1 00:00:00,180 --> 00:00:02,760 -: And why don't you bring the students up to speed? 2 00:00:02,760 --> 00:00:06,210 'Cause we were speaking a little bit before this interview 3 00:00:06,210 --> 00:00:08,039 about the process you're at right now, 4 00:00:08,039 --> 00:00:10,410 hiring on some people for your company, 5 00:00:10,410 --> 00:00:13,980 and mention that again, of how you were using ChatGPT 6 00:00:13,980 --> 00:00:18,150 and one of the different facets of something at OpenAI 7 00:00:18,150 --> 00:00:22,170 to speed up that specific hiring process. 8 00:00:22,170 --> 00:00:23,760 -: Absolutely. Yeah. 9 00:00:23,760 --> 00:00:28,760 So, OpenAI, one of their OpenAI is if you don't know that, 10 00:00:28,807 --> 00:00:30,060 for anyone who doesn't know, 11 00:00:30,060 --> 00:00:34,050 it's the people who make ChatGPT the minds behind that. 12 00:00:34,050 --> 00:00:36,660 And they have another a number of other APIs, 13 00:00:36,660 --> 00:00:39,840 and one of them is a similarity scoring algorithm, 14 00:00:39,840 --> 00:00:42,720 which is like, what it's doing in the backend 15 00:00:42,720 --> 00:00:43,860 is like fairly complex, 16 00:00:43,860 --> 00:00:46,320 but you can basically dump it a resume 17 00:00:46,320 --> 00:00:48,270 or any given body of text and say, 18 00:00:48,270 --> 00:00:50,220 hey, does it meet this certain criteria? 19 00:00:50,220 --> 00:00:52,500 Or how similar is it to another set of criteria? 20 00:00:52,500 --> 00:00:54,720 So, in my case, it's the job description 21 00:00:54,720 --> 00:00:56,310 that I wrote for this job, 22 00:00:56,310 --> 00:00:58,961 which I actually got help from ChatGPT writing 23 00:00:58,961 --> 00:01:01,020 to make sure that I was compliant 24 00:01:01,020 --> 00:01:03,210 in every way with the state and local laws, 25 00:01:03,210 --> 00:01:04,769 because I'm not an HR person 26 00:01:04,769 --> 00:01:06,120 who's an expert on those things. 27 00:01:06,120 --> 00:01:08,280 So, ChatGPT gave me a little hand there. 28 00:01:08,280 --> 00:01:11,100 I'm not a lawyer. (chuckles) 29 00:01:11,100 --> 00:01:13,062 Must have the disclaimer in there, right? 30 00:01:13,062 --> 00:01:13,895 -: Yeah. 31 00:01:14,790 --> 00:01:16,320 -: Taking those similarity scores 32 00:01:16,320 --> 00:01:18,450 between the job description and the resume, 33 00:01:18,450 --> 00:01:20,400 that all happens just automatically 34 00:01:20,400 --> 00:01:22,020 with my applicant tracking system. 35 00:01:22,020 --> 00:01:24,205 There's a bit of software where you apply to a job. 36 00:01:24,205 --> 00:01:26,640 It goes into something called an applicant tracking system. 37 00:01:26,640 --> 00:01:28,020 It's like an HR tool basically 38 00:01:28,020 --> 00:01:30,281 to manage applications to a job. 39 00:01:30,281 --> 00:01:32,700 It just pulls those new resumes down, 40 00:01:32,700 --> 00:01:34,500 compares them to the job description. 41 00:01:34,500 --> 00:01:35,640 If it meets a certain level, 42 00:01:35,640 --> 00:01:37,980 a threshold level of which they're similar, 43 00:01:37,980 --> 00:01:40,260 it'll pass them on to the next stage automatically. 44 00:01:40,260 --> 00:01:42,510 So, yeah, when you're hiring for a job, 45 00:01:42,510 --> 00:01:44,884 there's a lot of people who apply, 46 00:01:44,884 --> 00:01:46,830 some of them are very ambitious 47 00:01:46,830 --> 00:01:50,160 and think they can handle maybe a lot more than they can, 48 00:01:50,160 --> 00:01:54,330 which it's lovely to see and I love the confidence, 49 00:01:54,330 --> 00:01:57,090 but the job that we're hiring for 50 00:01:57,090 --> 00:01:58,920 has a very specific set of criteria 51 00:01:58,920 --> 00:02:00,840 and we need to find people 52 00:02:00,840 --> 00:02:01,920 who are really well-qualified 53 00:02:01,920 --> 00:02:03,180 for the things that we're doing here. 54 00:02:03,180 --> 00:02:05,340 So, yeah, 55 00:02:05,340 --> 00:02:06,870 we need to find those edges -: Yeah. 56 00:02:06,870 --> 00:02:08,039 -: where we can and save time, 57 00:02:08,039 --> 00:02:09,948 because having to go through hundreds of resumes, 58 00:02:09,948 --> 00:02:13,710 I also get fatigued, I make mistakes, I overlook things. 59 00:02:13,710 --> 00:02:15,180 AI never gets fatigued. 60 00:02:15,180 --> 00:02:17,340 AI doesn't, I mean, it makes mistakes, 61 00:02:17,340 --> 00:02:19,845 but eh, for the most part, 62 00:02:19,845 --> 00:02:21,821 it's gonna be better than me. -: Right. 63 00:02:21,821 --> 00:02:23,070 -: Yeah, exactly. (Yan chuckles) 64 00:02:23,070 --> 00:02:25,470 There's like hallucinations and things like that, 65 00:02:25,470 --> 00:02:27,840 but it's not gonna get tired. 66 00:02:27,840 --> 00:02:30,450 It's not going to take a lunch break, 67 00:02:30,450 --> 00:02:33,660 which is very strange facet. 68 00:02:33,660 --> 00:02:35,790 As it starts to work its way more and more 69 00:02:35,790 --> 00:02:38,040 into the modern work life 70 00:02:38,040 --> 00:02:42,030 and the ability to do more of our tasks 71 00:02:42,030 --> 00:02:44,310 that have been reserved just for having 72 00:02:44,310 --> 00:02:46,560 to have the human capabilities to do these tasks 73 00:02:46,560 --> 00:02:50,670 is now so quickly getting taken over by technology. 74 00:02:50,670 --> 00:02:53,329 How do you see it positioning yourself, 75 00:02:53,329 --> 00:02:57,270 your company, in a field that is changing so dramatically? 76 00:02:57,270 --> 00:02:59,836 It's kind of funny, before you answer that, 77 00:02:59,836 --> 00:03:04,626 updating our courses specifically that involve AI, 78 00:03:04,626 --> 00:03:07,290 we are having to reshoot lectures every week, 79 00:03:07,290 --> 00:03:08,123 every other week, 80 00:03:08,123 --> 00:03:09,930 because the layouts, the interfaces, 81 00:03:09,930 --> 00:03:11,730 the capabilities and the functionalities 82 00:03:11,730 --> 00:03:13,830 are all changing so rapidly. 83 00:03:13,830 --> 00:03:16,770 Back to the question, how do you see yourself 84 00:03:16,770 --> 00:03:17,820 positioning the company 85 00:03:17,820 --> 00:03:21,453 that you're growing to surf that wave? 86 00:03:23,040 --> 00:03:23,940 -: That's a great question 87 00:03:23,940 --> 00:03:26,700 and it's something I think about all the time. 88 00:03:26,700 --> 00:03:29,040 Something, me and my co-founder, Nate, 89 00:03:29,040 --> 00:03:30,457 we talk about this all the time. 90 00:03:30,457 --> 00:03:33,060 How do we use the tools 91 00:03:33,060 --> 00:03:34,620 that are available to us to make sure 92 00:03:34,620 --> 00:03:36,723 that we move faster than everyone else? 93 00:03:38,880 --> 00:03:41,880 As an entrepreneur, I'm constantly thinking about the fact 94 00:03:41,880 --> 00:03:45,180 that there's some amount of competi... 95 00:03:45,180 --> 00:03:48,300 It's running a business you're always in the arena. 96 00:03:48,300 --> 00:03:49,174 Like there's someone, 97 00:03:49,174 --> 00:03:52,207 if you don't out-execute the next business, 98 00:03:52,207 --> 00:03:55,110 they're gonna beat you and you're gonna be homeless. 99 00:03:55,110 --> 00:03:57,360 So, you gotta find every edge you can. 100 00:03:57,360 --> 00:03:59,670 And if there's an edge to be had, 101 00:03:59,670 --> 00:04:02,160 it is an imperative that you take that edge 102 00:04:02,160 --> 00:04:03,900 and the biggest edge we have right now 103 00:04:03,900 --> 00:04:06,240 is this changing landscape of artificial intelligence. 104 00:04:06,240 --> 00:04:08,392 So, part of my job, the way I see it nowadays, 105 00:04:08,392 --> 00:04:13,258 is to understand what's possible, stay at the cutting edge. 106 00:04:13,258 --> 00:04:15,480 Maybe some things aren't really there yet. 107 00:04:15,480 --> 00:04:17,550 Logo design was one of 'em where I was like, 108 00:04:17,550 --> 00:04:20,058 okay, we need to get a proper logo designed. 109 00:04:20,058 --> 00:04:23,272 And I felt like all the tools I researched, 110 00:04:23,272 --> 00:04:25,320 it was like, we're not there yet. 111 00:04:25,320 --> 00:04:27,390 For whatever reason, typography isn't, 112 00:04:27,390 --> 00:04:28,980 generative typography is just not, 113 00:04:28,980 --> 00:04:31,290 there's like a lot of rules and the human eyes 114 00:04:31,290 --> 00:04:34,440 expecting a certain thing that we're not there yet. 115 00:04:34,440 --> 00:04:37,320 But I made sure to evaluate it 116 00:04:37,320 --> 00:04:39,404 and it was in the conversation. 117 00:04:39,404 --> 00:04:43,920 Should we try to find some type of AI service designer logo? 118 00:04:43,920 --> 00:04:45,090 So, that's maybe an example 119 00:04:45,090 --> 00:04:46,620 of where it's not a great tool, 120 00:04:46,620 --> 00:04:48,210 but there's plenty of other things. 121 00:04:48,210 --> 00:04:50,580 Like for example, all of our stock imagery, it's like, 122 00:04:50,580 --> 00:04:52,290 I was looking through Shuttershock the other day, 123 00:04:52,290 --> 00:04:53,580 funnily enough, a place I used to work 124 00:04:53,580 --> 00:04:54,810 as a software engineer, 125 00:04:54,810 --> 00:04:58,694 and I thought to myself, jeez, their days are very numbered. 126 00:04:58,694 --> 00:05:01,360 Their business model is predicated on 127 00:05:02,520 --> 00:05:05,430 it being somewhat difficult to find an image 128 00:05:05,430 --> 00:05:07,200 that represents something you want to do. 129 00:05:07,200 --> 00:05:09,780 Now, you can just say, Midjourney, hey, do this thing. 130 00:05:09,780 --> 00:05:11,310 I don't care exactly what it looks like. 131 00:05:11,310 --> 00:05:12,720 I don't have something exact in mind, 132 00:05:12,720 --> 00:05:14,700 but I have a general idea. 133 00:05:14,700 --> 00:05:17,550 Midjourney just kills Shutterstock every time. 134 00:05:17,550 --> 00:05:18,690 And I can't imagine a world 135 00:05:18,690 --> 00:05:21,450 where once that gets far enough along 136 00:05:21,450 --> 00:05:23,130 and that's a mature enough product 137 00:05:23,130 --> 00:05:24,030 that it won't be something 138 00:05:24,030 --> 00:05:26,460 that everyone needs to use every day in order... 139 00:05:26,460 --> 00:05:29,310 And if you're spending money on Shutterstock, 140 00:05:29,310 --> 00:05:31,143 it's like a waste. 141 00:05:33,180 --> 00:05:34,740 We'll see, but maybe I'll be wrong. 142 00:05:34,740 --> 00:05:35,790 And there's always a blowback 143 00:05:35,790 --> 00:05:37,230 and there's always certain edge cases, 144 00:05:37,230 --> 00:05:38,550 but I think for the average business, 145 00:05:38,550 --> 00:05:41,280 it's like, why would you not use Midjourney 146 00:05:41,280 --> 00:05:42,113 or at least try it? 147 00:05:42,113 --> 00:05:43,860 Give it a shot first before... 148 00:05:43,860 --> 00:05:45,210 You reach for Midjourney first, 149 00:05:45,210 --> 00:05:47,700 and then you go for stock imagery. 150 00:05:47,700 --> 00:05:49,710 I think in every facet of our business, 151 00:05:49,710 --> 00:05:51,180 that's something that's constantly on my mind. 152 00:05:51,180 --> 00:05:53,147 Is there an AI tool that's becoming available 153 00:05:53,147 --> 00:05:57,450 or an AI technique that I'm not thinking about 154 00:05:57,450 --> 00:05:58,290 or don't know about yet 155 00:05:58,290 --> 00:06:00,930 that could be a good candidate for solving this problem? 156 00:06:00,930 --> 00:06:01,763 -: Got it. 157 00:06:01,763 --> 00:06:03,090 So, you're finding new ones, 158 00:06:03,090 --> 00:06:07,110 trying to put into the workflow and stay riding that wave. 159 00:06:07,110 --> 00:06:08,550 Things are changing so quickly. 160 00:06:08,550 --> 00:06:11,490 What would take 10 years is now 10 days, essentially. 161 00:06:11,490 --> 00:06:13,110 Basically, 10 months sometimes. 162 00:06:13,110 --> 00:06:17,760 But yeah, that's really when I started to see the impact 163 00:06:17,760 --> 00:06:20,610 of what was happening was when I saw that, 164 00:06:20,610 --> 00:06:22,770 exactly the Shutterstock is a perfect example, 165 00:06:22,770 --> 00:06:24,720 'cause I went through this whole process mentally 166 00:06:24,720 --> 00:06:28,110 as well of the old paradigm is I go 167 00:06:28,110 --> 00:06:30,390 and I pay for something that's available online 168 00:06:30,390 --> 00:06:32,227 that fits in with what I want. 169 00:06:32,227 --> 00:06:33,570 Same thing with Netflix. 170 00:06:33,570 --> 00:06:34,403 I get on Netflix 171 00:06:34,403 --> 00:06:37,890 and I wanna find a show that's a comedy and it's fun 172 00:06:37,890 --> 00:06:41,872 and there's a female lead role, whatever it is. 173 00:06:41,872 --> 00:06:44,040 Now, what can happen with Shutterstock 174 00:06:44,040 --> 00:06:46,440 is you type in exactly what you want on Midjourney. 175 00:06:46,440 --> 00:06:47,970 You don't have to go get a stock image. 176 00:06:47,970 --> 00:06:49,980 It's custom directly to you. 177 00:06:49,980 --> 00:06:51,450 The future of what's gonna be happening 178 00:06:51,450 --> 00:06:54,810 with generative video as well is that you're gonna type in, 179 00:06:54,810 --> 00:06:57,000 I wanna watch a "Star Wars" movie 180 00:06:57,000 --> 00:06:58,950 that has aspects of "Lord of the Rings" 181 00:06:58,950 --> 00:07:02,040 with a strong male lead character. 182 00:07:02,040 --> 00:07:03,360 And it's gonna go, (tongue rolls) 183 00:07:03,360 --> 00:07:05,370 and instead of going through Netflix 184 00:07:05,370 --> 00:07:07,290 for 20 minutes trying to find the movie you want, 185 00:07:07,290 --> 00:07:11,310 you're gonna get generated your custom movie script, 186 00:07:11,310 --> 00:07:12,474 character, styling, 187 00:07:12,474 --> 00:07:16,200 everything in minutes and moments, essentially. 188 00:07:16,200 --> 00:07:18,840 We're not there yet, but that's where it's all going. 189 00:07:18,840 --> 00:07:21,960 And that's why it's an interesting conversation to have 190 00:07:21,960 --> 00:07:25,110 of like staying on the cutting edge of that 191 00:07:25,110 --> 00:07:27,570 is going to, almost need to hire a full-time person. 192 00:07:27,570 --> 00:07:29,580 That's just the future. 193 00:07:29,580 --> 00:07:32,730 They have that job at Samsung, I forget what it's called. 194 00:07:32,730 --> 00:07:33,780 It has a really cool name -: Mm. 195 00:07:33,780 --> 00:07:34,920 -: for the role. 196 00:07:34,920 --> 00:07:36,030 But -: Yeah. 197 00:07:36,030 --> 00:07:37,680 -: it's a really interesting space. 198 00:07:37,680 --> 00:07:42,150 And yeah, I guess that leads me to the question 199 00:07:42,150 --> 00:07:44,040 of the pros and cons 200 00:07:44,040 --> 00:07:47,802 that you're seeing of being in this space. 201 00:07:47,802 --> 00:07:49,680 What's been fun and exciting, 202 00:07:49,680 --> 00:07:51,240 and then what's been also the downside? 203 00:07:51,240 --> 00:07:54,150 -: Oh, it's a really exciting place to be, that's for sure. 204 00:07:54,150 --> 00:07:55,307 It moves really fast. 205 00:07:55,307 --> 00:07:57,360 I think that's the pro and the con. 206 00:07:57,360 --> 00:07:59,059 If you're trying to build a business, 207 00:07:59,059 --> 00:08:02,070 it's very shaky ground to build on, 208 00:08:02,070 --> 00:08:05,610 because for a lot of the, you can see it, 209 00:08:05,610 --> 00:08:08,610 it's like, mm, do I wanna build a, 210 00:08:08,610 --> 00:08:10,740 spend a bunch of time to build something here 211 00:08:10,740 --> 00:08:13,710 and invest in solving this one problem 212 00:08:13,710 --> 00:08:17,742 with the existential threat looming that this gets, 213 00:08:17,742 --> 00:08:22,623 something else, some technology, some AI technique comes in 214 00:08:23,760 --> 00:08:26,845 and obsoletes all the hard work I put into this. 215 00:08:26,845 --> 00:08:29,938 So, I think that is something of a downside 216 00:08:29,938 --> 00:08:33,210 to be in this space and I do worry about that a lot, 217 00:08:33,210 --> 00:08:35,250 but that's a small worry 218 00:08:35,250 --> 00:08:38,730 compared to what I focus mostly on the upside, which is 219 00:08:38,730 --> 00:08:39,870 -: Upside. 220 00:08:39,870 --> 00:08:42,393 -: there's just so much cool stuff that's possible. 221 00:08:43,345 --> 00:08:44,820 And if this technique doesn't work, 222 00:08:44,820 --> 00:08:47,130 like if this business model doesn't work, 223 00:08:47,130 --> 00:08:48,210 we can make a small tweak 224 00:08:48,210 --> 00:08:49,920 and there's another way we might be able 225 00:08:49,920 --> 00:08:53,823 to find a way to build a product that people love.