk. 384 00:23:21,000 --> 00:23:25,100 If it goes south that It's to sort of selling or financial 385 00:23:25,100 --> 00:23:30,000 products kind of went away to some extent and progressively we 386 00:23:30,000 --> 00:23:32,500 sort of ended up in a world and obviously then you know, rates 387 00:23:32,500 --> 00:23:36,900 came down and that led to sort of a completely different 388 00:23:37,100 --> 00:23:40,900 approach where, you know, it was very clear that we were going to 389 00:23:40,900 --> 00:23:45,000 be sort of, you know, we had multiple multiple rounds of QE 390 00:23:45,000 --> 00:23:49,400 and differentials in rates between M, and DM, go down a lot 391 00:23:49,400 --> 00:23:54,700 of, let's call it macro factors. Ders, you have slowly made this 392 00:23:54,700 --> 00:23:56,800 industry progressively less interesting. 393 00:23:57,100 --> 00:23:59,300 Right? I mean, yeah, number of reasons. 394 00:23:59,800 --> 00:24:02,400 The most interesting aspect of that, was that we sort of moved 395 00:24:02,400 --> 00:24:06,500 away from Financial engineering to what we call, you know, 396 00:24:06,600 --> 00:24:09,100 rattle based engineering regulatory accounting tax and 397 00:24:09,100 --> 00:24:10,200 legal, right? Okay. 398 00:24:10,300 --> 00:24:12,900 Okay. So how do we sort of create 399 00:24:12,900 --> 00:24:17,000 products that, you know, and and, and sort of do business 400 00:24:17,100 --> 00:24:21,300 that falls squarely and clearly within the boundary conditions 401 00:24:21,300 --> 00:24:24,500 created by Regulations accounting tax and legal. 402 00:24:25,400 --> 00:24:32,700 And all four were sort of, you know, thing rethinking how the 403 00:24:32,700 --> 00:24:34,900 lens from which they were viewing Financial products in 404 00:24:34,908 --> 00:24:36,100 the financial industry as a whole. 405 00:24:36,600 --> 00:24:38,500 Now. These I would say are all the, 406 00:24:38,700 --> 00:24:41,000 let's call it the headwinds in some ways, right? 407 00:24:41,000 --> 00:24:45,300 Yeah, and now, let's talk about the Tailwind, which sort of led 408 00:24:45,300 --> 00:24:49,400 to a drift away, right? And and sort of created new and 409 00:24:49,400 --> 00:24:51,800 interesting things for the financial industry. 410 00:24:51,800 --> 00:24:54,400 To look at. So, I think the first thing I 411 00:24:54,400 --> 00:24:58,100 would say there is really at some high level digital 412 00:24:58,100 --> 00:25:03,600 transformation, right? Yep, where you had a number of 413 00:25:04,500 --> 00:25:09,600 Industries companies, Etc, move in the direction of recording 414 00:25:09,600 --> 00:25:14,300 and measuring things and putting them inside, you know, a 415 00:25:14,300 --> 00:25:17,400 database. Yep, which they were not doing 416 00:25:17,400 --> 00:25:20,600 before. So people were getting more data 417 00:25:20,600 --> 00:25:21,800 Centric. Broadly. 418 00:25:21,800 --> 00:25:24,200 I think this was one Sort of, let's say a market wide 419 00:25:24,200 --> 00:25:27,200 phenomenon. The second thing that happened 420 00:25:27,200 --> 00:25:30,900 was I would see the Revival of Silicon Valley in some in some 421 00:25:30,900 --> 00:25:36,700 way where, you know, you had, you know, the the rise and Rise 422 00:25:36,700 --> 00:25:40,800 of fang. Plus the creation of 423 00:25:40,800 --> 00:25:47,500 interesting, you know, jobs rolls ideas things to work on 424 00:25:47,500 --> 00:25:53,000 for people who are let's say math oriented or Computer 425 00:25:53,000 --> 00:25:56,200 science oriented and coming out of colleges and universities in 426 00:25:56,208 --> 00:25:59,800 the US and elsewhere. Yeah, I think that created a 427 00:25:59,900 --> 00:26:05,800 second change and then you know, the suddenly you're sort of 428 00:26:05,800 --> 00:26:07,800 ended up in this world very quickly. 429 00:26:08,200 --> 00:26:12,400 Where Capital was broadly accessible to everyone, right? 430 00:26:12,500 --> 00:26:15,000 Yes. So it no longer became and I 431 00:26:15,000 --> 00:26:19,200 think we're seeing that today where you know, it's not Capital 432 00:26:19,200 --> 00:26:23,300 Access to Capital is it might be a differentiator for a But for a 433 00:26:23,300 --> 00:26:26,200 large company, it's no longer a differentiator, right? 434 00:26:26,200 --> 00:26:30,100 Yep, and the things that you needed to do. 435 00:26:30,100 --> 00:26:33,600 The story is you needed to be able to tell and how you 436 00:26:33,600 --> 00:26:38,900 convince investors to give you Capital changed right there. 437 00:26:38,900 --> 00:26:42,500 For a couple of changes started to happen. 438 00:26:42,500 --> 00:26:46,300 First of all, you know, people started looking and equities. 439 00:26:46,300 --> 00:26:52,400 I think far more closely than they did before GFC and focus. 440 00:26:52,700 --> 00:26:56,300 I came especially in Tekken Healthcare to sort of find 441 00:26:56,500 --> 00:27:00,700 assets, which truly had an edge right now. 442 00:27:00,700 --> 00:27:04,700 If you have to measure loyalty, for example, a great way to do 443 00:27:04,700 --> 00:27:07,600 that is reviews, right? Yeah. 444 00:27:07,700 --> 00:27:11,800 So if you have, you know, a product that is directly being 445 00:27:11,800 --> 00:27:16,300 sold to a customer and at the same time, you have, you know, 446 00:27:16,800 --> 00:27:20,300 Facebook, Instagram, Twitter, and all of these widely 447 00:27:20,300 --> 00:27:22,500 available as well as a number of forums that existed. 448 00:27:22,700 --> 00:27:28,100 Before of Hardcore aficionados their response and reaction to a 449 00:27:28,100 --> 00:27:32,300 new product launch became very important, you know, 100 other 450 00:27:32,600 --> 00:27:35,100 of these data sources. I mean, ranging from you know, 451 00:27:35,100 --> 00:27:37,800 satellite images parking lot pictures of parking, lots 452 00:27:37,800 --> 00:27:44,300 infrared, cameras whole sort of flurry of information, which you 453 00:27:44,300 --> 00:27:49,000 would call all data, right? And which basically then start 454 00:27:49,000 --> 00:27:53,900 to become interesting sources of Information, right? 455 00:27:54,000 --> 00:27:56,900 And for the first time, you know, post your the Advent of 456 00:27:56,900 --> 00:28:01,800 big data and you know certain let's call it breakthroughs in 457 00:28:02,900 --> 00:28:07,200 the area of NLP, broadly, right and everything associated with 458 00:28:07,200 --> 00:28:10,900 neural networks. All of a sudden if you had data 459 00:28:10,900 --> 00:28:14,100 and you could throw it at a neural network, you had the 460 00:28:14,100 --> 00:28:17,100 potential to basically make some sense of it. 461 00:28:17,400 --> 00:28:20,000 Yep. Late and, you know things that 462 00:28:20,000 --> 00:28:23,300 were not interesting since the 70s and which were actually Make 463 00:28:23,400 --> 00:28:28,400 ideas suddenly became reality in the current in the last decade, 464 00:28:28,400 --> 00:28:29,100 right? Yep. 465 00:28:29,100 --> 00:28:33,500 And so, you had, you know, you had compute. 466 00:28:33,900 --> 00:28:40,100 You had the, let's call it the data ology you had data. 467 00:28:40,500 --> 00:28:46,200 And most importantly, I would say that a lot of the 468 00:28:46,200 --> 00:28:50,100 breakthroughs and ideas were in the free and open-source 469 00:28:50,900 --> 00:28:56,500 software Community, then I think Typically Wall Street is never 470 00:28:56,700 --> 00:29:01,000 sort of a good first mover but a great, you know, follower, 471 00:29:01,900 --> 00:29:05,300 right? So so I think you had many of 472 00:29:05,300 --> 00:29:09,600 these Trends becoming mainstream and interesting and you know, 473 00:29:10,400 --> 00:29:15,400 frankly, I would say the the how useful is all this data and 474 00:29:15,400 --> 00:29:17,900 like, you know, as all data really taken over and have 475 00:29:17,900 --> 00:29:22,800 people made money off of it and has it been sort of has the The 476 00:29:22,800 --> 00:29:26,200 of this been proven I would say it's quite hard to say. 477 00:29:26,500 --> 00:29:29,500 Okay, obviously the very systematic and you know, 478 00:29:29,500 --> 00:29:31,200 long-standing players in this market. 479 00:29:31,200 --> 00:29:33,800 Like, you know, whether like The Medallion fund you have 480 00:29:33,800 --> 00:29:37,800 consistently shown themselves to be all-weather winners, right? 481 00:29:37,800 --> 00:29:41,700 If I can use that term but as is the average fund really been 482 00:29:41,700 --> 00:29:44,700 able to has the electron really been able to extract Alpha from 483 00:29:45,700 --> 00:29:48,400 these data feeds. It's unclear to me. 484 00:29:49,000 --> 00:29:53,000 And from the all data perspective and you all data 485 00:29:53,000 --> 00:29:56,400 Universe. I do think that if you sort of 486 00:29:56,400 --> 00:30:01,300 look at the curve, right? And one possible, an example of 487 00:30:01,300 --> 00:30:06,900 this of a company that has lived through the entire cycle would 488 00:30:06,900 --> 00:30:11,400 be Kwon Do, Right. Yeah, which sort of, you know, 489 00:30:11,400 --> 00:30:16,000 was birthed, I would say at the Advent of panel data was still 490 00:30:16,000 --> 00:30:18,300 sort of, you know, let's call it a niche idea. 491 00:30:18,500 --> 00:30:20,600 Yeah. Found a way to create a 492 00:30:20,600 --> 00:30:25,100 Marketplace soft. Nominal growth was sold to a 493 00:30:25,100 --> 00:30:30,200 snack and now just this week or last week, the founders of spent 494 00:30:30,200 --> 00:30:33,500 the time as NASDAQ and I are have left to sort of, you know, 495 00:30:33,508 --> 00:30:37,400 take on other, do other things. Yeah, I would say that, you 496 00:30:37,400 --> 00:30:39,500 know, in some ways, it's the end of an era, right? 497 00:30:39,700 --> 00:30:46,700 And, you know, I would be lying if I said cuando was not one of 498 00:30:46,708 --> 00:30:48,800 the things that inspired me to get into this because it was 499 00:30:48,800 --> 00:30:52,600 such an interesting idea. And frankly, it did seem like 500 00:30:52,600 --> 00:30:56,600 Like, you know, something that would be phenomenal and 501 00:30:56,600 --> 00:31:00,900 explosive and it was, but now that we are sort of, you know, a 502 00:31:00,908 --> 00:31:07,400 couple of years in, I, I'm, I would say that while it sits at 503 00:31:07,400 --> 00:31:09,200 the end of the day. This is a zero-sum game, right? 504 00:31:09,200 --> 00:31:13,800 So there are people who have learned, you know, through 505 00:31:14,000 --> 00:31:17,800 through this Red Queen Race of, you know, Finding better faster, 506 00:31:19,300 --> 00:31:21,600 better quality data and analyzing it better there have 507 00:31:21,700 --> 00:31:25,800 emerged somewhere. But the majority are losers and 508 00:31:25,800 --> 00:31:29,500 there's no two ways about it. And, you know, and so I think 509 00:31:29,500 --> 00:31:32,000 that's kind of it's gotten. It's gotten to a point where 510 00:31:32,100 --> 00:31:35,000 people understand the market and this is steady state to the 511 00:31:35,000 --> 00:31:39,300 world of all data. And, you know, and and, and sort 512 00:31:39,300 --> 00:31:41,400 of that's where it is. Okay. 513 00:31:42,100 --> 00:31:43,400 Okay. Now you bring me to another 514 00:31:43,400 --> 00:31:45,100 question. So you while talking about a, I 515 00:31:45,100 --> 00:31:48,000 know you were talking about, like, you mentioned this court 516 00:31:48,000 --> 00:31:52,000 that rules Plus data is like, whatever software engineering, 517 00:31:52,200 --> 00:31:55,700 but But Discerning the rules Based on data and outcomes that 518 00:31:55,700 --> 00:31:58,900 say I now. So now let me ask you what is 519 00:31:58,900 --> 00:32:01,800 the difference between AI + ml? This is again, something that a 520 00:32:02,500 --> 00:32:03,600 lot of people conflate. So. 521 00:32:03,600 --> 00:32:09,300 So so so frankly in today's day and age, it's like asking, 522 00:32:09,300 --> 00:32:13,300 what's the difference between let's say crypto and you know, 523 00:32:13,300 --> 00:32:15,100 cryptocurrency, right? Okay. 524 00:32:15,100 --> 00:32:18,900 Yeah, to go back to the analogy. So frankly, they've been used 525 00:32:18,900 --> 00:32:23,700 interchangeably today and you know, I Sort of come up, try and 526 00:32:23,700 --> 00:32:26,900 come up with a definition that, you know, some Venn diagram, 527 00:32:26,900 --> 00:32:29,000 which says okay, you know, these are all the things that belong 528 00:32:29,000 --> 00:32:30,600 to the world of machine learning but not AI. 529 00:32:30,900 --> 00:32:33,700 Yeah, but frankly, I think this is a, this is not a very useful 530 00:32:33,700 --> 00:32:36,700 discussion to have the under point and we have gotten to a 531 00:32:36,700 --> 00:32:41,200 place where a lot of these words are being used. 532 00:32:41,200 --> 00:32:45,800 So interchangeably. Yeah, that while they may have 533 00:32:45,800 --> 00:32:49,700 had distinct meanings five years ago, they no longer do so. 534 00:32:50,100 --> 00:32:53,900 Okay, right. So it It will be it will be sort 535 00:32:53,900 --> 00:32:56,400 of you know, let's say a very theoretical argument at this 536 00:32:56,400 --> 00:33:01,800 point in time to try and create distinction between these words. 537 00:33:02,400 --> 00:33:04,900 You say it's pretty much. It's your interpretation. 538 00:33:04,900 --> 00:33:06,700 It doesn't matter. What is what? 539 00:33:06,700 --> 00:33:08,800 Yeah, I mean you can you can you can look up Wikipedia and you'll 540 00:33:08,800 --> 00:33:11,500 get some very distinct some more somewhat distinct definitions, 541 00:33:11,500 --> 00:33:14,800 but frankly people are using these words so interchangeably 542 00:33:14,800 --> 00:33:18,400 today that it doesn't, it doesn't matter, right? 543 00:33:18,400 --> 00:33:21,600 Like so for instance, you know, it brings back that idea that, 544 00:33:21,600 --> 00:33:24,000 you know, if it's in Python. It's machine learning. 545 00:33:24,000 --> 00:33:26,600 If it's in PowerPoint. It's AI stuff like that. 546 00:33:26,600 --> 00:33:28,400 Right? So we have that world today. 547 00:33:28,600 --> 00:33:31,500 Yeah, right. So yeah. 548 00:33:32,700 --> 00:33:35,300 Okay. So I again want to bring you 549 00:33:35,300 --> 00:33:38,200 back to another thing that you had maintained off hand in the 550 00:33:38,200 --> 00:33:42,600 middle and then like we sort of. So you mentioned about how for 551 00:33:42,600 --> 00:33:45,800 example that your company and you philosophically opposed 552 00:33:45,800 --> 00:33:49,000 philosophically look at sort of data science as a sort of a 553 00:33:49,200 --> 00:33:51,900 toolbox kind of a thing where we right where you were. 554 00:33:52,500 --> 00:33:55,200 So how is it generally in finance nowadays? 555 00:33:55,200 --> 00:33:57,900 As a luckier? I know that all all big Banks 556 00:33:57,900 --> 00:34:00,400 now have massive data science teams. 557 00:34:00,700 --> 00:34:03,900 They are, they, they employ lots of machine learning people and 558 00:34:03,900 --> 00:34:07,000 things like that. Is it, is it generally? 559 00:34:07,200 --> 00:34:09,800 But one thing I've noticed in the industries that like 560 00:34:09,800 --> 00:34:12,100 especially among people who call themselves data scientists. 561 00:34:12,300 --> 00:34:16,000 Is that like they tend to be more of the hammered kind of 562 00:34:16,000 --> 00:34:17,900 people rather than their toolbox kind of people. 563 00:34:17,900 --> 00:34:19,400 But what is it? What do you notice in your 564 00:34:19,400 --> 00:34:21,300 industry as in like, how is it now? 565 00:34:21,300 --> 00:34:23,800 There is Italy if you. Put your competition clients and 566 00:34:23,800 --> 00:34:29,500 so on sure. So let's start by saying that 567 00:34:30,100 --> 00:34:35,800 typically when a large company and established firm sets up a 568 00:34:35,800 --> 00:34:41,400 data science team, right? The objective is not to solve a 569 00:34:41,400 --> 00:34:44,100 problem per se. The objective is to have a data 570 00:34:44,100 --> 00:34:48,900 science team, fair enough. Yeah, and you could say the same 571 00:34:48,900 --> 00:34:53,000 thing for, you know, typically, when you could extend the same 572 00:34:53,000 --> 00:34:56,000 logic to let you know, if I have a large established company is 573 00:34:56,000 --> 00:34:58,100 setting up a Bitcoin trading desk. 574 00:34:58,200 --> 00:35:00,400 Yeah, then your objective is not to make money. 575 00:35:00,400 --> 00:35:02,200 It is to set up a Bitcoin trading desk. 576 00:35:02,200 --> 00:35:07,600 Yep. 8 it is to service some customer need or some customer 577 00:35:07,600 --> 00:35:12,000 asked saying, hey. You provide me Bitcoin Services. 578 00:35:12,000 --> 00:35:15,200 Is that something that you do and you want to be able to say 579 00:35:15,200 --> 00:35:18,800 yes, yep, hate. And likewise. 580 00:35:18,800 --> 00:35:21,800 I think so, in many sense, in many cases. 581 00:35:21,800 --> 00:35:26,300 I think, at least when look, I have, I'm not sort of in the, I 582 00:35:26,308 --> 00:35:30,400 have not been on the, in the sell side for about, you know, 583 00:35:30,400 --> 00:35:34,200 close to four years now. Yeah, so all my opinions need to 584 00:35:34,200 --> 00:35:38,600 be sort of treated with a big spoonful of salt is not a pinch 585 00:35:38,600 --> 00:35:41,900 of salt, right? but I think now we are at a 586 00:35:41,908 --> 00:35:49,800 place where There are dedicated data science teams, perhaps 587 00:35:49,800 --> 00:35:52,600 working on problems of Interest. It could range anywhere. 588 00:35:52,700 --> 00:35:55,800 I think the one of the core problems of interest for all 589 00:35:56,800 --> 00:36:01,800 firms, irrespective of financial services are not is are they 590 00:36:01,800 --> 00:36:09,100 doing a good job of resourcing? Their customers with the right 591 00:36:09,100 --> 00:36:10,900 sales. People with the right? 592 00:36:12,000 --> 00:36:13,500 Let's call it. Research with the right 593 00:36:13,500 --> 00:36:18,800 resources and This is something that I would say is an 594 00:36:18,800 --> 00:36:24,000 interesting scale, data science problem, which I am sure most 595 00:36:24,000 --> 00:36:26,000 Financial Services firms are already working on. 596 00:36:26,200 --> 00:36:27,300 Yep. For instance. 597 00:36:27,400 --> 00:36:29,700 Let's say, you know, I have a customer. 598 00:36:29,700 --> 00:36:32,700 I know who they sales coverage is. 599 00:36:32,700 --> 00:36:35,000 I know the kind of questions, they ask me all the time. 600 00:36:35,300 --> 00:36:37,300 I know the kind of Trades they have done over the last ten 601 00:36:37,300 --> 00:36:41,400 years. I have a general area of our 602 00:36:41,400 --> 00:36:43,800 general knowledge understanding of what their areas of Interest 603 00:36:43,800 --> 00:36:45,300 are and today. Let's see. 604 00:36:45,400 --> 00:36:50,400 The customer who I make two million dollars a year from now 605 00:36:50,700 --> 00:36:54,100 an interesting data science problem could be what are the 606 00:36:54,100 --> 00:36:57,800 five things that we could do or let's say we have these limited 607 00:36:57,800 --> 00:37:00,500 set of resources. How do we allocate them? 608 00:37:00,500 --> 00:37:04,200 It could be, you know, time that my research person spends with 609 00:37:04,200 --> 00:37:06,000 this customer. It could be, no. 610 00:37:06,000 --> 00:37:09,600 I have built this IP or this toolkit, or this toolbox, which 611 00:37:09,600 --> 00:37:12,900 helps understand X. I have this webinar which has, 612 00:37:13,200 --> 00:37:15,300 which I can only give out 50 exclusives. 613 00:37:15,500 --> 00:37:18,800 It's yep. Right home should who should be 614 00:37:19,000 --> 00:37:21,800 given the right to sit in this webinar. 615 00:37:22,000 --> 00:37:23,400 Yep. These are all I would say 616 00:37:23,400 --> 00:37:31,300 optimization of effort of resources to maximize Revenue 617 00:37:32,300 --> 00:37:33,800 that you get from your customers. 618 00:37:34,100 --> 00:37:36,200 I think this is a fairly Universal problem. 619 00:37:36,600 --> 00:37:41,500 It's a fan, it's a problem that is Broad large wide enough and 620 00:37:41,500 --> 00:37:45,200 has a sufficient impact on bottom line. 621 00:37:45,400 --> 00:37:49,500 That is something that I'm sure every data science team is every 622 00:37:49,500 --> 00:37:51,900 bank has a data science team of some sort or someone the data 623 00:37:51,900 --> 00:37:53,300 science skills working on this problem. 624 00:37:53,400 --> 00:37:54,300 Yep. Yep. 625 00:37:54,700 --> 00:37:56,800 Now let's drill down massively. Right? 626 00:37:56,800 --> 00:38:03,500 Like for example, we had let's say model which let's say was 627 00:38:03,500 --> 00:38:05,800 used for option pricing. Okay, great. 628 00:38:05,800 --> 00:38:10,200 Yep, and you know and at this point we've moved, you know, 629 00:38:10,600 --> 00:38:13,500 five six Generations down from Lexia black-scholes model, 630 00:38:13,500 --> 00:38:14,100 right? Yep. 631 00:38:15,500 --> 00:38:17,900 You've got, let's say, two Factor model. 632 00:38:17,900 --> 00:38:20,300 You've got some Sable. You've got a bunch of other 633 00:38:21,100 --> 00:38:24,400 ideas in there. And now the question is that, 634 00:38:24,400 --> 00:38:28,300 you know, hey, can we sort of? Throw away these models. 635 00:38:28,700 --> 00:38:33,900 And we have enough, you know, data out there of underlying 636 00:38:33,900 --> 00:38:41,200 parameters and of, you know, option prices on the market that 637 00:38:41,400 --> 00:38:43,500 can we replace this with a machine learning model. 638 00:38:43,800 --> 00:38:47,000 However, if we have Machinery at least like a statistical model 639 00:38:47,000 --> 00:38:51,500 of some sort, right, so I'm sure this work is already also being 640 00:38:51,500 --> 00:38:52,900 undertaken. I'm sure that this is also 641 00:38:52,900 --> 00:38:58,800 something that most people are have done something about I had 642 00:38:59,000 --> 00:39:00,900 you know, it may have worked in some situations. 643 00:39:00,900 --> 00:39:03,200 It may not have worked in certain other situations, but 644 00:39:03,200 --> 00:39:06,600 this is probably something that also people have worked on. 645 00:39:06,900 --> 00:39:11,800 So these are all probably instances where, you know, let's 646 00:39:11,800 --> 00:39:17,200 say there is a data science team or someone would dedicated data 647 00:39:17,200 --> 00:39:21,200 science, background thinking about these problems are working 648 00:39:21,200 --> 00:39:22,200 on them. Yeah. 649 00:39:22,500 --> 00:39:27,700 Now, the bigger change I would say and You have sort of you 650 00:39:27,700 --> 00:39:30,500 know been on LinkedIn and looked up your Goldman colleagues. 651 00:39:30,500 --> 00:39:34,900 Who are still at Goldman. You will notice that data 652 00:39:34,900 --> 00:39:37,300 science or let's see machine learning has become mainstream. 653 00:39:37,300 --> 00:39:41,000 Most people have done some Coursera course or some have 654 00:39:41,000 --> 00:39:43,700 threatened to have played around or our weekend. 655 00:39:43,800 --> 00:39:46,400 Let's say machine learning folks, right? 656 00:39:46,400 --> 00:39:47,100 Yep. Yep. 657 00:39:48,000 --> 00:39:53,700 And so that I think is the most significant change where people 658 00:39:53,700 --> 00:39:58,900 are given sufficient. Amount of familiarity with this 659 00:39:58,900 --> 00:40:03,000 toolkit that they're able to incorporate it into their 660 00:40:03,000 --> 00:40:08,000 everyday sort of work, and there is a strong interest need. 661 00:40:09,800 --> 00:40:16,200 I would say, you know, compulsion almost to sort of get 662 00:40:16,200 --> 00:40:21,900 up to speed these kind of ideas and to find situations at work 663 00:40:21,900 --> 00:40:24,000 where these can be applied and that's what leads to the sort of 664 00:40:24,200 --> 00:40:26,700 the hammer problem that you spoke about. 665 00:40:26,900 --> 00:40:28,600 Let's say they're someone's day job. 666 00:40:28,600 --> 00:40:31,500 Does not really require them to apply machine learning but they 667 00:40:31,500 --> 00:40:33,700 are still sort of saying, you know, what? 668 00:40:34,300 --> 00:40:35,800 I've learned about this new thing. 669 00:40:35,800 --> 00:40:39,300 I've got some awareness and familiarity with it, you know, 670 00:40:39,300 --> 00:40:41,500 can I find a way to demonstrate that? 671 00:40:41,500 --> 00:40:44,300 I know this is well at work. Yeah. 672 00:40:47,500 --> 00:40:53,300 I'll sort of. close all this by saying something else, which is 673 00:40:56,000 --> 00:40:59,800 Going back to what we spoke about I mentioned earlier, which 674 00:40:59,800 --> 00:41:02,800 is that Wall Street or the financial services industry as 675 00:41:02,800 --> 00:41:07,800 such the established players are not first-movers. 676 00:41:07,800 --> 00:41:10,400 They are good s mobiles. Yep. 677 00:41:10,400 --> 00:41:13,000 They do good for followers fast followers. 678 00:41:13,000 --> 00:41:17,900 I suppose is the word. So when something becomes 679 00:41:17,900 --> 00:41:22,500 established, you will see that they're sort of very quick in 680 00:41:22,508 --> 00:41:26,400 adopting that. And that also means that You 681 00:41:26,400 --> 00:41:31,400 know, the the base at which things are getting adopted 682 00:41:31,400 --> 00:41:35,800 moving changing within large firms is relatively glacial 683 00:41:36,200 --> 00:41:39,200 versus what you would see inside. 684 00:41:39,400 --> 00:41:42,400 Let's see the startup world or the tech World in general, 685 00:41:42,900 --> 00:41:46,400 right? I mean, the last five years have 686 00:41:46,400 --> 00:41:51,600 been like mean the the rate of change in terms of the ability 687 00:41:51,600 --> 00:41:55,600 of what neural networks can deliver has been like nothing. 688 00:41:55,700 --> 00:41:59,800 A short of phenomenal. I mean, if I were favoritest I 689 00:41:59,800 --> 00:42:04,000 would use the word miraculous. Yeah, and but this is not 690 00:42:04,000 --> 00:42:08,300 something that lets say you could have ever come about 691 00:42:08,300 --> 00:42:12,800 within a large established player tinkering by themselves 692 00:42:12,800 --> 00:42:16,000 with this kind of Technology, right? 693 00:42:16,300 --> 00:42:20,500 I think the fact that A lot of these changes were happening in 694 00:42:20,500 --> 00:42:26,500 the open source community and where code or ideas were openly 695 00:42:26,500 --> 00:42:31,400 being exchanged by means of papers and archive or, you know, 696 00:42:31,400 --> 00:42:35,200 simply code available on GitHub. Everything has been is, no. 697 00:42:35,200 --> 00:42:37,900 I mean, I think that's been sort of the core contributors to 698 00:42:37,900 --> 00:42:41,200 where, you know, where we are today. 699 00:42:41,200 --> 00:42:45,300 And I think that is something, which is very hard to recreate 700 00:42:45,300 --> 00:42:49,300 within a particular industry, which sort of We don't values 701 00:42:49,600 --> 00:42:53,800 highly and secretive about their models or there. 702 00:42:54,100 --> 00:42:58,200 Yeah, you know, ideas, right? And frankly, out of necessity, 703 00:42:58,400 --> 00:43:00,700 right? Of course, and I think that is, 704 00:43:00,700 --> 00:43:04,100 though, that is the challenge here, right? 705 00:43:04,100 --> 00:43:08,700 So, just to summarize the very likely that every large form 706 00:43:08,700 --> 00:43:11,400 out, there has a data science team, who's working on some 707 00:43:11,700 --> 00:43:13,500 firm, white problems, very likely. 708 00:43:13,500 --> 00:43:17,500 That every individual Quantum is applying his kind of self-taught 709 00:43:17,500 --> 00:43:20,800 in data science. And applying some ideas and 710 00:43:20,800 --> 00:43:25,200 thirdly, you know, if you look across these two teams, the rate 711 00:43:25,200 --> 00:43:29,200 of change is perhaps or the the amount of let's see. 712 00:43:29,200 --> 00:43:32,000 New idea is and create new ideas. 713 00:43:32,000 --> 00:43:35,900 New development that's happening is probably not at the same Pace 714 00:43:35,900 --> 00:43:39,100 that we are seeing elsewhere, right? 715 00:43:39,900 --> 00:43:45,100 Because it is it is the kind of environment where which sort of 716 00:43:45,100 --> 00:43:50,000 encourages being of very good. Follower as opposed to a first 717 00:43:50,000 --> 00:43:51,500 mover. Is that a bit clearer? 718 00:43:51,500 --> 00:43:52,400 Right? I think that. 719 00:43:52,400 --> 00:43:55,700 Yeah, I mean, I think just two clicks plane, the last bit. 720 00:43:56,300 --> 00:44:02,600 It's a feature and not a bug. Yep, right because I think, the 721 00:44:02,600 --> 00:44:06,000 tech World operates, very differently as a very different 722 00:44:06,000 --> 00:44:08,400 appetite. For risk, has very different 723 00:44:08,400 --> 00:44:12,800 appetite for, you know, they're sort of that not star if you 724 00:44:12,808 --> 00:44:15,300 will, in terms of what should and should be done, is quite 725 00:44:15,300 --> 00:44:17,600 different from the way the financial industry operates. 726 00:44:17,700 --> 00:44:20,400 Yeah, right. We're in a world where, you 727 00:44:20,400 --> 00:44:23,100 know, it's it's not a shoot. First. 728 00:44:23,100 --> 00:44:28,300 Ask questions, later World. You have Regulators all over the 729 00:44:28,300 --> 00:44:32,600 place. You have reputational risk. 730 00:44:32,800 --> 00:44:36,400 Your typical customer is someone you know, who's looking to 731 00:44:36,400 --> 00:44:41,200 conserve wealth first before creating wealth. 732 00:44:41,300 --> 00:44:45,500 Yeah State don't lose my money. That's the first rule, right? 733 00:44:46,300 --> 00:44:50,400 So it's a very different. Industry with a very different 734 00:44:50,500 --> 00:44:53,600 sort of value system. And therefore, I think that is 735 00:44:53,600 --> 00:44:59,600 what sort of, you know, creates this very cautious approach 736 00:45:00,000 --> 00:45:02,700 towards risk-taking of all forms. 737 00:45:03,000 --> 00:45:08,100 Yeah, I mean B8 B Financial Risk taking but also technology 738 00:45:08,100 --> 00:45:11,500 technological risk, right? In terms of, you know, adopting 739 00:45:11,500 --> 00:45:15,800 a new system or you know, migrating to a new framework. 740 00:45:15,800 --> 00:45:18,200 All of these things have to be done very prudently. 741 00:45:18,600 --> 00:45:23,100 Okay, and so by, and if they don't sort of one way in which, 742 00:45:23,100 --> 00:45:26,800 I think a lot of firms they say from the Google Cisco, probably 743 00:45:26,800 --> 00:45:29,000 have grown. Is that like they may not 744 00:45:29,100 --> 00:45:31,700 innovate so much internally, but as soon as they see somebody 745 00:45:31,700 --> 00:45:34,500 outside who's innovating they quickly go gobble them up, 746 00:45:34,700 --> 00:45:37,500 right? So, do do for large Financial 747 00:45:37,500 --> 00:45:39,900 firms. Also, like have they also got a 748 00:45:39,908 --> 00:45:45,000 reputation for that, we hiding, or, is it just a sort of still 749 00:45:45,600 --> 00:45:49,900 the they try to replicate it, rather than sort of Xh, so, I 750 00:45:49,900 --> 00:45:52,300 would say that to answer this question. 751 00:45:52,300 --> 00:45:56,300 We have to go back to what is the core asset in this business? 752 00:45:56,300 --> 00:46:00,000 That generates value for a Google or a Facebook, or anyone 753 00:46:00,000 --> 00:46:04,300 else that core acid is typically you can argue brand and so on, 754 00:46:04,300 --> 00:46:05,900 but it's IP. Yeah. 755 00:46:06,000 --> 00:46:08,300 Yep. It's basically being able to do 756 00:46:08,300 --> 00:46:12,100 something quicker faster better than someone else being first to 757 00:46:12,100 --> 00:46:17,200 Market, capturing the customer before with a better experience 758 00:46:17,200 --> 00:46:18,600 before anyone else can. At. 759 00:46:19,100 --> 00:46:22,800 Yeah, in the financial services industry. 760 00:46:23,600 --> 00:46:28,800 It is not so much your IP or technological Edge, but it is 761 00:46:28,800 --> 00:46:33,200 the ability to day after day year after year, deliver 762 00:46:33,200 --> 00:46:38,400 consistent, good quality service to your best customers to be 763 00:46:38,400 --> 00:46:41,200 able to sort of it is it is basically, you know, being 764 00:46:41,200 --> 00:46:45,300 reliable. It's a big being sort of trust, 765 00:46:45,300 --> 00:46:48,300 is a big, big element of the traditional banking industry. 766 00:46:48,800 --> 00:46:54,000 Sort of a something that, you know, also is something that's a 767 00:46:54,000 --> 00:46:56,100 value that that exists, even today with the financial 768 00:46:56,100 --> 00:46:59,000 services industry. And it's, and, and now that's 769 00:46:59,000 --> 00:47:02,200 something which is very difficult to earn and very easy 770 00:47:02,200 --> 00:47:06,700 to lose. Therefore, while your customers 771 00:47:06,700 --> 00:47:11,600 might be asking you questions. Like, you know, hey if I want to 772 00:47:11,600 --> 00:47:13,500 trade Bitcoin, can I do that with you? 773 00:47:14,600 --> 00:47:18,100 They're not asking you. Are you sort of using the most? 774 00:47:18,400 --> 00:47:23,300 Latest most sophisticated, you know, machine learning Tech and 775 00:47:23,300 --> 00:47:27,400 more than your customers. That's not what the shareholders 776 00:47:27,400 --> 00:47:30,000 are looking for. Shareholdings shareholders, are 777 00:47:30,000 --> 00:47:32,500 looking for you to beat earnings estimates. 778 00:47:32,900 --> 00:47:36,500 The shareholders are looking for you not to get find you not to 779 00:47:36,508 --> 00:47:41,600 get sued. They are not looking for you to 780 00:47:41,600 --> 00:47:43,700 have the latest and the greatest IP. 781 00:47:44,200 --> 00:47:48,300 Yep. Therefore, the expectations from 782 00:47:48,500 --> 00:47:52,500 Wall Street Forum are vastly different from the expectations 783 00:47:52,900 --> 00:47:56,000 that are on the shoulders of a valley company. 784 00:48:17,800 --> 00:48:19,900 Thank you for listening to data shatter. 785 00:48:20,400 --> 00:48:24,000 If you like this show, please leave a comment, share and 786 00:48:24,000 --> 00:48:27,200 subscribe to the podcast. You can find this podcast on 787 00:48:27,200 --> 00:48:30,600 Apple podcasts Spotify or wherever else you go. 788 00:48:30,700 --> 00:48:34,100 To get your podcasts. Once again, this is Karthik 789 00:48:34,100 --> 00:48:35,700 signing off. Thank you.