1 00:00:00,080 --> 00:00:03,840 In this 349th episode of data driven, we are pleased 2 00:00:03,840 --> 00:00:07,460 to interview Pavel Goldman Khaledin, where he's the head of artificial 3 00:00:07,520 --> 00:00:09,974 intelligence and machine learning at Sumsub. 4 00:00:10,995 --> 00:00:14,594 Sumsub isn't your average AI startup. They're 5 00:00:14,594 --> 00:00:18,130 globally recognized for their work in k y c, AML, 6 00:00:18,269 --> 00:00:21,869 and anti fraud technologies. Our guest is the 7 00:00:21,869 --> 00:00:25,470 wizard behind the curtain, crafting tech to outsmart financial 8 00:00:25,470 --> 00:00:29,075 fraud does and deep fake artists. Quite the 9 00:00:29,075 --> 00:00:32,835 digital Sherlock Holmes, if you will. Now here are 10 00:00:32,835 --> 00:00:34,750 Frank, Andy, and Pavel. 11 00:00:39,129 --> 00:00:42,829 Hello, and welcome to Data Driven, the podcast where we explore the emergent 12 00:00:43,295 --> 00:00:47,055 Fields of data science, artificial intelligence, and, 13 00:00:47,055 --> 00:00:50,895 of course, data engineering, which is basically the underpinning of it 14 00:00:50,895 --> 00:00:54,500 all. And with me on this, journey is my favorite data 15 00:00:54,500 --> 00:00:58,340 engineer of them all, Andy Leonard. How's it going, Andy? Good, Frank. 16 00:00:58,340 --> 00:01:02,055 How are you? I'm doing alright. We we were recording this, the day 17 00:01:02,055 --> 00:01:04,475 after we did a 2 hour show, 18 00:01:07,020 --> 00:01:10,780 Kinda by accident, don't I see our guest in, it look kinda had this 19 00:01:10,780 --> 00:01:14,160 look of, uh-oh. No. It's not gonna turn. I can't do that today. 20 00:01:16,104 --> 00:01:19,625 But we are very excited here to in spite of our issues with Microsoft 21 00:01:19,625 --> 00:01:23,420 Bookings, in spite of our crazy hectic schedules, And in 22 00:01:23,420 --> 00:01:27,140 spite of your allergies and, really tasty jelly jam and 23 00:01:27,140 --> 00:01:30,740 and and biscuits Really sorry about that. No. 24 00:01:30,740 --> 00:01:34,475 I I don't know what it is on the East Coast this week, man. It's 25 00:01:34,475 --> 00:01:37,534 it's well below freezing, and I'm sneezing. Oh, that rhymed. 26 00:01:38,075 --> 00:01:41,034 Allergy station should be over for me. I don't know what's going on. For real. 27 00:01:42,220 --> 00:01:45,680 But our guest is actually, from Berlin, 28 00:01:46,620 --> 00:01:49,740 and one of my favorite cities in the world. In fact, they were singing the 29 00:01:49,740 --> 00:01:53,505 virtual green room. Had I lived in Berlin instead of Frankfurt, I probably 30 00:01:53,505 --> 00:01:55,284 never would have come back to New York, 31 00:01:57,265 --> 00:02:00,179 or the US, but he is 32 00:02:01,380 --> 00:02:05,060 our guest today is Pavel Goldman Kaledin. Hopefully, I said that 33 00:02:05,060 --> 00:02:08,625 right. He is the head of AI and ML 34 00:02:08,685 --> 00:02:12,385 at Sumsub, a global know your customer anti 35 00:02:12,525 --> 00:02:16,050 money laundering, anti fraud company, and, 36 00:02:16,530 --> 00:02:19,490 we're we're welcome to we're happy to have him. Although, I don't think he's in 37 00:02:19,490 --> 00:02:23,190 Berlin today. I think he's somewhere a bit warmer. Welcome to the show, Pavel. 38 00:02:23,925 --> 00:02:27,765 Yeah. Hi, guys. Happy to be here. Good. Good. So I have 39 00:02:27,765 --> 00:02:31,500 a lot of questions. You know, 40 00:02:31,500 --> 00:02:32,240 first off, 41 00:02:35,420 --> 00:02:38,944 I think I can kinda see the map, but What's the 42 00:02:38,944 --> 00:02:42,165 connection between know your customer, KYC, 43 00:02:44,305 --> 00:02:48,150 anti money laundering, and anti fraud? I think I think 44 00:02:48,150 --> 00:02:51,510 I see it, but I wanna hear you you kinda walk me through it because 45 00:02:51,510 --> 00:02:55,350 I haven't had enough coffee either today. So so what's the, like, what's 46 00:02:55,350 --> 00:02:58,845 the common thread? Because, like, because I I've not seen those 3 47 00:02:58,845 --> 00:03:01,965 kinda put together in kinda 1, 48 00:03:02,365 --> 00:03:06,200 sentence, but I can kinda see why. But 49 00:03:06,200 --> 00:03:09,800 I I I I can try to explain. But the thing is and we actually 50 00:03:10,440 --> 00:03:13,980 this is what we focus on. So we try to secure as a company. 51 00:03:14,235 --> 00:03:17,935 We try to secure the whole customer journey from 52 00:03:18,315 --> 00:03:21,754 onboarding. So this is the first step of when, for instance, like, I'm in a 53 00:03:21,754 --> 00:03:24,700 bank. So So I want to onboard some of my customers, and I want to 54 00:03:24,700 --> 00:03:28,480 make sure that this has real persons, for instance, that are not fraudsters. 55 00:03:28,700 --> 00:03:31,280 So I want to onboard them, make sure they are, 56 00:03:32,955 --> 00:03:36,635 that person, they actually pretend to be. And then and 57 00:03:36,635 --> 00:03:40,390 here's the thing. If I can, for instance, like, I'm a Journey 58 00:03:40,769 --> 00:03:44,610 person. But a month later. There could 59 00:03:44,610 --> 00:03:48,015 be some, you know, strange patterns of, you know, 60 00:03:48,895 --> 00:03:52,575 financial transaction happening. So probably, there are some sort of a pattern of 61 00:03:52,575 --> 00:03:56,380 money laundering. So this is where transaction monitoring comes. 62 00:03:56,460 --> 00:03:59,580 So you can actually this is a person. So this is but knowing customers are 63 00:03:59,580 --> 00:04:03,180 very simple. You can actually I mean, you 64 00:04:03,180 --> 00:04:06,905 can So basic basic attack is to be just pretend to be, 65 00:04:07,125 --> 00:04:10,885 a person. You you are not, basically. But then even if I'm 66 00:04:10,885 --> 00:04:14,230 not, I'm just a real person, I can actually, yeah, come up with some sort 67 00:04:14,230 --> 00:04:17,829 of, you know, few things to 68 00:04:17,829 --> 00:04:21,430 do. And then where just we try to monitor it, and then from a permit, 69 00:04:21,430 --> 00:04:25,205 make sure that, Okay. We can actually flag the transaction and 70 00:04:25,205 --> 00:04:29,045 then make sure it's it's it's getting looped. And then, I mean, there is a 71 00:04:29,045 --> 00:04:32,669 flag raised, and then, Probably, we can do 72 00:04:32,669 --> 00:04:36,030 something about that. This is just, like 73 00:04:36,030 --> 00:04:39,605 this. If we're talking about anti fraud, and here's the 74 00:04:39,605 --> 00:04:43,365 thing. Sometimes it's very easy to see that something fish is 75 00:04:43,365 --> 00:04:47,050 happening. So for instance, like, A very like, 2 years ago, it 76 00:04:47,050 --> 00:04:50,810 was a very typical attack. So I tried to, you know, open a bank 77 00:04:50,810 --> 00:04:54,345 account or, like, remotely, And I actually, I'll leave somewhere 78 00:04:54,345 --> 00:04:58,125 else, or I don't I I use a stolen document. What 79 00:04:58,425 --> 00:05:02,200 what I can do To do that, I can actually just print out the 80 00:05:02,200 --> 00:05:05,960 image of a person and just try to make sure that actually the 81 00:05:05,960 --> 00:05:09,574 KFC provider like us Tried to make us 82 00:05:09,574 --> 00:05:13,254 believe that I'm a real person. That was a very, you know, typical attack 2 83 00:05:13,254 --> 00:05:16,935 years ago. Now it's very easy to detect. Still peep some people use 84 00:05:16,935 --> 00:05:20,229 it. And that's it. And that's that for us. It is very easy to do 85 00:05:20,229 --> 00:05:23,750 that. But probably, I mean, this is not a real person. Some of you trying 86 00:05:23,750 --> 00:05:27,050 to use the printed out images. This is 87 00:05:27,455 --> 00:05:31,295 Fraud. We can actually or reject it or or ask a person. Can 88 00:05:31,295 --> 00:05:34,870 you well, I mean, we need your real real pay real real image. 89 00:05:35,190 --> 00:05:38,949 Or we can just tell our customers that, you have to take a look because 90 00:05:38,949 --> 00:05:42,710 there was something fishy going. And then it goes and goes and goes. And the 91 00:05:42,710 --> 00:05:46,505 whole customer journey, We try to make sure that the fraud is not happening. This 92 00:05:46,505 --> 00:05:49,645 is basically it. So 93 00:05:50,264 --> 00:05:53,960 fraud is kind of, I think, Cyber fraud or whatever the cool 94 00:05:53,960 --> 00:05:57,340 kids call it, I think is has has infected 95 00:05:57,480 --> 00:06:01,205 every industry. I mean, if I just I 96 00:06:01,205 --> 00:06:04,645 mean, I I get 2 factor authentication logging in the 97 00:06:04,645 --> 00:06:08,300 roadblocks, like, for my kids. Right. And I'm like, 98 00:06:08,300 --> 00:06:12,060 they'll they'll they'll they'll get in front of their device, and they'll be like, can 99 00:06:12,060 --> 00:06:15,775 you tell me what the passcode is that they texted you? Like, Sometimes 100 00:06:15,854 --> 00:06:19,155 some days it's the only way I see 1 of my kids. But, 101 00:06:21,694 --> 00:06:25,460 has the because I I wonder, like, has the pandemic kind of Accelerated 102 00:06:25,520 --> 00:06:28,980 kind of virtual fraud, or is that just independent? 103 00:06:29,680 --> 00:06:33,380 I think it I think it is. Because it, right now, it's but it's not 104 00:06:33,664 --> 00:06:37,125 Related to fraud. Exactly. But the thing is is that now 105 00:06:37,345 --> 00:06:40,405 people are used to actually work 106 00:06:40,784 --> 00:06:44,630 remotely, Or it's so it's not that common for you 107 00:06:44,630 --> 00:06:48,310 to go to bank in person. So you just call there. You just I 108 00:06:48,310 --> 00:06:51,895 mean, use over the internet, basically. It's like easier 109 00:06:52,375 --> 00:06:56,074 So and now you can actually, there is no way, you can actually verify 110 00:06:56,134 --> 00:06:59,655 that this is the only person. Right. Yep. And this is a final thing because 111 00:06:59,655 --> 00:07:03,320 for instance, in Germany, where I reside, most of the 112 00:07:03,320 --> 00:07:07,080 time. There is a regulation called it's called 113 00:07:07,080 --> 00:07:10,845 video ident. So for in Germany, in order For for 114 00:07:10,845 --> 00:07:14,605 me, if you are going to open an account, anyway, I really have to call 115 00:07:14,605 --> 00:07:18,420 a person, a live in person operator, And talk to him, and he makes 116 00:07:18,420 --> 00:07:22,200 sure that or she makes sure that, a a million person. But everybody 117 00:07:22,420 --> 00:07:25,620 do not like it, basically. Because, I mean, it it takes time. You have to 118 00:07:25,620 --> 00:07:28,805 talk, talk to a person. I I just want to open an account. So it's 119 00:07:28,805 --> 00:07:32,605 it's it's it's fast as I'm but but except Germany, all of the rest 120 00:07:32,605 --> 00:07:36,045 of European Union, I think across the world as well. It's, I mean, you 121 00:07:36,045 --> 00:07:39,600 just Send your image or video, some of your documents, 122 00:07:39,740 --> 00:07:42,940 and then the the account is up. So it's very easy. And people get you, 123 00:07:43,100 --> 00:07:46,905 getting used to it. And that's why it's easier to to to actually, 124 00:07:48,165 --> 00:07:51,890 do fraud because it's, I mean, it's it's a soldier to trade off, 125 00:07:52,050 --> 00:07:55,650 Make it easier, and then it's easier for fraudsters to actually do their business. So 126 00:07:55,650 --> 00:07:59,330 that's that's the thing. Gotcha. Do you see, 127 00:07:59,650 --> 00:08:03,405 you mentioned you see, Like, new scams, people 128 00:08:03,405 --> 00:08:07,245 are running as well. And you also mentioned a lot 129 00:08:07,245 --> 00:08:10,620 of what I I thought would be pretty effective ways to to 130 00:08:10,620 --> 00:08:14,159 combat those scams, without really 131 00:08:14,620 --> 00:08:18,465 giving anybody any ideas. Are there, like, brand new 132 00:08:18,465 --> 00:08:22,245 scams that have happened maybe in in the very recent past 133 00:08:22,705 --> 00:08:25,525 that, you're still working on ways to combat? 134 00:08:27,310 --> 00:08:31,150 I must say that, there is there will always 135 00:08:31,150 --> 00:08:34,725 be some sort of, you know, arms, right. 136 00:08:34,725 --> 00:08:38,565 Competition? Yeah. So you have to say or. There will always be, 137 00:08:38,565 --> 00:08:42,260 like, a new prod Of yours. 138 00:08:42,260 --> 00:08:45,060 And then we have to actually deal with that. But I can tell you a 139 00:08:45,060 --> 00:08:48,420 story. So for instance, like, so we asked him so not a big company. Yeah. 140 00:08:48,420 --> 00:08:51,755 The technology team is not that So big, we have to move fast. But 141 00:08:52,055 --> 00:08:55,355 in my team, the AI slash, ML, it's not 142 00:08:55,735 --> 00:08:59,510 anti money laundering, but artificial intelligence slash machine learning. We have 143 00:08:59,510 --> 00:09:03,290 a very small department aimed at creating defects. 144 00:09:03,990 --> 00:09:07,510 So we do not detect defects. We have to actually learn how to create 145 00:09:07,510 --> 00:09:11,355 them So you actually know how I mean, how people actually read Oh, 146 00:09:11,815 --> 00:09:15,595 that makes sense. So synthetic data. Interesting. Yeah. Yeah. 147 00:09:15,710 --> 00:09:18,830 And this is at and I can also tell you that I mean, and this 148 00:09:18,830 --> 00:09:21,810 is for me, it was, like, so sorry if, you know, a surprise because, 149 00:09:23,075 --> 00:09:26,755 Most of the like, let's talk about defects. So, yes, then what what is like 150 00:09:26,755 --> 00:09:30,595 recent type of fraud? Deepest, for sure. We had a report. I 151 00:09:30,595 --> 00:09:34,329 I think it, We published it 3 years 2 days ago or like 152 00:09:34,329 --> 00:09:38,089 yesterday on friends. So what's actually happening right now? 153 00:09:38,089 --> 00:09:41,475 And the thing is that deep fakes, They use usage of 154 00:09:41,475 --> 00:09:45,154 defects for fraud. It maybe it rest 155 00:09:45,154 --> 00:09:48,535 like 5 times. So like 2 years ago, like nobody actually 156 00:09:48,915 --> 00:09:52,630 knew so About defects. But now it's it's very easy to craft. It's 157 00:09:52,630 --> 00:09:55,830 very easy to craft. I mean, people like I mean, you are a fraudster. You 158 00:09:55,830 --> 00:09:58,410 have to actually, it's very rare 159 00:09:59,335 --> 00:10:03,095 prefer for you to just craft just 1 defect. It's usually something 160 00:10:03,095 --> 00:10:06,695 we call the serial fraud. You create like hundreds of defects. So now it's easy, 161 00:10:06,695 --> 00:10:10,420 very easy to create them. So now it's like a craft, like, hundreds 162 00:10:10,480 --> 00:10:14,160 of identities. And then I tried to bypass our security checks. So that's why this 163 00:10:14,160 --> 00:10:17,775 is like the recent trend. I mean, as so it's on the news, 164 00:10:17,775 --> 00:10:21,155 basically. And then we have to actually try to make sure that our solution, 165 00:10:21,535 --> 00:10:25,279 can detect it. And it's not sometimes, it's not that easy. Well, 166 00:10:25,279 --> 00:10:29,120 it sounds like, you know, there's there's stuff that people used 167 00:10:29,120 --> 00:10:32,959 years ago, and you've got that figured out. And it's probably not being used 168 00:10:32,959 --> 00:10:36,305 as much, at least alone. But now you've got, 169 00:10:36,625 --> 00:10:40,225 people coming up with, first, new ideas, and then second, they're 170 00:10:40,225 --> 00:10:43,910 doing combinations new plus older ideas. Is that 171 00:10:43,910 --> 00:10:47,590 accurate? But but, it is actually. And the thing is Okay. So, 172 00:10:48,070 --> 00:10:51,525 these are also like, Okay. Just imagine. We have a very 173 00:10:51,905 --> 00:10:55,745 sophisticated deep fake detector. So I I'm pretty sure that our, like, 174 00:10:55,745 --> 00:10:59,220 models are more or less, good. So, like, 175 00:10:59,760 --> 00:11:03,600 I mean, it's not 100% for sure. Mhmm. But what 176 00:11:03,600 --> 00:11:07,165 happens next? So can I actually, I mean, combat defects, 5 177 00:11:07,165 --> 00:11:11,005 years later? Maybe it's I'm so advanced. I so make like, 178 00:11:11,005 --> 00:11:14,685 our customers, like, ask us about it, like, once in a 179 00:11:14,685 --> 00:11:18,510 month. So what do you actually what is your plan, to talk about defects 180 00:11:18,510 --> 00:11:22,110 in 2 years. Right. Because now, you know, AI is like, it's very hard problem 181 00:11:22,110 --> 00:11:25,865 to solve. But here's also problem. There is a thing 182 00:11:25,865 --> 00:11:29,465 called mules. Have you heard about mules or money 183 00:11:29,465 --> 00:11:33,260 mules? This is, the the thing is 184 00:11:33,260 --> 00:11:36,160 that you actually go, hire a person. 185 00:11:36,700 --> 00:11:40,335 Usually, buy, pay some €50. And then 186 00:11:40,335 --> 00:11:44,175 actually this person passes a KVST check for you. 187 00:11:44,175 --> 00:11:47,775 And then Oh, wow. The person just sells sells here his or her 188 00:11:47,775 --> 00:11:50,660 account to you. And then this is a real person. I mean, it's not a 189 00:11:50,660 --> 00:11:53,800 defect. I found it that I could defect. Wow. It's not obvious and not defect. 190 00:11:53,860 --> 00:11:56,745 Yeah. But that well, this is that looks suspicious. But 191 00:11:58,165 --> 00:12:01,765 but I if I'm in a bank, I'm in a I'm a bank, for me, 192 00:12:01,765 --> 00:12:05,340 it's like a real person just trying to open up in a bank account. Yeah. 193 00:12:05,340 --> 00:12:09,180 And now we actually have to look around. So that's why so I 194 00:12:09,180 --> 00:12:12,060 like working with Deepgrams. I mean, it's very, you know, cool technology. You have to, 195 00:12:12,060 --> 00:12:15,904 like Yeah. It's technology. But Now you actually have 196 00:12:15,904 --> 00:12:19,665 to look around. You have to make sure what is, I mean, the 197 00:12:19,665 --> 00:12:23,345 pattern. What are the devices do you use? It's like lots 198 00:12:23,345 --> 00:12:27,110 of small Features or, signals, you have to actually 199 00:12:27,410 --> 00:12:31,090 combine or merge them altogether and then make a decision. Is it, like, 200 00:12:31,090 --> 00:12:34,935 specia or suspicious sorta? And this is like, but this is fun. This is 201 00:12:34,935 --> 00:12:38,695 like, you have to really look around, look collect lots of data, and then try 202 00:12:38,695 --> 00:12:41,515 to find, you know, your way into making a decision. 203 00:12:43,140 --> 00:12:46,740 Interesting. It's it's it's a fascinating the simple things are no 204 00:12:46,740 --> 00:12:50,555 longer simple. Right? Just signing up for an account, You know, 205 00:12:51,255 --> 00:12:54,954 it's just now it's become like this massive multinational worldwide 206 00:12:55,334 --> 00:12:59,150 cyber Security kind of exercise. It's a 207 00:12:59,150 --> 00:13:02,590 fascinating, Yes. For a 208 00:13:02,590 --> 00:13:06,315 customer, it is it must remain easy. Yes. I don't know like 209 00:13:06,435 --> 00:13:10,035 I mean, since, like even, you know, the really, really 210 00:13:10,035 --> 00:13:13,635 typical KBC check is includes recording your 211 00:13:13,635 --> 00:13:17,440 video. You usually have to do something like, you know, turn your head 212 00:13:17,440 --> 00:13:21,120 or something. I mean, if you have this experience. People do not like it. For 213 00:13:21,120 --> 00:13:24,020 them, it's like, why do you have to do this? That's it's it looks strange. 214 00:13:24,194 --> 00:13:27,235 I mean, just can I just open an account? And then it's like so it's 215 00:13:27,235 --> 00:13:30,615 also trade off unless you have to be simultaneously 216 00:13:31,634 --> 00:13:35,250 secure and busy. And this is Yeah. Those those are 217 00:13:35,250 --> 00:13:39,090 those are very much contradictory, forces. Yeah. 218 00:13:39,090 --> 00:13:42,615 Well, the other thing too, like, if I'm if I'm If I'm an average 219 00:13:42,675 --> 00:13:46,435 customer or paranoid me. Right? Like, if I go to a 220 00:13:46,435 --> 00:13:49,910 thing and they want me to look this way, look that way, Am I training 221 00:13:49,910 --> 00:13:53,750 their deep fake model of me? Do you know what I mean? Like, I mean, 222 00:13:53,750 --> 00:13:57,590 I'm kinda like, you know, obviously, I've done a lot of live streams and stuff 223 00:13:57,590 --> 00:14:00,955 like that, so I shudder Better to think what you know, where that could lead. 224 00:14:00,955 --> 00:14:04,215 But, what are your thoughts on that? Like, I mean, are do do you have 225 00:14:04,215 --> 00:14:07,570 people who are Do savvy customers 226 00:14:08,270 --> 00:14:11,730 do they get a little suspicious? Like, 227 00:14:12,670 --> 00:14:16,475 what are your thoughts on I'm not. I I 228 00:14:16,475 --> 00:14:20,155 must said that I mean, the defects that we see, they they 229 00:14:20,155 --> 00:14:23,970 can be crafted just for 1 1 image. Right. So like, 230 00:14:23,970 --> 00:14:27,670 here's the problem. So so like, there are, none of that, I mean, 231 00:14:28,210 --> 00:14:31,834 you can see them, but Usually, people send, you know, 232 00:14:31,834 --> 00:14:35,595 low quality images. So it's even harder for us to see it. Even harder for 233 00:14:35,595 --> 00:14:38,495 for human person for human to see that this is a problem. 234 00:14:39,319 --> 00:14:42,839 But there is also, I think, if I find a story that I 235 00:14:42,839 --> 00:14:46,600 know, that some of our models 236 00:14:46,600 --> 00:14:50,045 actually detect defects better than humans. So 237 00:14:50,045 --> 00:14:53,805 it's actually easier for a fraudsters to treat a leading 238 00:14:53,805 --> 00:14:57,560 person than a model. This model, like, can look back from certain artifacts with 239 00:14:57,560 --> 00:15:01,340 eyes or just, like, some sort of, you know, glitches. 240 00:15:01,640 --> 00:15:05,265 It's easy. But for person, especially the quality of the image is It's bad. 241 00:15:05,265 --> 00:15:08,625 It's like there is no way anybody can actually spot this is the 242 00:15:08,625 --> 00:15:12,385 problem. And this is great. It it is a problem. I I I must I 243 00:15:12,385 --> 00:15:15,560 must admit this is, I think, this is what we 244 00:15:15,860 --> 00:15:19,220 actually have to be have to hear 245 00:15:19,220 --> 00:15:22,965 about about creating deep fakes. I know that that 246 00:15:22,965 --> 00:15:26,005 is a very interesting thing. So, you know, about I mean, there are lots of 247 00:15:26,005 --> 00:15:29,720 things happening, around AR regulations, Especially in the 248 00:15:29,720 --> 00:15:33,480 European Union. Sure. And then so we actually tried to follow and then to 249 00:15:33,480 --> 00:15:37,080 make sure that everything is compliant. And actually, I wanted to say that we touched 250 00:15:37,080 --> 00:15:40,485 upon k y c KYT, which is know your 251 00:15:40,485 --> 00:15:44,245 transaction. There was also KYB and all your business, which is basically, you 252 00:15:44,245 --> 00:15:47,350 know, how we make sure that the company you work with is is 253 00:15:48,550 --> 00:15:52,310 I know fraudsters. And there is also a thing called k y 254 00:15:52,310 --> 00:15:55,850 a I, know your AI. And it says about transparency. 255 00:15:56,215 --> 00:15:59,835 So many people out there want to be to know actually how AI is used. 256 00:15:59,975 --> 00:16:03,495 So the k l it's it's a very new trend, I think. You have never 257 00:16:03,495 --> 00:16:06,060 heard about it because, I mean, it was going to be a week ago. Since 258 00:16:06,060 --> 00:16:09,660 I like, I want to actually know what's happening with all of this model of 259 00:16:09,660 --> 00:16:13,260 error, not just about touch prod, ground everywhere. But back 260 00:16:13,260 --> 00:16:16,605 to the problem with defects. The thing is, 261 00:16:20,265 --> 00:16:21,565 what to to say that, 262 00:16:24,290 --> 00:16:27,490 Oh, sorry. I lost the my my train of thought. But this is the all 263 00:16:27,490 --> 00:16:30,370 the time. Yeah. We I was just about to say that. But what you know, 264 00:16:30,370 --> 00:16:34,204 one solution to this, I I think, Pavel, would be 265 00:16:34,204 --> 00:16:38,045 if people did something, you know, like, I don't know, colored their 266 00:16:38,045 --> 00:16:41,579 hair Or grew a cool beard. I'm just 267 00:16:41,579 --> 00:16:44,940 throwing that out and with apologies to people listening and not 268 00:16:44,940 --> 00:16:48,300 watching. No. You know? I'm just 269 00:16:48,300 --> 00:16:52,045 saying. But but if you did but if you did 270 00:16:52,045 --> 00:16:55,805 grow a beard, would would or or or change your hair color or 271 00:16:55,805 --> 00:16:59,310 altered their face? Like, I know that, like, facial most facial recognitions 272 00:16:59,449 --> 00:17:03,050 use landmarks on, like, the eye sockets. Right. The a lot harder to change I 273 00:17:03,050 --> 00:17:06,270 was joking. Didn't mind. But, like, would it would it would that 274 00:17:07,295 --> 00:17:10,734 I don't know. Like, does that have any impact on these kind of systems or 275 00:17:10,734 --> 00:17:14,570 are they more like facial recognition systems? They are, 276 00:17:15,270 --> 00:17:19,030 it's, so we operate on the if you're talking about defect detectors or 277 00:17:19,030 --> 00:17:22,470 defect, models for defect detection. Yeah. There are 278 00:17:22,470 --> 00:17:26,135 some, I can't say that I face recognition. The 279 00:17:26,135 --> 00:17:29,895 models, they mostly focus on artifacts. So so for 280 00:17:29,895 --> 00:17:33,610 instance, like, a defect of a year ago, usually, 281 00:17:34,550 --> 00:17:38,070 had problems with eyes. Your eyes of a defect, they usually are 282 00:17:38,070 --> 00:17:41,530 very, you know, not really human. 283 00:17:42,184 --> 00:17:45,965 So it will be changed. It will be like as as as the technology, 284 00:17:46,585 --> 00:17:49,950 is getting more advanced. But like a few years ago, you can actually just crop 285 00:17:50,429 --> 00:17:54,190 Eyes of an image of a person, pretending to be a human person, then they'd 286 00:17:54,190 --> 00:17:57,389 make sure that this is actually a defect. Also I must say that 287 00:17:57,825 --> 00:18:01,345 Yeah. So a video is is is easier to detect because you can actually 288 00:18:01,505 --> 00:18:05,210 so, there is a thing called, I don't like the term in blindness because 289 00:18:05,370 --> 00:18:08,890 No, but nobody actually know what Linus is, but Linus is a detection. 290 00:18:08,890 --> 00:18:12,730 Linus detection is detection. If this is a 291 00:18:12,730 --> 00:18:16,554 leading person or not. And before, like, 5 years ago, it was 292 00:18:16,774 --> 00:18:20,534 mostly a distinction between, a video of a person or 293 00:18:20,534 --> 00:18:23,420 a printed out image. Now it's a detection of an image, 294 00:18:24,360 --> 00:18:27,500 defect, and the linear person. And at that time, 295 00:18:28,440 --> 00:18:31,695 you actually there are 2 types of fly misses. One tool that's passive, 296 00:18:32,315 --> 00:18:35,835 and we actually use also sometimes our customers actually ask us for 297 00:18:35,835 --> 00:18:39,480 pacifying. Let's adjust 1 image. But it's easier for 298 00:18:39,480 --> 00:18:43,020 us and for everybody else to ask a person to actually do something. 299 00:18:43,400 --> 00:18:47,175 And for defects, for instance, like, if I ask them to rotate, Sometimes some 300 00:18:47,175 --> 00:18:51,015 artifacts can appear. Some artifact. And then you can actually see that probably. I 301 00:18:51,015 --> 00:18:54,315 mean, this is not the only person. There are some sort of problems with visual 302 00:18:54,615 --> 00:18:58,250 artifacts. So it is it is like this. 303 00:18:58,710 --> 00:19:02,310 Also, I must say that there was also a challenge for us because there 304 00:19:02,310 --> 00:19:05,815 are, certain cameras. They have some sort of a 305 00:19:05,815 --> 00:19:08,794 beautifiers. So I'm pretty sure as I'm calling from my, 306 00:19:09,575 --> 00:19:12,394 my computer, and then my camera actually 307 00:19:13,400 --> 00:19:16,940 Advances my image. So my image is a little bit, better 308 00:19:17,000 --> 00:19:20,360 than I'm in the real life. So my my skin is is is a little 309 00:19:20,360 --> 00:19:23,805 bit better. So it's it is actually, Embedded into 310 00:19:23,805 --> 00:19:27,645 hardware. And for us, it looks like, some sort of, you know so there is 311 00:19:27,645 --> 00:19:31,025 a signal for us. It does some sort of, you know it's Oh, I see. 312 00:19:31,885 --> 00:19:34,410 So It's hard. You know? And you have to make sure that make sure that, 313 00:19:34,410 --> 00:19:38,090 okay, it's not defect. It's just the person using that, camera off my, 314 00:19:39,130 --> 00:19:42,125 computer. It's like, you know, you have you have to be really, a 315 00:19:43,145 --> 00:19:46,825 yellow error. Apple, I mean, installs 316 00:19:46,825 --> 00:19:50,640 another camera, and then you have to be actually tune your models to make 317 00:19:50,640 --> 00:19:54,080 sure that you actually do not penalize people from with 318 00:19:54,400 --> 00:19:58,240 I think about that. Yeah. The cameras are gonna behave differently if you use different 319 00:19:58,240 --> 00:20:01,375 cameras. So I'm here using my 4 k, 320 00:20:01,995 --> 00:20:05,675 camera. Kind of an outdated one, but it's still it does the job. But what 321 00:20:05,675 --> 00:20:09,430 if I pick up my droid Or, you know, my wife 322 00:20:09,430 --> 00:20:13,030 my wife, you know, she's the the device. She's got an 323 00:20:13,030 --> 00:20:16,835 iPhone. And if I'm trying to log in through her device, That would be different 324 00:20:16,835 --> 00:20:20,515 images, and it may change. You know, it may tell me, nope. That's not 325 00:20:20,515 --> 00:20:24,210 you. Those are gonna be different artifacts. That's fascinating. And I also 326 00:20:24,210 --> 00:20:27,830 think it's funny that you have an old four k camera, which 327 00:20:29,090 --> 00:20:32,605 is a pretty funny thing to say. Like For for podcasting, I won't 328 00:20:32,765 --> 00:20:36,465 No. I know. I don't wanna throw back to, theme from yesterday's 329 00:20:36,684 --> 00:20:40,044 2 hour show, but I'll just make this note. We we 330 00:20:40,044 --> 00:20:43,820 learned that we're in the top 2 a half percent of podcasts. 331 00:20:43,820 --> 00:20:47,600 So now I feel like I should have, I don't know, 16 k studio 332 00:20:47,820 --> 00:20:51,535 and Yeah. I should have a lot of time like Joe Rogan has in a 333 00:20:51,535 --> 00:20:54,975 brick wall. Exactly. Right. I don't I need something better than this 334 00:20:54,975 --> 00:20:58,355 old four k camera. But 335 00:20:58,975 --> 00:21:02,610 if all of a sudden You just want to open a bank account right 336 00:21:02,610 --> 00:21:06,450 now. Yeah. It looks strange because, I mean, a typical person is like you 337 00:21:06,450 --> 00:21:09,925 use your iPhone or you're like a regular computer. Like, with 4 k or 16 338 00:21:09,925 --> 00:21:13,365 k camera, it's like very strange. It's some something, you know. It's it's a signal 339 00:21:13,365 --> 00:21:17,090 for for every model and make sure that It's an outlier. Right? And 340 00:21:17,090 --> 00:21:20,690 it sounds like a big this is still obviously, there's way 341 00:21:20,690 --> 00:21:24,434 more complicated things than what you do, But outliers 342 00:21:24,735 --> 00:21:28,434 detecting outliers is probably 1 1 big tool in your tool belt. 343 00:21:28,654 --> 00:21:32,470 It is. Yeah. That's very hard if you have a Genuine person, 344 00:21:32,470 --> 00:21:36,070 and you are an outlier somehow. I mean, everybody can be an 345 00:21:36,070 --> 00:21:39,210 outlier in some sense. It's very hard because, yeah, 346 00:21:40,525 --> 00:21:44,285 So this is hard. So, like, at some point, yeah, colored hairs 347 00:21:44,285 --> 00:21:47,885 can be also an outlier. I don't No. It's just interesting. So I imagine, like, 348 00:21:47,885 --> 00:21:51,450 Instagram filters and things like that probably also cause 349 00:21:51,450 --> 00:21:55,210 chaos and things like that. Yeah. Of course. But, yeah, I 350 00:21:55,210 --> 00:21:58,375 mean So usually use, yeah, filters, 351 00:21:59,154 --> 00:22:02,674 a strong signal for us. I mean Right. And also I must I must have 352 00:22:02,674 --> 00:22:06,370 this defects. So going back, thing with defects is that 353 00:22:06,370 --> 00:22:10,050 it's not, like, specifically use the fraudsters. Here's the 354 00:22:10,050 --> 00:22:13,515 problem. You know, there are lots of cool things for defects. You can press 355 00:22:13,515 --> 00:22:16,975 advertising. Right. I don't know what what else. But, usually, 356 00:22:18,075 --> 00:22:21,799 you can actually adopt a person to, like, Replaced an 357 00:22:21,799 --> 00:22:25,240 actor in the movie. This is also a defect. It's a very cool defect, very 358 00:22:25,240 --> 00:22:28,934 sophisticated defect, very high quality defect. Still a defect. So those 359 00:22:28,934 --> 00:22:32,615 are our usage is actually for for that, I mean, not just for fraud. 360 00:22:32,615 --> 00:22:36,230 And then going back to our problems, it's like, I mean, And the 361 00:22:36,309 --> 00:22:40,070 even even that and even that from that, I like this example, 362 00:22:40,070 --> 00:22:42,010 but, the guys from the, 363 00:22:45,044 --> 00:22:48,804 I mean so we focus on financial fraud. Yeah. So it's more or less like 364 00:22:48,804 --> 00:22:52,645 people trying to actually sue money on, like, take over your account, something like 365 00:22:52,645 --> 00:22:56,389 that. But the thing is the defects, they are mostly created 366 00:22:56,389 --> 00:22:59,669 not for that. And this is a very interesting thing, I think. They are created. 367 00:22:59,669 --> 00:23:02,945 And, actually, I didn't know about that, but we actually knew that When they started 368 00:23:02,945 --> 00:23:06,405 to try and to create our Deepak's. So we went, you know, to the Internet, 369 00:23:06,545 --> 00:23:09,765 some strange forms to make sure what what people actually use 370 00:23:10,179 --> 00:23:13,960 What they create deep eggs for. And they create 371 00:23:14,100 --> 00:23:17,799 deep eggs for porn. It's like 98%, 89% 372 00:23:19,245 --> 00:23:22,765 Deepex, I slide 4. And this is also a problem because in in there is 373 00:23:22,765 --> 00:23:26,600 a thing called nonconsensual port. Deepex are used for that, And this 374 00:23:26,600 --> 00:23:29,080 is also a problem. So it's not our business, but the thing is that the 375 00:23:29,080 --> 00:23:32,539 same technologies is there. And you actually I mean, if you, 376 00:23:33,215 --> 00:23:36,735 I mean, work in the area, you can actually so the same model can actually 377 00:23:36,735 --> 00:23:40,255 be applied to detect, this type of defects. Right. So it's 378 00:23:40,255 --> 00:23:44,090 different, but, I mean yeah. Yes. It's, That was expressed to 379 00:23:44,090 --> 00:23:47,850 me maybe a year ago. It's fascinating how 380 00:23:47,850 --> 00:23:51,505 quickly this space is just Evolving or 381 00:23:51,885 --> 00:23:55,505 devolving, I guess, depending on your point of view. Yeah. 382 00:23:56,125 --> 00:23:58,465 But, no, you're right. Like, most of it is 383 00:23:59,630 --> 00:24:03,090 Those a lot of the deep fake kind of work is done 384 00:24:03,390 --> 00:24:07,230 for adult content. And, you know, and it's there 385 00:24:07,310 --> 00:24:11,125 the The legislation around this is gonna vary 386 00:24:11,125 --> 00:24:14,565 widely from place to place. But, like, you know, 387 00:24:14,565 --> 00:24:18,370 revenge porn laws don't apply. And there. I I think that was a big thing 388 00:24:18,370 --> 00:24:22,210 in, and there was a controversy somewhere. I think it 389 00:24:22,210 --> 00:24:25,695 was New Jersey, Where somebody had 390 00:24:25,755 --> 00:24:29,515 created deep fake images of either high 391 00:24:29,515 --> 00:24:32,875 school or middle school girls, which adds an extra level of 392 00:24:32,875 --> 00:24:36,520 legal Concern I have a whole lots of extra 393 00:24:36,520 --> 00:24:40,360 levels of concern. Let's be honest. But, like, you know and and and and 394 00:24:40,360 --> 00:24:43,895 there was this, you know, the big debate. And my first reaction was, I'm 395 00:24:43,895 --> 00:24:47,434 actually kinda surprised it took this long for that to happen, 396 00:24:48,375 --> 00:24:51,495 which is a very cynical take, I'll admit. But I can tell I I can 397 00:24:51,495 --> 00:24:54,929 tell you the reason. The thing is that Technology moves so fast. Yes. And 398 00:24:54,929 --> 00:24:57,570 legislation actually is always, like 399 00:24:58,929 --> 00:25:02,150 so even with with EAU, AI act, 400 00:25:02,725 --> 00:25:06,485 those I mentioned defects just a little because they started working on 401 00:25:06,485 --> 00:25:09,945 the regulations 2 years ago. And 2 years ago, it was not a problem. 402 00:25:10,169 --> 00:25:13,049 And now it's, like, all over, you know, the Internet, and then you have to 403 00:25:13,049 --> 00:25:16,350 actually tweak the, wording, 404 00:25:16,890 --> 00:25:20,585 but it takes time. Well, even still, like, you know, like, there's, 405 00:25:21,545 --> 00:25:25,145 a few months ago, they had these fake commercials that were created by with 406 00:25:25,145 --> 00:25:28,779 combination of 11 Labs and A few other companies to name them, so I 407 00:25:28,779 --> 00:25:32,460 forget. But, you know, they had a picture of Elon Musk, you 408 00:25:32,460 --> 00:25:36,299 know, eating spaghetti, and it looked weird. But you can easily see, 409 00:25:36,299 --> 00:25:40,015 like, You know, I was messing around with v q early versions of v 410 00:25:40,015 --> 00:25:43,375 q grant d q GANs in early 411 00:25:43,375 --> 00:25:47,090 2022, And that stuff looked 412 00:25:47,090 --> 00:25:50,610 weird, and it it really evolved. And this morning, I saw 413 00:25:50,610 --> 00:25:54,310 Pika AI, I guess, just went Yeah. Yeah. Yeah. Went to a wider beta. 414 00:25:54,585 --> 00:25:58,424 And, yeah, released and and and, like, I'm seeing what's created with that, 415 00:25:58,424 --> 00:26:02,125 and, you know, it still looks weird, it still looks cartoonish, 416 00:26:02,745 --> 00:26:06,419 but it's not The fact that we've gone that far in the span 417 00:26:06,419 --> 00:26:10,179 of, you know, less than 2 years, like, I think says something, like and to 418 00:26:10,179 --> 00:26:13,835 your point, legislation Usually takes years, to 419 00:26:13,835 --> 00:26:17,135 make. So, like, by the time these laws are written, they may not be valid. 420 00:26:17,195 --> 00:26:20,415 In the case of New Jersey, I think there's some debate over, 421 00:26:22,990 --> 00:26:26,290 does what sorts of laws that applies to? Because 422 00:26:26,430 --> 00:26:29,785 the the original, The faces 423 00:26:30,005 --> 00:26:33,605 were mapped on to something else, but that the 424 00:26:33,605 --> 00:26:37,205 something else I'm trying to keep our clean rating here. The something else were 425 00:26:37,205 --> 00:26:40,870 people over 18, but the bases were mapped onto it. So there's 426 00:26:40,870 --> 00:26:43,770 some debate over, do existing laws cover that? 427 00:26:44,710 --> 00:26:48,145 I'm not a lawyer. Don't look at me, and I'm not. But, 428 00:26:48,225 --> 00:26:51,684 it's just fascinating to your point. Like, this is moving quickly. 429 00:26:52,865 --> 00:26:56,559 Yep. It's definitely complicated. So we've 430 00:26:56,559 --> 00:27:00,240 reached the point in our show, Pavel, where we, like to 431 00:27:00,240 --> 00:27:03,795 ask a set of questions. They're in the chat. And 432 00:27:03,795 --> 00:27:07,175 I'll start out, with the, the very first question. 433 00:27:07,635 --> 00:27:11,395 How did you find your way into this field? Did this field find you, 434 00:27:11,395 --> 00:27:15,240 or did you find it? Yeah. I must say I have a 435 00:27:15,240 --> 00:27:18,840 story to tell. I just studied yeah. Studied computer 436 00:27:18,840 --> 00:27:22,575 science at, university And I actually worked as a software engineer 437 00:27:22,575 --> 00:27:26,095 at Motorola. You may remember this company, with 438 00:27:26,095 --> 00:27:29,475 HQ in Chicago back then, for 5 years. 439 00:27:29,730 --> 00:27:33,169 And then it was, 2011, which is, like, long time 440 00:27:33,169 --> 00:27:36,850 ago, the very first, massive 441 00:27:36,850 --> 00:27:40,155 online courses appeared. There was a one called AI class, 442 00:27:40,535 --> 00:27:44,375 and it later turned out to be a Udacity. And there 443 00:27:44,375 --> 00:27:48,000 was also a m l called ML class. It's a ML class. And 444 00:27:48,000 --> 00:27:51,720 this now this Coursera. It's like 10 years ago. And I was like, okay. 445 00:27:51,720 --> 00:27:55,080 Cool. I enrolled and actually, I pushed because it is like it was it was 446 00:27:55,080 --> 00:27:58,695 hard. It was like, you have to really, be involved. And 447 00:27:58,695 --> 00:28:01,895 then I felt like, okay, this is a cool thing. This is like a next 448 00:28:01,895 --> 00:28:04,795 big thing for me and, like, for everybody else. It was like 449 00:28:06,010 --> 00:28:09,850 12 years ago. So I quit my job, and I actually, so 450 00:28:09,850 --> 00:28:13,130 at the same time, I started to try to run a small startup with my 451 00:28:13,130 --> 00:28:16,655 friend, failed miserably. But I take, took my time, studied, 452 00:28:17,435 --> 00:28:20,795 for maybe half a year, and then joined a small data 453 00:28:20,795 --> 00:28:24,440 startup as a data scientist. And then it just 454 00:28:24,440 --> 00:28:28,040 started there. So it's I think I I find, my way into 455 00:28:28,040 --> 00:28:31,320 data. But Yeah. I don't know. So You want to 456 00:28:31,924 --> 00:28:35,125 I'm sorry. Go ahead. I just I just say it sounds like you were very 457 00:28:35,125 --> 00:28:38,585 intentional about finding your way into it. So that's cool. Yeah. 458 00:28:39,125 --> 00:28:42,970 That's cool. And I see you were You were at VK for a while too, 459 00:28:42,970 --> 00:28:46,730 which I've never seen VK, but I hear it's like a like 460 00:28:46,730 --> 00:28:50,534 a Russian language version of Twitter slash Facebook. It used 461 00:28:50,534 --> 00:28:54,015 to be. Yes. Yeah. Yeah. I don't I yeah. Obviously, now things are different, but 462 00:28:54,135 --> 00:28:57,335 yeah. Yeah. Yeah. Yeah. I worked there for 5 years, a long time ago. Oh, 463 00:28:57,335 --> 00:29:00,880 interesting. And, you know, if you're talking about the data, I mean, 464 00:29:00,880 --> 00:29:04,720 the, where it's like the the place where you can 465 00:29:04,720 --> 00:29:08,305 actually play with data. You can actually cool do many cool things. 466 00:29:08,785 --> 00:29:12,465 Oh, yeah. Nice. Nice. And he's being modest. According to LinkedIn, he was director 467 00:29:12,465 --> 00:29:14,565 of AI research, so he's super smart. 468 00:29:16,580 --> 00:29:20,260 But, what's your favorite part of 469 00:29:20,260 --> 00:29:24,075 your current job? Oh, I can't say it 470 00:29:24,075 --> 00:29:27,515 could create some defects, but, it's not 471 00:29:27,515 --> 00:29:29,775 it. I think 472 00:29:31,900 --> 00:29:34,960 no. I mean, I would say that what I like is, they, 473 00:29:36,780 --> 00:29:40,365 the the Samsung, Samsung is is now it's it's a product or any company. So 474 00:29:40,365 --> 00:29:44,205 have our own own products, whether, like, a technology company, yet we have our 475 00:29:44,205 --> 00:29:47,965 own product. And having that, 476 00:29:47,965 --> 00:29:51,799 actually, our own product, Actually helps us, you know, I know what our 477 00:29:51,799 --> 00:29:55,639 customer wants. Wonderful. I know the 478 00:29:55,639 --> 00:29:58,120 data. So it's like, you know, I mean, you have to actually so you have 479 00:29:58,120 --> 00:30:01,664 to look around. Okay. There is a problem with defects. I have to, 480 00:30:01,664 --> 00:30:05,265 like, make sure that I mean, I had, I actually have to understand this. This 481 00:30:05,265 --> 00:30:08,890 is a problem. And for many of our customers, I mean, I 482 00:30:08,890 --> 00:30:12,170 don't I would not like to say that we have to educate them or actually 483 00:30:12,170 --> 00:30:15,745 make make sure that they understand this is a problem with defects. And now we 484 00:30:15,745 --> 00:30:19,525 have when they understand, we can actually help them with their their, 485 00:30:20,785 --> 00:30:24,304 safety and security. One thing that this is, like, a little bit, I 486 00:30:24,304 --> 00:30:27,420 mean, Clumsy answer, but I'm sorry if you know. 487 00:30:28,600 --> 00:30:32,140 Yeah. Being closer to the product is is is is fun. 488 00:30:32,835 --> 00:30:36,375 Oh, sorry. Cool. So we have 3 complete 489 00:30:36,435 --> 00:30:40,115 sentence. And the first one is when I'm not 490 00:30:40,115 --> 00:30:42,420 working, I enjoy blank. 491 00:30:43,940 --> 00:30:47,780 Okay. Okay. Let me think for a while. There are many things I can 492 00:30:47,780 --> 00:30:50,340 say. No. I can say no. This is I think of this as I can, 493 00:30:50,500 --> 00:30:54,165 I can share? No. I I I I run or I can see job. 494 00:30:54,305 --> 00:30:57,765 Mhmm. Oh, cool. Cool. I run-in the the ring marathon. 495 00:30:58,465 --> 00:31:02,159 This is my Nice. There are Major Martins, like, 5, 496 00:31:02,480 --> 00:31:05,919 6 Martins across the world. So that's New York, Paris, 497 00:31:05,919 --> 00:31:09,139 London, Tokyo, Berlin, and, 498 00:31:10,135 --> 00:31:13,735 London. Nice. Like, 6 so that Very So Berlin was my 1st major 499 00:31:13,735 --> 00:31:17,335 marathon. So I ran it, this this September, and it was great. No. That's 500 00:31:17,335 --> 00:31:18,475 awesome. That's awesome. 501 00:31:21,690 --> 00:31:25,289 When you said Berlin, the first thing that popped in my mind was, Berliner 502 00:31:25,289 --> 00:31:28,110 Kendall wrote, which is like this local kinda drink. 503 00:31:29,205 --> 00:31:32,505 Yeah. Yeah. Yeah. Yeah. I know. That's like Yeah. 504 00:31:32,565 --> 00:31:36,245 Yeah. But I prefer there is a it's a vehicle. It's 505 00:31:36,245 --> 00:31:39,650 like a craft. Oh, yeah. From Berlin. 506 00:31:39,650 --> 00:31:43,410 Right. But I talking about Berlin, so I run. It was 507 00:31:43,410 --> 00:31:47,065 super fun, but, on my finishing picture, so 508 00:31:47,065 --> 00:31:50,825 it's my me, Ryan. So close to Bernsberg. It's a 509 00:31:50,825 --> 00:31:54,505 very central grid. Mhmm. And there is also a guy in the 510 00:31:54,505 --> 00:31:58,270 bottle question. And and I 511 00:31:58,270 --> 00:32:01,790 wasn't it was not slow. I wasn't slow. Yeah. There was a guy in a 512 00:32:01,790 --> 00:32:05,395 huge ball, like, I still running, like, finishing with me. Like, so it was, Oh, 513 00:32:05,395 --> 00:32:08,535 that's funny. That's fun. It's that's fun. That's funny. Very cool. 514 00:32:09,715 --> 00:32:13,235 Next, complete the sentence. I think the coolest thing in 515 00:32:13,235 --> 00:32:15,015 technology today is 516 00:32:18,260 --> 00:32:22,020 blank. Oh, it's it's it's hard to say. Let me I'll just 517 00:32:22,020 --> 00:32:24,280 think for a while. But, I mean, 518 00:32:25,835 --> 00:32:29,675 I think that so my my area 519 00:32:29,675 --> 00:32:33,355 seems like I expert a personally specified natural 520 00:32:33,355 --> 00:32:37,159 language processing. So I know about language models. And, 521 00:32:37,159 --> 00:32:40,600 actually, we had papers on language models, like, before they they 522 00:32:40,600 --> 00:32:44,424 were super big. So, like, on tuning language models. Yes. I 523 00:32:44,424 --> 00:32:47,945 found it really, really exciting that it in a 524 00:32:47,945 --> 00:32:50,845 year, it went from, you know, research 525 00:32:51,480 --> 00:32:55,320 Prototypes to, like, everyday product. This is Yeah. This was 526 00:32:55,320 --> 00:32:59,054 a compelling. So, like, my parents used Chargebee PCs. Like, I mean, this 527 00:32:59,054 --> 00:33:01,955 is like this is like a mobile phone. This is I mean, this is what, 528 00:33:02,095 --> 00:33:05,635 like, some sort of a milestone, last year. 529 00:33:05,909 --> 00:33:09,529 I think this is this is it. And he is that the actual unit 530 00:33:09,830 --> 00:33:13,669 for main things. You can build products on on language models. And 531 00:33:13,669 --> 00:33:17,475 this is also like. It's wild, isn't it? Like, you know, 532 00:33:17,535 --> 00:33:21,295 and and it's captured everybody's imagination in in good and bad ways. 533 00:33:21,295 --> 00:33:24,929 But, like, my father-in-law, you know, So he used to 534 00:33:24,929 --> 00:33:28,230 say Frank works with computers. Now he says Frank works in AI. 535 00:33:28,450 --> 00:33:31,669 Okay. You know? That's good. 536 00:33:32,345 --> 00:33:35,545 But I also like we used to say machine learning. So now you have to 537 00:33:35,545 --> 00:33:39,225 say AI. That's right. That's right. You have to say that data mining 538 00:33:39,225 --> 00:33:42,045 core something. So it's like, you know That's right. It definitely would. 539 00:33:43,050 --> 00:33:45,950 I wonder what it'll be next year. Who knows? Gen AI probably. 540 00:33:47,770 --> 00:33:51,565 Probably. So our next one, complete this Regulate, I think. Oh, that's 541 00:33:51,565 --> 00:33:54,925 right. Regulation. That's right. Regular. Our our last completes the 542 00:33:54,925 --> 00:33:58,705 sentence is I look forward to the day when I can use technology 543 00:33:59,005 --> 00:34:02,790 to blank. Uh-huh. I 544 00:34:02,790 --> 00:34:06,470 can't it's hard to answer because, I mean, like, I 545 00:34:06,470 --> 00:34:10,105 can't say it would be cool If I can, you know, 546 00:34:10,105 --> 00:34:13,704 develop drugs. And then there are very cool startups for drug design 547 00:34:13,704 --> 00:34:17,070 with AI. Yet, I mean, Just imagine we have 548 00:34:17,070 --> 00:34:20,670 a a a cure for cancer, but Right. We have so 549 00:34:20,670 --> 00:34:24,110 many diseases to care to cure. So let's say, I think I 550 00:34:24,110 --> 00:34:27,855 hope Once we fix anything, then there is gonna 551 00:34:27,855 --> 00:34:31,135 be a next, you know, next milestone for us to look forward. So I'm sorry 552 00:34:31,135 --> 00:34:34,950 if, you know, there's never I hope there will be 553 00:34:34,950 --> 00:34:38,790 no such date, I can say. Right. Right. That's 554 00:34:38,790 --> 00:34:41,690 a good one. I'm pretty sure you will agree with me. Like Yeah. 555 00:34:42,474 --> 00:34:45,855 Especially work with the technology. I mean So true. For sure. 556 00:34:46,635 --> 00:34:50,420 The next question, share something different about yourself, but remember, It's a family 557 00:34:50,420 --> 00:34:54,100 oriented well, not family oriented, but we like we we like it so that 558 00:34:54,100 --> 00:34:57,164 you can list it with your kids in the in the car. Right? Like, That's 559 00:34:57,164 --> 00:35:00,684 kind of a Yeah. Yeah. Yeah. And, yeah, and I live in Berlin across, very 560 00:35:00,684 --> 00:35:04,384 close. There's a very, how to say, kinky club, which is Berlin. 561 00:35:05,960 --> 00:35:09,660 Was that the the the tier garden? It's it's 562 00:35:10,119 --> 00:35:13,960 it's a it's a it is family friendly. It's it's like the most family 563 00:35:13,960 --> 00:35:17,745 friendly place in in in Berlin. You got some. Yeah. No. It's it's 564 00:35:17,745 --> 00:35:21,285 called KitKat. Yes. What I can say. 565 00:35:21,905 --> 00:35:25,630 I have, purple hair. Since last month. 566 00:35:26,250 --> 00:35:29,230 I don't know. So I can say that I speak 567 00:35:31,609 --> 00:35:35,425 a few languages, all of that. But, no, I'm I'm 568 00:35:35,425 --> 00:35:39,185 joking. So I speak Japanese. I don't I don't Japanese, for a 569 00:35:39,185 --> 00:35:41,925 long time. So I I can speak Japanese. I speak 570 00:35:43,030 --> 00:35:46,710 English, obviously, Russian. My parents are from 571 00:35:46,710 --> 00:35:50,385 Russia. And I also speak German. So I actually Studied 572 00:35:50,385 --> 00:35:53,905 German for 2 years. So I actually studied right now. So I had, like, my 573 00:35:53,905 --> 00:35:57,505 German classes 3 or 4 times per week, which is 574 00:35:57,665 --> 00:36:00,670 let me just go. Sorry. So I hope in a year, I will be able 575 00:36:00,670 --> 00:36:04,350 to do a podcast in German as well. Oh, Wendeschon. That is 576 00:36:04,350 --> 00:36:04,850 not 577 00:36:10,815 --> 00:36:14,400 Yeah. Yeah. And we just lost, like, We we just 578 00:36:14,400 --> 00:36:18,240 looked at our analytics, and, like, most of our listeners are from English language countries. 579 00:36:18,240 --> 00:36:22,080 So I think we just lost them. Maybe we 580 00:36:22,080 --> 00:36:25,225 can attract new listeners. Oh, I like it. I like the way you think. We 581 00:36:25,225 --> 00:36:28,125 wanna we wanna get to the top 2.4% now. 582 00:36:29,065 --> 00:36:30,845 Our new goal. So, 583 00:36:33,090 --> 00:36:36,690 Audible is a sponsor of the show, and I'm not sure if 584 00:36:36,690 --> 00:36:39,010 Audible is big in Europe. I think it is because I've seen a lot of 585 00:36:39,010 --> 00:36:42,835 German language audiobooks. It is a no. Okay. 586 00:36:43,155 --> 00:36:46,355 So do you do you listen to audiobooks? And if so, you have a good 587 00:36:46,355 --> 00:36:49,930 recommendation. Otherwise, we'll take a recommendation on the regular good Fashion 588 00:36:49,930 --> 00:36:53,690 paper dead tree book. No. I have a couple. I think I can 589 00:36:53,690 --> 00:36:56,110 give you a couple of examples. This is like, 590 00:36:58,035 --> 00:37:01,795 I like this was the most, you know so so I'm so my 591 00:37:01,795 --> 00:37:04,695 background is from many, places, 592 00:37:05,490 --> 00:37:09,330 since Israel, Russia, and Germany in some extent. So 593 00:37:09,330 --> 00:37:13,095 I would recommend, there is a very Good book. It is 594 00:37:13,255 --> 00:37:15,895 in my opinion, this is very known, but not many people know about it for 595 00:37:15,895 --> 00:37:19,275 some reason. It's called the good soldier's make. Okay. 596 00:37:19,760 --> 00:37:23,440 Like it said, didn't Not heard of. About the, sort of 597 00:37:23,520 --> 00:37:27,145 third world war by Oh, interesting. But this is 598 00:37:27,185 --> 00:37:30,805 it's very good. Like, you can actually learn a lot about 599 00:37:31,185 --> 00:37:34,865 Czech Republic, Germany, Austria in the beginning of 600 00:37:34,865 --> 00:37:38,630 the, Last century. Oh, interesting. 601 00:37:38,630 --> 00:37:41,109 Especially now, it's the very thing. It's called in the park. This is a very 602 00:37:41,109 --> 00:37:44,809 good thing too. And it's very funny. It's like one of the funniest, books 603 00:37:45,109 --> 00:37:48,805 ever written. And also the the second one, I have 2. 604 00:37:49,425 --> 00:37:52,885 This called Arc of Triumph, by remark. Okay. 605 00:37:53,270 --> 00:37:57,110 This is also about the pre war Europe, pre second World War 606 00:37:57,110 --> 00:38:00,790 Europe, like, Southeast, years of the 607 00:38:00,790 --> 00:38:04,365 last century. And this is also very, like, you know, you 608 00:38:04,365 --> 00:38:08,145 really you really feel like what what was the I mean, living in 609 00:38:08,365 --> 00:38:12,140 Germany and, France, during that time, it's very, very interesting. 610 00:38:12,140 --> 00:38:15,980 So one of my favorites. So I can definitely recommend both of these 611 00:38:15,980 --> 00:38:19,565 videos. Very cool. So audible detecting I'm sorry. I'm 612 00:38:19,565 --> 00:38:22,925 detecting a history theme. Yes. Yeah. 613 00:38:22,925 --> 00:38:26,740 Yeah. Yeah? Cool. There's a really good book. Since you live in Berlin, 614 00:38:26,740 --> 00:38:30,500 you might like it. It's called Faust's Metropolis, and 615 00:38:30,500 --> 00:38:34,225 it's about the history of Berlin from, like, you know, Almost 616 00:38:34,225 --> 00:38:37,905 stone age time till Okay. Cool. You know, the 20 617 00:38:38,145 --> 00:38:41,770 you know, early 21st century is kind of like And the basic 618 00:38:41,770 --> 00:38:45,530 gist is, like, you know, a lot has happened in Berlin. Good. 619 00:38:45,530 --> 00:38:49,135 Sure. Yeah. We all know the bad. Right? But, like, some good things have 620 00:38:49,135 --> 00:38:52,575 happened, kinda everything in between. It's kind of it's an interesting look at, like, the 621 00:38:52,575 --> 00:38:56,255 history of the city and how it apparently was built on a swamp or something 622 00:38:56,255 --> 00:38:59,829 like that. Like Yeah. It's just, it's it's 623 00:38:59,829 --> 00:39:03,510 interesting. And Audible is a sponsor of 624 00:39:03,510 --> 00:39:06,329 Data Driven. If you go to the data driven book .com, 625 00:39:07,454 --> 00:39:11,295 I think even the data driven book .com might work. Uh-huh. That was 626 00:39:11,295 --> 00:39:14,974 a pronunciation joke. You'll get a free, on 1 free 627 00:39:14,974 --> 00:39:18,670 audiobook on us, and And we'll get a kickback if you sign up for a 628 00:39:18,670 --> 00:39:22,430 subscription. And finally, where can 629 00:39:22,430 --> 00:39:25,810 folks find out about you, more about you, and what you're up to at Sumsub 630 00:39:26,494 --> 00:39:29,715 And, some of the other things you you're up to. 631 00:39:31,855 --> 00:39:35,610 What's up? My my connection was, Oh, where can folks find out 632 00:39:35,610 --> 00:39:38,890 more about you and what you're up to? Oh, yes. It's, 633 00:39:40,330 --> 00:39:43,710 yes. It's, it's a company. It's called Samsung. So Samsung dot com. 634 00:39:44,355 --> 00:39:48,115 Also, like, what we have is, today is 635 00:39:48,115 --> 00:39:51,859 with anti fraud. And you have to I mean, It's not about 636 00:39:51,859 --> 00:39:55,700 all the product. It's actually about making people helping people 637 00:39:55,700 --> 00:39:59,540 learn about, security. So how they can actually navigate the Internet or, 638 00:39:59,540 --> 00:40:03,365 like, their life More safely. So we have a portal called 639 00:40:03,365 --> 00:40:06,665 some suburb where we actually post a lot 640 00:40:07,605 --> 00:40:11,230 of stuff on Making your Internet life, 641 00:40:11,230 --> 00:40:15,069 can I say like this, safer? So, actually, I I advise you 642 00:40:15,069 --> 00:40:18,655 to take a look, and then probably you'll find something interesting there. 643 00:40:18,895 --> 00:40:22,275 We definitely will. And, any parting thoughts before 644 00:40:22,575 --> 00:40:26,415 we end the show? Any final thoughts? I just 645 00:40:26,415 --> 00:40:30,060 want to say, yeah, Just I was very happy to, to be here 646 00:40:30,060 --> 00:40:33,820 and hope, it was Cool. Interesting. This is a great show. It's always good to 647 00:40:33,820 --> 00:40:37,234 it's always good to kinda understand The the the intersection 648 00:40:37,375 --> 00:40:41,214 of of AI data and security because some people still see 649 00:40:41,214 --> 00:40:44,630 those as separate things. But I think as time goes on, 650 00:40:45,170 --> 00:40:48,930 we're gonna I'm gonna we're gonna wonder how we ever saw it as separate 651 00:40:48,930 --> 00:40:52,585 things. There are so many things to talk about that. Yeah. Yeah. Yeah. 652 00:40:52,585 --> 00:40:56,205 Yeah. Well, awesome. Any parting thoughts, Andy? 653 00:40:56,585 --> 00:41:00,280 No. Just a great show. Pavel, thank you for, for joining us. 654 00:41:00,280 --> 00:41:03,720 It was our honor. Yes. Likewise. And we'll let 655 00:41:03,720 --> 00:41:06,700 Bailey finish the show. That was some show. 656 00:41:07,455 --> 00:41:11,135 We appreciate you listening to Data Driven. We know you're 657 00:41:11,135 --> 00:41:14,840 busy and we appreciate you listening to our podcast. But 658 00:41:14,840 --> 00:41:18,440 we have a favor to ask. Please rate and review our 659 00:41:18,440 --> 00:41:22,120 podcast on Itunes, Stitcher, or wherever you subscribe to 660 00:41:22,120 --> 00:41:24,994 us. You have subscribed to us, 661 00:41:25,855 --> 00:41:29,535 haven't you? Having high ratings and reviews helps us 662 00:41:29,535 --> 00:41:33,110 improve the quality of our show and rank us more favorably with the search 663 00:41:33,110 --> 00:41:36,869 algorithms. That means more people listen to us, 664 00:41:36,869 --> 00:41:40,630 spreading the joy. And, can't the world use a little 665 00:41:40,630 --> 00:41:44,375 more joy these days? So go do your part to 666 00:41:44,375 --> 00:41:47,815 make the world just a little better and be sure to rate and review the 667 00:41:47,815 --> 00:41:48,315 show.