1 00:00:00,000 --> 00:00:10,000 Welcome to artificially intelligent marketing, a weekly podcast where we stay on top of the 2 00:00:10,000 --> 00:00:15,720 latest trends, tips and tools in the world of marketing AI, helping you get the best 3 00:00:15,720 --> 00:00:18,560 results from your marketing efforts. 4 00:00:18,560 --> 00:00:23,960 Now let's join our hosts, Paul Avery and Martin Broadhurst. 5 00:00:23,960 --> 00:00:30,760 I'm delighted to announce a guest to artificially intelligent marketing this week. 6 00:00:30,760 --> 00:00:38,320 Say hello to Ankur Pandey, CEO and co-founder of Longshot, which is an AI driven content 7 00:00:38,320 --> 00:00:43,840 marketing suite of tools with everything from content generation to content optimization 8 00:00:43,840 --> 00:00:49,260 with some clever sophisticated workflows to make your life easier as a content marketer 9 00:00:49,260 --> 00:00:51,840 baked right into the tool. 10 00:00:51,840 --> 00:00:59,280 Ankur has founded multiple SaaS startups, always in the AI space and launched Longshot 11 00:00:59,280 --> 00:01:04,400 about two and a half years ago, which in the world of AI feels like several millennia, 12 00:01:04,400 --> 00:01:05,400 I guess. 13 00:01:05,400 --> 00:01:07,680 We've seen quite a few changes in that time. 14 00:01:07,680 --> 00:01:09,680 So welcome to the show. 15 00:01:09,680 --> 00:01:10,680 Thanks a lot, Martin. 16 00:01:10,680 --> 00:01:13,040 It's my pleasure to be here on this show. 17 00:01:13,040 --> 00:01:14,040 Excited. 18 00:01:14,040 --> 00:01:16,960 You've been around the content marketing space for a while now. 19 00:01:16,960 --> 00:01:20,840 You've had a tool helping content marketers. 20 00:01:20,840 --> 00:01:25,760 I'm interested to get your thoughts on how that content marketing landscape has really 21 00:01:25,760 --> 00:01:31,920 changed over the last few years in no small part due to AI. 22 00:01:31,920 --> 00:01:37,960 First of all, when it comes to content marketing, by now pretty much every serious content marketers 23 00:01:37,960 --> 00:01:42,040 would use AI in their workflows. 24 00:01:42,040 --> 00:01:47,840 The degrees might differ or where they use it, how they use it, et cetera, but then it 25 00:01:47,840 --> 00:01:50,640 has become pretty universally accepted. 26 00:01:50,640 --> 00:01:57,320 So the biggest change, if I can just call it one change, is the overwhelming acceptance. 27 00:01:57,320 --> 00:02:02,360 What I used to see a couple of years back was that marketers would be a bit skeptical. 28 00:02:02,360 --> 00:02:06,080 And I think I understand the reason. 29 00:02:06,080 --> 00:02:10,720 Over the last decade or so, there had been over-promises in terms of content marketing 30 00:02:10,720 --> 00:02:16,880 space, especially a lot of AI or technology tools promising things which would rather 31 00:02:16,880 --> 00:02:21,720 turn out to be not at least as much as the claims would be. 32 00:02:21,720 --> 00:02:28,760 The initial skepticism was probably justified, but I think by now everybody knows that AI 33 00:02:28,760 --> 00:02:35,040 in content marketing is here to stay and that content marketers of all types, of all sizes 34 00:02:35,040 --> 00:02:39,240 in every niche and every industry would use it. 35 00:02:39,240 --> 00:02:42,600 So that's like the one biggest change I would point out. 36 00:02:42,600 --> 00:02:47,280 What I'm particularly interested in and looking to get into the discussion with you is how 37 00:02:47,280 --> 00:02:55,480 AI sits alongside the latest developments that we've seen in search engine optimisation, 38 00:02:55,480 --> 00:03:02,080 such as Google's helpful content update, which now tells us that you've got to be writing 39 00:03:02,080 --> 00:03:05,920 high quality, helpful content for humans. 40 00:03:05,920 --> 00:03:14,240 And obviously that's got to be done in conjunction with the EAT guidelines as well. 41 00:03:14,240 --> 00:03:22,780 So in that context, in the modern contemporary 2023 content marketing landscape, tell us 42 00:03:22,780 --> 00:03:25,880 how does Longshot work? 43 00:03:25,880 --> 00:03:30,600 I would kind of start from the point of view of content marketers who would be interested 44 00:03:30,600 --> 00:03:31,600 in SEOs. 45 00:03:31,600 --> 00:03:36,120 The answer is likely more direct for those kind of content marketing folks. 46 00:03:36,120 --> 00:03:43,080 The impact on SEO and merely by using AI content was kind of either overblown or was kind of 47 00:03:43,080 --> 00:03:46,840 in some sense a lot of confusions around it and they have gone. 48 00:03:46,840 --> 00:03:49,000 And there are two main reasons for that. 49 00:03:49,000 --> 00:03:53,960 One is that people have understood that or at least learned some tricks of the trade, 50 00:03:53,960 --> 00:03:57,000 like where exactly, how should they use it? 51 00:03:57,000 --> 00:04:01,480 And second thing is that, for example, Google has taken a very open position that initially 52 00:04:01,480 --> 00:04:06,000 there was some kind of confusion around it, but now it was pretty clear that they're not 53 00:04:06,000 --> 00:04:07,800 after AI content, so to speak. 54 00:04:07,800 --> 00:04:14,440 And we are also seeing that specific sort of, you know, ask or other behavior in our customers 55 00:04:14,440 --> 00:04:15,720 or in our users. 56 00:04:15,720 --> 00:04:22,200 So I guess get lesser less confusions with respect to will this AI content not rank or 57 00:04:22,200 --> 00:04:23,840 will that be penalized? 58 00:04:23,840 --> 00:04:27,120 I think that is pretty much negligible. 59 00:04:27,120 --> 00:04:33,600 People have seen that just because AI, the impact was not there in the sense that if 60 00:04:33,600 --> 00:04:37,680 the content was bad, the content did not resonate with the overall domain, etc. 61 00:04:37,680 --> 00:04:44,200 Of course, there may be an SEO sort of downside or there'll be an impact, but not merely because 62 00:04:44,200 --> 00:04:46,360 it was an AI content per se. 63 00:04:46,360 --> 00:04:53,360 Another reason is that we are seeing like, you know, both tools making like my own tool, 64 00:04:53,360 --> 00:05:01,120 we try to, for example, you know, create sophisticated enough workflows so that we tend to create 65 00:05:01,120 --> 00:05:04,640 as much human-like on brand content as possible. 66 00:05:04,640 --> 00:05:07,240 Plus, the AI technology itself has improved. 67 00:05:07,240 --> 00:05:10,440 It has itself become less synthetic over the time. 68 00:05:10,440 --> 00:05:11,920 So I think it's a culmination of many factors. 69 00:05:11,920 --> 00:05:16,880 So I would say unless if you're trying to abuse it, if you're trying to create bulk 70 00:05:16,880 --> 00:05:23,200 content without any oversight and with like a primarily spammy intent, you should watch 71 00:05:23,200 --> 00:05:24,200 it, watch out. 72 00:05:24,200 --> 00:05:29,840 But beyond that, it's reasonable to accept that if the content is good, you're good. 73 00:05:29,840 --> 00:05:31,880 Google's position on that is very clear. 74 00:05:31,880 --> 00:05:37,920 They've said that they won't punish you as long as it is helpful to the readers. 75 00:05:37,920 --> 00:05:40,120 So you touched on your tool there. 76 00:05:40,120 --> 00:05:43,200 Many people listening won't have tried Longshot. 77 00:05:43,200 --> 00:05:44,200 Tell us about it. 78 00:05:44,200 --> 00:05:50,480 So Longshot, I mean, as Martin already, you mentioned that it is a content suite of products. 79 00:05:50,480 --> 00:05:55,600 We observe content marketers and a variety of workflows, primarily long-form content. 80 00:05:55,600 --> 00:06:00,920 That's why the name and we basically try to solve the problem, which was rather confusing 81 00:06:00,920 --> 00:06:02,240 for serious content teams. 82 00:06:02,240 --> 00:06:05,760 For example, think like hallucination, factuality, etc. 83 00:06:05,760 --> 00:06:10,680 So if you're creating content for, let's say a brand, so you would not want content which 84 00:06:10,680 --> 00:06:15,360 would be very generic, would have factual inaccuracies or would hallucinate. 85 00:06:15,360 --> 00:06:20,360 So in order to do that, we developed the workflows, which would try to fix that by a variety of 86 00:06:20,360 --> 00:06:21,360 methods. 87 00:06:21,360 --> 00:06:27,000 So the product is a combination of many features and that had to be the case because there's 88 00:06:27,000 --> 00:06:30,680 no one single use case by content teams. 89 00:06:30,680 --> 00:06:34,600 They do a variety of jobs and there's a field which keeps on evolving. 90 00:06:34,600 --> 00:06:37,120 Therefore, we have to also cater it. 91 00:06:37,120 --> 00:06:42,000 So today, if you go to Longshot, you would be able to write really complex and long content 92 00:06:42,000 --> 00:06:45,040 with an extremely guided approach. 93 00:06:45,040 --> 00:06:50,880 Let's say you want to create a listicle article and you can upload or give inputs your own 94 00:06:50,880 --> 00:06:55,120 context, your own personality, the brand voice, things of that nature. 95 00:06:55,120 --> 00:06:58,080 And then you can guide it along the way and finally optimize. 96 00:06:58,080 --> 00:07:02,340 When I say optimize, it's not just SEO, it's a thing like plagiarism and readability. 97 00:07:02,340 --> 00:07:07,240 So we try to be at a space where it's an overall assistant to a content team. 98 00:07:07,240 --> 00:07:11,680 And then of course, if there are interventions required, there are also features for that. 99 00:07:11,680 --> 00:07:16,040 For example, you can send it to a team review within the product and a lot of people do 100 00:07:16,040 --> 00:07:17,040 that. 101 00:07:17,040 --> 00:07:19,000 Pack their team members, et cetera, et cetera. 102 00:07:19,000 --> 00:07:21,720 That's our goal and we are of course progressing towards that goal. 103 00:07:21,720 --> 00:07:28,320 You were one of the first out of the traps in the industry to really focus on reliable 104 00:07:28,320 --> 00:07:29,320 content. 105 00:07:29,320 --> 00:07:36,200 You spoke a moment ago about inaccuracies and hallucinations and you launched Fact GPT 106 00:07:36,200 --> 00:07:37,200 as it was originally called. 107 00:07:37,200 --> 00:07:41,560 I don't know if the feature is still called Fact GPT now, but you were one of the first 108 00:07:41,560 --> 00:07:42,560 out of the traps with that. 109 00:07:42,560 --> 00:07:48,560 Tell us what was it that you saw in the industry that led you to go down that route? 110 00:07:48,560 --> 00:07:50,800 Fact GPT is a feature inside Longshot. 111 00:07:50,800 --> 00:07:52,640 It is still called Fact GPT. 112 00:07:52,640 --> 00:07:56,800 And of course the feature has since grown to do many other things. 113 00:07:56,800 --> 00:08:02,840 So basically as the name says, we realized that the generic AI tools, including chat 114 00:08:02,840 --> 00:08:08,200 GPTs, et cetera, were really good in creating high quality content styles, but they did 115 00:08:08,200 --> 00:08:11,560 not do that well when it comes to two, three things. 116 00:08:11,560 --> 00:08:15,560 For example, if you want to produce something which is very recent or if you want to upload 117 00:08:15,560 --> 00:08:21,100 your own context, you have an applethora of data, you want to sort of give it to the AI. 118 00:08:21,100 --> 00:08:22,100 This was not easy. 119 00:08:22,100 --> 00:08:25,320 I wouldn't say that it was impossible, but for a content team, there's a lot of work 120 00:08:25,320 --> 00:08:28,200 to be done and a lot of things under the hood. 121 00:08:28,200 --> 00:08:32,800 So we decided to solve this problem just to summarize if you want to make sure the content 122 00:08:32,800 --> 00:08:37,720 is either recent or you want to give your own context or if you want to just focus your 123 00:08:37,720 --> 00:08:39,680 content on some specific websites. 124 00:08:39,680 --> 00:08:46,200 Let's say you trust 10 websites for a specific type of news, a specific type of content. 125 00:08:46,200 --> 00:08:48,920 You want to base your content only on these 10, let's say. 126 00:08:48,920 --> 00:08:50,200 So what do you do? 127 00:08:50,200 --> 00:08:54,480 So those are the kind of things which Fact GPT attempted to solve, still solves. 128 00:08:54,480 --> 00:09:00,200 And we were of course motivated primarily by our users asked that they wanted a method 129 00:09:00,200 --> 00:09:01,200 to do that. 130 00:09:01,200 --> 00:09:04,960 They would initially what we saw that they were producing a generic draft and then they're 131 00:09:04,960 --> 00:09:08,880 in some sense updating the recent fact, right? 132 00:09:08,880 --> 00:09:13,480 So they were doing it in two parts and we thought can't we combine this? 133 00:09:13,480 --> 00:09:15,280 So this was the primary motivation. 134 00:09:15,280 --> 00:09:20,720 We eventually arrived at something which did work and it has been growing since then. 135 00:09:20,720 --> 00:09:22,760 Since then it has grown to be much more sophisticated. 136 00:09:22,760 --> 00:09:26,800 It can handle lots of varieties of data and much more context. 137 00:09:26,800 --> 00:09:32,120 And thanks to both my team's work as well as upgradation of the AI itself, which we 138 00:09:32,120 --> 00:09:33,400 also leverage. 139 00:09:33,400 --> 00:09:37,600 So it's basically the technology itself has evolved and the product therefore has also 140 00:09:37,600 --> 00:09:38,600 evolved. 141 00:09:38,600 --> 00:09:41,200 You mentioned the AI has evolved. 142 00:09:41,200 --> 00:09:45,920 You've built a tool which has connections to external APIs and you've built your own 143 00:09:45,920 --> 00:09:49,720 products on top and you've also developed your own products behind the scenes. 144 00:09:49,720 --> 00:09:55,360 Can you talk to us about bringing the product together and the models that you're using 145 00:09:55,360 --> 00:10:03,880 and how you go about choosing whether to develop your own model or application or integrating 146 00:10:03,880 --> 00:10:06,280 into something that already exists? 147 00:10:06,280 --> 00:10:09,280 I would say it's been an interesting scenario. 148 00:10:09,280 --> 00:10:14,880 The reason is that we leverage AI technologies like OpenAI and now we also leverage others. 149 00:10:14,880 --> 00:10:19,960 For example, Claude Coher, many other open source fine-tuned llamas, et cetera, and some 150 00:10:19,960 --> 00:10:21,480 homegrown ones. 151 00:10:21,480 --> 00:10:27,040 What really happened is that when we were developing it, or I'm basically summarizing 152 00:10:27,040 --> 00:10:30,840 the whole two and a half years journey, we were developing and testing and seeing is 153 00:10:30,840 --> 00:10:33,320 it okay or is it good for our users. 154 00:10:33,320 --> 00:10:37,160 The technologies for whom we were customers were also learning along the way. 155 00:10:37,160 --> 00:10:44,040 So OpenAI of the world and Coher's and Claude's of the world have actually also improved thanks 156 00:10:44,040 --> 00:10:46,760 to their users, including me. 157 00:10:46,760 --> 00:10:52,560 We were, for example, I remember we had lots of deep dives, we still do, with the OpenAI 158 00:10:52,560 --> 00:10:54,440 teams, Claude teams, et cetera. 159 00:10:54,440 --> 00:10:58,160 In that sense, we kind of became a front for these technologies to also go. 160 00:10:58,160 --> 00:11:00,040 It was pretty symbiotic that way. 161 00:11:00,040 --> 00:11:04,400 To answer your question specifically about what models to choose, this was based on both 162 00:11:04,400 --> 00:11:05,720 external tests and internal tests. 163 00:11:05,720 --> 00:11:10,440 We would go back to the research literature, seeing which model performed best, but we 164 00:11:10,440 --> 00:11:13,240 will also do that testing on our own sample. 165 00:11:13,240 --> 00:11:19,600 As our user base grew, we knew where to test, what to test for, where are these issues which 166 00:11:19,600 --> 00:11:20,600 come. 167 00:11:20,600 --> 00:11:25,600 So today when we test, we do not just believe in the openly available benchmarks. 168 00:11:25,600 --> 00:11:30,600 All the AI APIs which we develop, they would have to undergo some kind of testing process 169 00:11:30,600 --> 00:11:32,080 and they will publish these testing process. 170 00:11:32,080 --> 00:11:35,020 They're okay, this did plain X, Y, Z, et cetera. 171 00:11:35,020 --> 00:11:37,720 So that would be a good signal, not the complete signal. 172 00:11:37,720 --> 00:11:38,720 Why? 173 00:11:38,720 --> 00:11:42,040 Because your own specific type of user behavior might differ. 174 00:11:42,040 --> 00:11:46,520 So these testing platforms kind of look at general behavior. 175 00:11:46,520 --> 00:11:49,640 We wanted for our type of users, right? 176 00:11:49,640 --> 00:11:54,400 And we by now had extremely sophisticated testing sets for that. 177 00:11:54,400 --> 00:11:55,400 So we do that. 178 00:11:55,400 --> 00:11:56,400 We have been doing that. 179 00:11:56,400 --> 00:11:59,520 This is actually a pretty ongoing and a very real process. 180 00:11:59,520 --> 00:12:04,560 This is the primary motivator, which AI technologies to use where. 181 00:12:04,560 --> 00:12:09,600 Also I'd like to point out that today in long shot and by proxy, I can also say in many 182 00:12:09,600 --> 00:12:13,640 tools, there's no single sort of AI technology behind a feature. 183 00:12:13,640 --> 00:12:15,480 There is generally a chaining, right? 184 00:12:15,480 --> 00:12:19,040 So for example, you would use, you're not just depend on, for example, GPT-4, you'll 185 00:12:19,040 --> 00:12:24,040 do GPT-4, but for rack pipelines, for example, you would do like embedding search for that. 186 00:12:24,040 --> 00:12:28,520 You will either have something homegrown or, you know, so generally that's like a pipeline 187 00:12:28,520 --> 00:12:32,320 of a small or big AI technologies. 188 00:12:32,320 --> 00:12:38,400 So which has gone right to the user behavior and this is going to grow more sophisticated. 189 00:12:38,400 --> 00:12:44,120 Now with GPT marketplace launched and I think co-pilot side things coming soon, we will 190 00:12:44,120 --> 00:12:45,120 see it more and more. 191 00:12:45,120 --> 00:12:50,160 This chaining of entire things and we are being extremely sophisticated over it. 192 00:12:50,160 --> 00:12:54,080 So excited to see what part we will play with this. 193 00:12:54,080 --> 00:12:59,200 You mentioned there embeddings and embedding searches, and you also mentioned Co-Hare. 194 00:12:59,200 --> 00:13:07,280 Co-Hare is one of the most highly regarded models for search and embeddings. 195 00:13:07,280 --> 00:13:10,440 Is that how you're deploying that within your workflows? 196 00:13:10,440 --> 00:13:11,440 Great. 197 00:13:11,440 --> 00:13:13,880 Co-Hare also we use at few places, not everywhere. 198 00:13:13,880 --> 00:13:19,400 So Co-Hare, as you already said, Co-Hare is pretty highly regarded, but among their set 199 00:13:19,400 --> 00:13:22,680 of AI technologies, some are better than others. 200 00:13:22,680 --> 00:13:25,100 So they have something called Co-Hare reranker. 201 00:13:25,100 --> 00:13:30,960 So now what this reranker does is that if you, let's say, so you already mentioned FAC 202 00:13:30,960 --> 00:13:37,360 GPT, which is a feature where you would upload or link to a lot of data and we'll process 203 00:13:37,360 --> 00:13:38,360 it. 204 00:13:38,360 --> 00:13:40,800 Co-Hare is used somewhere there also. 205 00:13:40,800 --> 00:13:45,920 Does a good job at which document we should look at first or what is that order. 206 00:13:45,920 --> 00:13:48,800 So those kinds of things Co-Hare has done pretty well. 207 00:13:48,800 --> 00:13:50,800 So we use that in Co-Hare. 208 00:13:50,800 --> 00:13:52,680 As I said, this keeps on changing. 209 00:13:52,680 --> 00:13:56,720 If we are speaking two weeks from now, I might have a different answer. 210 00:13:56,720 --> 00:13:58,880 That's really how the nature of the game is. 211 00:13:58,880 --> 00:13:59,880 Absolutely. 212 00:13:59,880 --> 00:14:06,280 It was less than 24 hours ago that Clawed 2.1 was announced with the 200k context window 213 00:14:06,280 --> 00:14:12,200 and improvements and hallucination reduction and all the rest of it. 214 00:14:12,200 --> 00:14:15,240 So yeah, an ever moving landscape. 215 00:14:15,240 --> 00:14:16,240 Ever moving. 216 00:14:16,240 --> 00:14:17,240 For sure. 217 00:14:17,240 --> 00:14:23,280 So just to take it back to the content marketing piece then, you've clearly been working with 218 00:14:23,280 --> 00:14:27,600 lots of customers in all sorts of segments. 219 00:14:27,600 --> 00:14:34,400 Last time we spoke, I know you were working with journalists and news teams, marketing 220 00:14:34,400 --> 00:14:36,600 departments with education teams. 221 00:14:36,600 --> 00:14:37,600 I mean, you name it. 222 00:14:37,600 --> 00:14:41,760 I think you were just about across it. 223 00:14:41,760 --> 00:14:52,600 What learnings could you impart to our listeners about how to think about deploying AI tools 224 00:14:52,600 --> 00:14:54,840 within the content marketing workflow? 225 00:14:54,840 --> 00:15:00,120 First of all, I'll give you the answers which we are more clear on because this is also 226 00:15:00,120 --> 00:15:03,480 something which is bound to be a bit opinionated. 227 00:15:03,480 --> 00:15:08,280 But something which is less opinionated, I would say that if your content style or the 228 00:15:08,280 --> 00:15:14,840 kind of content you're writing has sensitive to a few things, for example, you are in FinTech 229 00:15:14,840 --> 00:15:18,120 or you could be in compliance, legal, areas like that. 230 00:15:18,120 --> 00:15:24,400 So you have to, by definition, make sure that this veracity and hallucination are checked. 231 00:15:24,400 --> 00:15:25,920 That's maybe the first thing you should do. 232 00:15:25,920 --> 00:15:30,000 The style and the creativity is maybe take a backseat here. 233 00:15:30,000 --> 00:15:34,920 Therefore, for example, if you are picking a tool, I mean, I would do whatever tool it 234 00:15:34,920 --> 00:15:35,920 is. 235 00:15:35,920 --> 00:15:38,640 So make sure that it's that prioritized. 236 00:15:38,640 --> 00:15:41,120 And similarly, then there are lots of use cases. 237 00:15:41,120 --> 00:15:44,960 We see that just after that, there are some other guidelines. 238 00:15:44,960 --> 00:15:49,400 Let's say we have teams which say that they want specific guidelines to be met. 239 00:15:49,400 --> 00:15:53,320 It could start from that they should have a specific tonality of the kind of content 240 00:15:53,320 --> 00:15:54,880 they have previously read. 241 00:15:54,880 --> 00:15:59,960 Now, so just understand that this is a bit complex because it's not like they know a 242 00:15:59,960 --> 00:16:00,960 name for this tonality. 243 00:16:00,960 --> 00:16:04,960 Let's say if I say write in the way how Martin has read. 244 00:16:04,960 --> 00:16:08,560 It's of course hard for anyone who has not read Martin to understand. 245 00:16:08,560 --> 00:16:15,160 So we have to, we do not have luxury to just know by if you just mention all the same. 246 00:16:15,160 --> 00:16:17,640 The style has to be in some sense discovered. 247 00:16:17,640 --> 00:16:23,000 So this sometimes this becomes important for teens who are very careful about brand voice. 248 00:16:23,000 --> 00:16:26,520 And then I, as I said about brand guidelines, so guidelines, example guidelines could be 249 00:16:26,520 --> 00:16:34,560 that for example, that they don't want abbreviations or they don't want any sort of names which 250 00:16:34,560 --> 00:16:37,240 or any legal technologies to be shared or things like that. 251 00:16:37,240 --> 00:16:41,920 So now the point is that if these kinds of things you can always do by hand, right, as 252 00:16:41,920 --> 00:16:43,440 traditionally has been done. 253 00:16:43,440 --> 00:16:46,920 But when you're selecting, just see that is that prioritized or not. 254 00:16:46,920 --> 00:16:49,400 So I think that could be the kind of answer. 255 00:16:49,400 --> 00:16:50,400 I'm pretty different. 256 00:16:50,400 --> 00:16:55,800 The reason is that we also tend to attract such users more because of course, just if 257 00:16:55,800 --> 00:16:58,840 you are interested to just write the content, right. 258 00:16:58,840 --> 00:17:03,260 If you have some quick idea and write a tweet, then I think that's a pretty co-ordination. 259 00:17:03,260 --> 00:17:04,840 You don't have to even use any tool. 260 00:17:04,840 --> 00:17:05,840 You just go on. 261 00:17:05,840 --> 00:17:08,520 I think Bing's and Google's of the world would solve it for you. 262 00:17:08,520 --> 00:17:09,520 Right. 263 00:17:09,520 --> 00:17:15,680 So there, but it depends on the sophistication and sensitivity or curiosity. 264 00:17:15,680 --> 00:17:20,120 On the flip side, I would say that if you other things which are like, which also matter, 265 00:17:20,120 --> 00:17:24,720 which we see, for example, we cater primarily in English, sometimes also in, I mean, in 266 00:17:24,720 --> 00:17:26,880 few major languages. 267 00:17:26,880 --> 00:17:29,920 And the reason being that some other languages like Arabic and even Hindi, right. 268 00:17:29,920 --> 00:17:32,280 Indian languages are a bit hard to tame. 269 00:17:32,280 --> 00:17:34,160 I mean, now it is improving. 270 00:17:34,160 --> 00:17:37,400 What I'm seeing is that there are dedicated tools coming for languages. 271 00:17:37,400 --> 00:17:38,720 Maybe that will grow. 272 00:17:38,720 --> 00:17:40,720 So that also kind of factors in. 273 00:17:40,720 --> 00:17:44,680 So I think there are a few angles one which has to be vigilant about. 274 00:17:44,680 --> 00:17:50,040 You've also introduced within the tool and you referenced this earlier, the more collaborative 275 00:17:50,040 --> 00:17:53,360 the team based features. 276 00:17:53,360 --> 00:17:58,800 Is there anything there that you think teams need to consider? 277 00:17:58,800 --> 00:18:04,640 One thing personally that I'm very keen on at the moment is pushing just the basic AI 278 00:18:04,640 --> 00:18:05,640 literacy. 279 00:18:05,640 --> 00:18:07,400 What are the models? 280 00:18:07,400 --> 00:18:09,080 What are the limitations of the models? 281 00:18:09,080 --> 00:18:11,120 What is a hallucination? 282 00:18:11,120 --> 00:18:12,120 Things like that. 283 00:18:12,120 --> 00:18:19,580 And so any other considerations that teams or leaders of teams and businesses might want 284 00:18:19,580 --> 00:18:24,360 to consider when deploying AI technologies in marketing? 285 00:18:24,360 --> 00:18:25,360 Yeah. 286 00:18:25,360 --> 00:18:27,200 So that's a good question. 287 00:18:27,200 --> 00:18:29,120 I'm not sure if I have a direct answer to that. 288 00:18:29,120 --> 00:18:36,160 I think part of it is that I'm actually surprised and very impressed that content team leaders, 289 00:18:36,160 --> 00:18:40,280 content heads of content, they are themselves pretty much on, it might not have been the 290 00:18:40,280 --> 00:18:46,440 case a year back, but now they are themselves dated at least to a great extent on a lot 291 00:18:46,440 --> 00:18:48,720 in some sense in context of AI literacy. 292 00:18:48,720 --> 00:18:50,560 I think they are catching up fast. 293 00:18:50,560 --> 00:18:55,360 I'm not sure if as a tool we have a part to play there, but that kind of thing has happened 294 00:18:55,360 --> 00:18:56,360 post-ChatGPT. 295 00:18:56,360 --> 00:18:59,320 I think people start to take it very seriously. 296 00:18:59,320 --> 00:19:04,100 Everybody and thanks in part to influencers, they did hype the thing, but then everybody 297 00:19:04,100 --> 00:19:06,440 and their mother decided to give it a shot. 298 00:19:06,440 --> 00:19:11,400 So that basically became like a, you know, when a lot of people did it for the FOMO, 299 00:19:11,400 --> 00:19:12,400 right? 300 00:19:12,400 --> 00:19:14,520 That they have to understand what's going on. 301 00:19:14,520 --> 00:19:19,200 So in no less part to all the constant drama, which generally open AI surrounds themselves 302 00:19:19,200 --> 00:19:20,680 with, including the recent one. 303 00:19:20,680 --> 00:19:23,000 So in India, for example, there is a world cup going on, right? 304 00:19:23,000 --> 00:19:26,240 So I was seeing the chat, a parallel chat between two things, right? 305 00:19:26,240 --> 00:19:29,560 The world cup, the cricket world cup that is, and the open AI, right? 306 00:19:29,560 --> 00:19:34,280 At least in the tech interested community, which was unprecedented. 307 00:19:34,280 --> 00:19:39,240 So yeah, I would say that a lot of evangelism has happened because of marketing and a lot 308 00:19:39,240 --> 00:19:41,240 of public understanding already. 309 00:19:41,240 --> 00:19:44,040 I think it would as any other technology, right? 310 00:19:44,040 --> 00:19:48,280 I think it would be in terms of what has to be, what should be prioritized and what should 311 00:19:48,280 --> 00:19:49,840 not be prioritized. 312 00:19:49,840 --> 00:19:53,600 That has to be also solved, I think, apart from the marketing language. 313 00:19:53,600 --> 00:19:58,160 I'm not sure what do I have to, what would my role or at least product role would be 314 00:19:58,160 --> 00:19:59,160 there. 315 00:19:59,160 --> 00:20:00,720 I think that would be in some educators role. 316 00:20:00,720 --> 00:20:05,320 But one thing I would definitely say, any new technology, of course, adds to confusion. 317 00:20:05,320 --> 00:20:08,520 But if you compare it to a few other technologies, which has happened in the past, things like 318 00:20:08,520 --> 00:20:10,920 blockchain and all, it is still much less confusing. 319 00:20:10,920 --> 00:20:13,560 I would say things are much more straightforward. 320 00:20:13,560 --> 00:20:15,360 You can see for yourself it is working. 321 00:20:15,360 --> 00:20:17,280 For example, in a blockchain world, right? 322 00:20:17,280 --> 00:20:20,640 It's not trivial for anybody to just start playing with it, but here you can do it. 323 00:20:20,640 --> 00:20:23,160 So you can in some sense see for yourself. 324 00:20:23,160 --> 00:20:31,440 So because of that very nature of things, I think the education has become much easier. 325 00:20:31,440 --> 00:20:33,280 Users can get value from it immediately. 326 00:20:33,280 --> 00:20:36,880 The time to value is very, very quick, isn't it? 327 00:20:36,880 --> 00:20:40,920 And like you say, there's no technical requirement. 328 00:20:40,920 --> 00:20:46,640 I can ask ChatGPT to write me a tweet about a topic and it will do it. 329 00:20:46,640 --> 00:20:47,640 And that's great. 330 00:20:47,640 --> 00:20:55,120 I say on the flip side, an NFT, I can buy a JPEG of a monkey. 331 00:20:55,120 --> 00:20:56,120 So what? 332 00:20:56,120 --> 00:20:57,120 Right? 333 00:20:57,120 --> 00:20:59,080 Where's the value there? 334 00:20:59,080 --> 00:21:02,520 Not that I'm dumping on the NFT and blockchain community. 335 00:21:02,520 --> 00:21:04,560 I would never do that. 336 00:21:04,560 --> 00:21:07,000 You mentioned there the open AI drama. 337 00:21:07,000 --> 00:21:13,360 Are you confident that your board are not going to sack you and then rehire you in four 338 00:21:13,360 --> 00:21:14,360 days? 339 00:21:14,360 --> 00:21:16,480 Yeah, I hope not. 340 00:21:16,480 --> 00:21:20,040 And the reason is that we don't have that kind of board, right? 341 00:21:20,040 --> 00:21:23,640 Which would have such a drastic disagreement. 342 00:21:23,640 --> 00:21:25,840 I think this is really one of them. 343 00:21:25,840 --> 00:21:29,760 This happens once in 30, 40 years, maybe like last time it happened was Steve Jobs, right? 344 00:21:29,760 --> 00:21:35,080 I don't think that founders or CEOs start doubting their boards. 345 00:21:35,080 --> 00:21:38,000 So mostly this will be one off. 346 00:21:38,000 --> 00:21:39,000 Yeah, hopefully. 347 00:21:39,000 --> 00:21:43,320 I think it does go to show the value of having a good board though, an experienced board 348 00:21:43,320 --> 00:21:44,320 for sure. 349 00:21:44,320 --> 00:21:49,200 Yeah, but I would definitely add this actually taught a lot of us and it's not like this. 350 00:21:49,200 --> 00:21:56,120 We were all prepared that no matter how big or how influential a particular, let's say, 351 00:21:56,120 --> 00:22:02,280 company becomes, our job is not to just entirely depend on open AI's of the world, right? 352 00:22:02,280 --> 00:22:07,480 Either by leveraging other elements or by using our homegrown versions, by fine tuning 353 00:22:07,480 --> 00:22:09,320 open sources, et cetera. 354 00:22:09,320 --> 00:22:15,120 But that is definitely what a general lesson for a builder in this space would be. 355 00:22:15,120 --> 00:22:20,000 So that I would definitely, in some sense, I would update my, this is something we already 356 00:22:20,000 --> 00:22:26,240 had a bit of careful about, but this kind of, I updated the sort of seriousness of this. 357 00:22:26,240 --> 00:22:34,400 I saw several people on threads at the weekend saying that they were switching over their 358 00:22:34,400 --> 00:22:42,520 open AI APIs and applying to the Microsoft Azure models. 359 00:22:42,520 --> 00:22:44,480 So they could just do a light replacement. 360 00:22:44,480 --> 00:22:51,080 They felt much more confident using it through Azure than open AI given the drama. 361 00:22:51,080 --> 00:22:54,840 Whether that's warranted or not, I don't know, but I agree. 362 00:22:54,840 --> 00:23:02,000 I think this gives developers cause for concern and to be aware of the over-reliance on one's 363 00:23:02,000 --> 00:23:04,720 supplier for sure. 364 00:23:04,720 --> 00:23:09,320 Looking ahead to the AI landscape, things are changing rapidly. 365 00:23:09,320 --> 00:23:13,040 We've seen in the two and a half years that your product has been on the market, things 366 00:23:13,040 --> 00:23:17,800 have evolved quickly and the interest in the space has obviously gone through the roof 367 00:23:17,800 --> 00:23:20,560 in the past 12 months. 368 00:23:20,560 --> 00:23:26,720 Fast forward 12, 24 months, where do you see the space heading? 369 00:23:26,720 --> 00:23:28,000 What's next for us? 370 00:23:28,000 --> 00:23:29,000 Yeah. 371 00:23:29,000 --> 00:23:33,440 This is something which I just had an opportunity to speak on just a few days back. 372 00:23:33,440 --> 00:23:36,880 So it was not exactly about content marketing, but it was content management system, which 373 00:23:36,880 --> 00:23:39,080 also has extraordinarily like overlap, right? 374 00:23:39,080 --> 00:23:43,200 So I would base the answer from there and whatever I learned there. 375 00:23:43,200 --> 00:23:47,240 So I would say that there are some really interesting trends which has come up, right? 376 00:23:47,240 --> 00:23:52,480 So the role of content would not be, would drastically change in the sense that, so what 377 00:23:52,480 --> 00:23:55,600 does, for example, what, how does a marketer treat content? 378 00:23:55,600 --> 00:23:57,960 Content is but a medium to inform your users, right? 379 00:23:57,960 --> 00:23:58,960 Things like that. 380 00:23:58,960 --> 00:24:02,760 So it's like you are putting some piece out there, somebody's reading it, understanding 381 00:24:02,760 --> 00:24:05,560 a few things or getting evangelized or stuff like that, right? 382 00:24:05,560 --> 00:24:08,920 So this whole philosophy might be updated. 383 00:24:08,920 --> 00:24:11,480 For example, this might sound futuristic. 384 00:24:11,480 --> 00:24:14,200 When we say futuristic, it might be like six words on the line, right? 385 00:24:14,200 --> 00:24:19,120 So whatever you are reading a blog and the blog updates itself on the real time, right? 386 00:24:19,120 --> 00:24:21,400 Based on what your signals are. 387 00:24:21,400 --> 00:24:25,120 I'm not saying that you have to chat everything and ask because sometimes when you try to 388 00:24:25,120 --> 00:24:26,240 ask, you don't know what to ask. 389 00:24:26,240 --> 00:24:28,120 You don't know what you don't know, right? 390 00:24:28,120 --> 00:24:33,240 But it's like taking cues of your sort of past historical data and what have you liked, 391 00:24:33,240 --> 00:24:35,280 where are your eye balls on things like that, right? 392 00:24:35,280 --> 00:24:40,640 So you would probably see a scenario where maybe like there'll be a draft content, but 393 00:24:40,640 --> 00:24:45,440 beyond that is truly personalized, right, with respect to this is, you know, I was also 394 00:24:45,440 --> 00:24:49,040 astonished when I kind of learned, but some people are taking this seriously, right? 395 00:24:49,040 --> 00:24:54,700 So you know, CMS of the world are trying to update themselves, what if this happens? 396 00:24:54,700 --> 00:24:55,700 What will their role be? 397 00:24:55,700 --> 00:24:57,760 How will you store the content, right? 398 00:24:57,760 --> 00:24:59,320 Or those kind of things. 399 00:24:59,320 --> 00:25:04,040 Second thing I would say is that I think in order to truly stand out, two types of context 400 00:25:04,040 --> 00:25:05,040 would matter. 401 00:25:05,040 --> 00:25:09,420 Your context as in the reader's context, which I already touched upon in the last point, 402 00:25:09,420 --> 00:25:10,980 as well as the brand context. 403 00:25:10,980 --> 00:25:11,980 It always used to matter. 404 00:25:11,980 --> 00:25:16,120 You have to, people would be interested in the messaging you have, et cetera, et cetera. 405 00:25:16,120 --> 00:25:18,120 It still works like that. 406 00:25:18,120 --> 00:25:21,840 But then I think it will go much deeper. 407 00:25:21,840 --> 00:25:27,920 I would not be surprised if every medium to large companies, when they use an AI tool, 408 00:25:27,920 --> 00:25:33,120 either a tool or AI directly, they have fine tuned it or adapted their use case. 409 00:25:33,120 --> 00:25:38,080 So the genericness will be taken care of. 410 00:25:38,080 --> 00:25:43,320 We just were in talks with one big media player and they had guidelines like this, for example. 411 00:25:43,320 --> 00:25:50,880 So both, they are like a media company and when they would talk about any person, right? 412 00:25:50,880 --> 00:25:56,440 Depending on what kind of content it is, they would shun away from writing anything negative, 413 00:25:56,440 --> 00:25:57,440 right? 414 00:25:57,440 --> 00:25:59,520 But at the same time, they also have to present the facts. 415 00:25:59,520 --> 00:26:02,040 So let's say somebody did cram it, right? 416 00:26:02,040 --> 00:26:05,320 All those things, it's a bit hard to capture today, right? 417 00:26:05,320 --> 00:26:07,800 And so we were trying to see what do you really mean? 418 00:26:07,800 --> 00:26:12,440 It's hard for users or humans to tell what they mean, but rather than show what they 419 00:26:12,440 --> 00:26:13,440 mean. 420 00:26:13,440 --> 00:26:15,640 And the last point I would like to say is this. 421 00:26:15,640 --> 00:26:19,720 I would also say that there'll be innovations on how users interact with all this, but maybe 422 00:26:19,720 --> 00:26:23,960 there is something which is completely unthought of in terms of how people interact with the 423 00:26:23,960 --> 00:26:26,280 AI first content tool. 424 00:26:26,280 --> 00:26:29,080 It can be something which we have not even arrived. 425 00:26:29,080 --> 00:26:36,200 So I think even designers have a role to play here because maybe the true value of AI is 426 00:26:36,200 --> 00:26:41,080 not hidden in the better model, but it might be hidden in the time. 427 00:26:41,080 --> 00:26:43,800 Like how do users interact with AI? 428 00:26:43,800 --> 00:26:46,560 So for example, chatgbt surprised everyone, including OpenAI. 429 00:26:46,560 --> 00:26:49,560 We're on record saying this. 430 00:26:49,560 --> 00:26:50,760 What comes after chatgbt? 431 00:26:50,760 --> 00:26:52,420 What is the next surprise? 432 00:26:52,420 --> 00:26:54,240 And I am confident that this will happen. 433 00:26:54,240 --> 00:26:58,160 And content marketing is a world where you see the first innovations, right? 434 00:26:58,160 --> 00:27:00,960 Because there are huge and very horizontal landscape, right? 435 00:27:00,960 --> 00:27:06,800 So this will definitely be one of the first areas to both get experimented upon as well 436 00:27:06,800 --> 00:27:09,640 as, you know, absorb it. 437 00:27:09,640 --> 00:27:17,360 Cathy Kozakov, the chief decision scientist at Google who recently left, she wrote earlier 438 00:27:17,360 --> 00:27:23,760 this year, a blog post that said the latest AI revolution is not an AI revolution. 439 00:27:23,760 --> 00:27:24,760 It's a design revolution. 440 00:27:24,760 --> 00:27:25,760 Yeah. 441 00:27:25,760 --> 00:27:27,400 And it was all about the user interface. 442 00:27:27,400 --> 00:27:30,260 And so I 100% agree with that. 443 00:27:30,260 --> 00:27:35,240 It was fascinating to see how getting a large language model into an interface, just like 444 00:27:35,240 --> 00:27:42,120 WhatsApp suddenly changed the appetite and demand for these technologies. 445 00:27:42,120 --> 00:27:43,380 Hey, who knows? 446 00:27:43,380 --> 00:27:50,360 Maybe the next frontier for content marketing is the humane wearable and we'll all be creating 447 00:27:50,360 --> 00:27:53,880 content through that kind of interface. 448 00:27:53,880 --> 00:27:54,880 Yeah. 449 00:27:54,880 --> 00:27:57,560 That's a dystopian idea. 450 00:27:57,560 --> 00:27:58,560 Let's see. 451 00:27:58,560 --> 00:27:59,560 I'm not judging. 452 00:27:59,560 --> 00:28:06,160 I'm not fully sold on that as a concept just yet. 453 00:28:06,160 --> 00:28:07,960 Let's look ahead for Longshot. 454 00:28:07,960 --> 00:28:08,960 What are you working on? 455 00:28:08,960 --> 00:28:12,160 What can you give us a peek behind the curtain on? 456 00:28:12,160 --> 00:28:17,840 So, so in Longshot, you know, what we are very bullish on the one aspect which we are 457 00:28:17,840 --> 00:28:21,560 convinced will be the next creation is the brand voice thing. 458 00:28:21,560 --> 00:28:23,560 Brand voice, our definitions also changed. 459 00:28:23,560 --> 00:28:28,760 We used to think that it's just the style of tonality, but it has actually evolved to 460 00:28:28,760 --> 00:28:32,440 multiple guidelines, rules, and the space keeps evolving. 461 00:28:32,440 --> 00:28:37,640 And of course, we now, we are actually working on a way that the most complex or the even 462 00:28:37,640 --> 00:28:42,560 most tracked, the most vague sort of, you know, ask can also be in some sense accepted 463 00:28:42,560 --> 00:28:43,560 by the voice. 464 00:28:43,560 --> 00:28:47,520 Like there is a feature or there is like a section of Longshot which is called auto blog 465 00:28:47,520 --> 00:28:54,920 workflows wherein we try to create blogs in nearly, I would not say 100% automated fashion, 466 00:28:54,920 --> 00:28:56,640 but nearly automated fashion. 467 00:28:56,640 --> 00:28:58,960 Like this is also proved to be valuable. 468 00:28:58,960 --> 00:29:04,200 So rather than creating one generic template, we tend to give like a lot of popular templates. 469 00:29:04,200 --> 00:29:08,680 For example, listicles and some types of blog, let's say FinTech blog would have a slightly 470 00:29:08,680 --> 00:29:10,440 different style and things like that. 471 00:29:10,440 --> 00:29:12,880 But then what if you don't even find all that? 472 00:29:12,880 --> 00:29:16,600 What if you have a style of your own or context of your own and then you want to create your 473 00:29:16,600 --> 00:29:17,600 own pattern? 474 00:29:17,600 --> 00:29:19,520 So that all those things are working on. 475 00:29:19,520 --> 00:29:23,240 I would also say finally that we have seen a lot of interest in APIs. 476 00:29:23,240 --> 00:29:28,400 A lot of people, especially larger companies for whatever reasons, they want to use kind 477 00:29:28,400 --> 00:29:33,960 of solutions we offer, but maybe in their own CMSs, in their own environment. 478 00:29:33,960 --> 00:29:39,080 So for that we had already brought up APIs which are used by lots of companies now, but 479 00:29:39,080 --> 00:29:41,040 this is also going to be an increased focus. 480 00:29:41,040 --> 00:29:44,400 So you don't have to depend on using Longshot, right? 481 00:29:44,400 --> 00:29:49,800 You can actually plug Longshot's APIs in your own systems and it will work like in C2. 482 00:29:49,800 --> 00:29:52,320 So those are the two, three main areas. 483 00:29:52,320 --> 00:29:56,600 We would be focusing on and for the readers who are interested, I don't intend to, as 484 00:29:56,600 --> 00:29:59,320 I said, be very focused on too much Longshot. 485 00:29:59,320 --> 00:30:00,320 You can see for yourself. 486 00:30:00,320 --> 00:30:04,960 If anybody's interested, you can just go to Longshot, get a free trial and I'll be happy 487 00:30:04,960 --> 00:30:07,880 to hear your feedback of any type. 488 00:30:07,880 --> 00:30:12,120 It is his own, you know, or critiques or praise, whatever, you know, bring it on. 489 00:30:12,120 --> 00:30:13,680 I would encourage people to check it out. 490 00:30:13,680 --> 00:30:15,160 I've been a long time user. 491 00:30:15,160 --> 00:30:19,880 I've seen the product evolve a fair amount over the past couple of years and looking 492 00:30:19,880 --> 00:30:21,920 forward to seeing where it heads. 493 00:30:21,920 --> 00:30:27,240 So actually building some tools with the API, getting that integrated into some of my own 494 00:30:27,240 --> 00:30:28,880 workflows as well. 495 00:30:28,880 --> 00:30:33,400 How do you feel about the future of content marketing and AI? 496 00:30:33,400 --> 00:30:36,960 What can you leave us with before we say goodbye? 497 00:30:36,960 --> 00:30:42,120 So I would say that I think if there are still folks who are skeptical of AI technologies, 498 00:30:42,120 --> 00:30:43,360 I think you should embrace it. 499 00:30:43,360 --> 00:30:48,160 I think the issue is that it will replace the job is overblown and it will. 500 00:30:48,160 --> 00:30:52,840 The point is that replacing job can be seen as a welcome change to get on to new things, 501 00:30:52,840 --> 00:30:53,840 right? 502 00:30:53,840 --> 00:30:57,800 In the sense that even in this very space, content marketing has never been a space which 503 00:30:57,800 --> 00:31:00,240 was set in stone, right? 504 00:31:00,240 --> 00:31:03,940 Every year it will go to any of these subject to changes, right? 505 00:31:03,940 --> 00:31:09,320 So I think, are you, and the reason I'm bringing it up is that we had the fair amount of such 506 00:31:09,320 --> 00:31:10,320 interaction previously. 507 00:31:10,320 --> 00:31:12,720 It has gone drastically down now. 508 00:31:12,720 --> 00:31:17,240 So my appeal would be to embrace it and buy by your own expertise with the products like 509 00:31:17,240 --> 00:31:18,800 us will also improve, right? 510 00:31:18,800 --> 00:31:19,800 So that's the whole point. 511 00:31:19,800 --> 00:31:24,720 If there are some legit concerns which product builders like ourselves have missed out so 512 00:31:24,720 --> 00:31:29,920 far, this is our opportunity to embrace them. 513 00:31:29,920 --> 00:31:35,800 Fantastic and optimistic and call to arms for content marketers to embrace AI. 514 00:31:35,800 --> 00:31:40,520 If people are interested in getting started with Longshot, they can check it out. 515 00:31:40,520 --> 00:31:46,040 It's longshot.ai and you get access to the full suite of tools via there. 516 00:31:46,040 --> 00:31:49,560 Thank you very much for joining me today. 517 00:31:49,560 --> 00:31:50,560 Thanks a lot, Martin. 518 00:31:50,560 --> 00:31:56,640 So it was a lovely talk and hope to exchange notes soon in future. 519 00:31:56,640 --> 00:31:57,640 Fantastic. 520 00:31:57,640 --> 00:31:58,640 Cheers, bye. 521 00:31:58,640 --> 00:32:02,640 Thank you for listening to Artificially Intelligent Marketing. 522 00:32:02,640 --> 00:32:08,680 To stay on top of the latest trends, tips and tools in the world of marketing AI, be 523 00:32:08,680 --> 00:32:10,440 sure to subscribe. 524 00:32:10,440 --> 00:32:17,760 We look forward to seeing you again next week.