WEBVTT 00:00:00.000 --> 00:00:02.940 My name is Janine Wright and I am the CEO of 00:00:02.940 --> 00:00:05.400 Inception Point AI. And what does Inception Point 00:00:05.400 --> 00:00:08.699 AI do and how do you make shows? Inception Point 00:00:08.699 --> 00:00:13.220 AI designs AI -generated personalities and then 00:00:13.220 --> 00:00:17.059 launches them into the world as 360 creators 00:00:17.059 --> 00:00:21.219 and we create AI -generated content on their 00:00:21.219 --> 00:00:25.010 behalf across multiple different platforms. And 00:00:25.010 --> 00:00:28.329 we've started and grown up in audio. In the process, 00:00:28.609 --> 00:00:31.670 we built what we believe to be the world's largest 00:00:31.670 --> 00:00:35.729 independent podcast company by Showcount. From 00:00:35.729 --> 00:00:39.909 the reviews that I've seen of your work, I'm 00:00:39.909 --> 00:00:41.649 wondering whether you're being judged a bit too 00:00:41.649 --> 00:00:44.009 much on the early stuff which is there and not 00:00:44.009 --> 00:00:47.170 necessarily some of the stuff which is there 00:00:47.170 --> 00:00:49.409 now. I mean, I think that's fair. I mean, obviously, 00:00:49.469 --> 00:00:53.090 that's part of what I said in my piece that we 00:00:53.090 --> 00:00:57.259 posted. Yeah, I mean, at this point, we've created 00:00:57.259 --> 00:01:02.939 more than 160 ,000 episodes. And we've been doing 00:01:02.939 --> 00:01:05.219 it for more than two years. We were very early 00:01:05.219 --> 00:01:09.879 in experimenting with AI. This was early in our 00:01:09.879 --> 00:01:12.260 own capabilities and early in AI's capabilities. 00:01:12.760 --> 00:01:15.359 And a lot of the stuff at the beginning, I mean, 00:01:15.359 --> 00:01:17.540 frankly, it really sucks. You know, we go back 00:01:17.540 --> 00:01:21.859 and listen to it and we cringe. But I understand 00:01:21.859 --> 00:01:25.140 this is a very... I've been working with creators 00:01:25.140 --> 00:01:28.359 in the podcasting space for many years, and pretty 00:01:28.359 --> 00:01:31.260 much without fail, everyone is embarrassed about 00:01:31.260 --> 00:01:34.980 their first episodes, their first season, right? 00:01:35.079 --> 00:01:38.939 Because it takes a while to get practice and 00:01:38.939 --> 00:01:43.859 to get good. But I've also been candid that even 00:01:43.859 --> 00:01:46.200 some of the stuff that we make today might not 00:01:46.200 --> 00:01:49.159 be good, right? We still do a lot of experimentation. 00:01:51.260 --> 00:01:53.340 Some of it lands and some of it doesn't. And 00:01:53.340 --> 00:01:55.799 some of it, some of the stuff I really love and 00:01:55.799 --> 00:01:57.540 some of the stuff doesn't do it for me, but it 00:01:57.540 --> 00:02:00.500 does it for other people. And all of those things 00:02:00.500 --> 00:02:02.799 are okay. I'm kind of looking at some of the 00:02:02.799 --> 00:02:06.900 ones that you have been promoting recently. And, 00:02:07.079 --> 00:02:11.539 you know, there are very few reviews, listener 00:02:11.539 --> 00:02:15.379 reviews on there. There's one listener review 00:02:15.379 --> 00:02:18.639 for Bloom and Banter, which is hosted by Nigel 00:02:18.639 --> 00:02:22.060 Thistledown, somebody with an accent almost as 00:02:22.060 --> 00:02:28.400 ridiculous as mine. And that review isn't particularly 00:02:28.400 --> 00:02:34.419 positive, shall we say. I'm wondering how many 00:02:34.419 --> 00:02:37.460 people are, I mean, you know, frankly, is this 00:02:37.460 --> 00:02:42.360 an SEO play? Are you there flooding podcast directories 00:02:42.360 --> 00:02:46.599 with... content um and hoping that people um 00:02:46.599 --> 00:02:51.340 find that content um you know in terms of seo 00:02:51.340 --> 00:02:53.879 or are you actually seeing people coming back 00:02:53.879 --> 00:02:55.860 again and again uh well we're definitely seeing 00:02:55.860 --> 00:02:58.000 people coming back again and again which is why 00:02:58.000 --> 00:02:59.860 we're leaning into many of our personalities 00:02:59.860 --> 00:03:02.719 um i don't know if it's fair to say like flooding 00:03:02.719 --> 00:03:05.560 with seo there's lots of businesses that make 00:03:05.560 --> 00:03:11.439 content based on uh what people are wanting and 00:03:11.439 --> 00:03:13.979 searching for. And perhaps that's been a problem 00:03:13.979 --> 00:03:15.780 with the way that we've been doing podcasting 00:03:15.780 --> 00:03:18.020 in the past. I mean, at some of the companies 00:03:18.020 --> 00:03:21.500 that I was at previously, you would make this 00:03:21.500 --> 00:03:26.939 beautiful, brilliant, incredible show and then 00:03:26.939 --> 00:03:29.080 title it something that nobody has ever heard 00:03:29.080 --> 00:03:32.289 of. And then you would have to spend... huge 00:03:32.289 --> 00:03:34.750 amounts of money to invest in telling people 00:03:34.750 --> 00:03:37.289 that you made this show and this is the title 00:03:37.289 --> 00:03:40.669 of this show and then they would they would go 00:03:40.669 --> 00:03:43.849 and they would have to type in to candidly what 00:03:43.849 --> 00:03:47.349 are these like rudimentary search directories 00:03:47.349 --> 00:03:50.110 that a lot of these podcast platforms have and 00:03:50.110 --> 00:03:52.250 you would hope and pray that they typed in exactly 00:03:52.250 --> 00:03:54.830 what you titled this show so that they could 00:03:54.830 --> 00:03:59.819 try and find that um and that was part of what 00:03:59.819 --> 00:04:04.120 drove up um the content production marketing 00:04:04.120 --> 00:04:07.680 and talent costs of making content under this 00:04:07.680 --> 00:04:12.280 traditional model um which has resulted in i 00:04:12.280 --> 00:04:16.680 mean really a widespread i don't mean to be dramatic 00:04:16.680 --> 00:04:20.180 but extinct extinction event across our industry 00:04:20.180 --> 00:04:23.490 where we're seeing you know, Stitcher and Gimlet 00:04:23.490 --> 00:04:27.350 and Wondery Layoffs and, you know, Odyssey. And 00:04:27.350 --> 00:04:31.490 I mean, these companies that are just not finding 00:04:31.490 --> 00:04:34.089 sustainable business models to continue to produce 00:04:34.089 --> 00:04:40.949 original content or even, you know, partner content 00:04:40.949 --> 00:04:45.660 profitably. Right. So, yes, I mean, if I'm going 00:04:45.660 --> 00:04:47.839 to make I know I've said this before, if I'm 00:04:47.839 --> 00:04:50.120 going to make a podcast about whales, we title 00:04:50.120 --> 00:04:52.620 it whales, because that way, when people are 00:04:52.620 --> 00:04:55.240 looking for content on whales, they can find 00:04:55.240 --> 00:04:58.819 our content. Yeah, it's the Max Cutler way of 00:04:58.819 --> 00:05:01.860 titling a show, isn't it? He had a had a serial 00:05:01.860 --> 00:05:04.620 killers show, which was called Serial Killers. 00:05:04.620 --> 00:05:08.000 And yeah, absolutely. Exactly. Yeah. And then, 00:05:08.000 --> 00:05:11.860 you know, and then it also means because. Our 00:05:11.860 --> 00:05:14.899 time to production is dramatically less and our 00:05:14.899 --> 00:05:17.379 costs of production are less. It means that we 00:05:17.379 --> 00:05:20.600 can surf trends much better than traditional 00:05:20.600 --> 00:05:24.160 podcasting organizations. So a week ago when 00:05:24.160 --> 00:05:29.220 Charlie Kirk was shot, we had content about Charlie 00:05:29.220 --> 00:05:31.459 Kirk up. We had a living biography. We had a 00:05:31.459 --> 00:05:35.660 content and a new show about his assassination 00:05:35.660 --> 00:05:39.569 up within an hour. So when people typed Charlie 00:05:39.569 --> 00:05:43.329 Kirk into Apple and Spotify, we were three of 00:05:43.329 --> 00:05:45.930 the top five shows, even beating out the Daily 00:05:45.930 --> 00:05:48.329 Mail. The only one that was beating us was the 00:05:48.329 --> 00:05:52.550 Charlie Kirk show. So that means that we're able 00:05:52.550 --> 00:05:56.269 to meet a need for content that people want, 00:05:56.430 --> 00:05:59.750 that they're looking for in that moment. And 00:05:59.750 --> 00:06:04.589 that has not really ever been part of the commercial 00:06:04.589 --> 00:06:07.699 podcast business model until now. When his assassin 00:06:07.699 --> 00:06:10.360 was arrested, we had a new episode up within 00:06:10.360 --> 00:06:13.459 15 minutes. Did a human being check that first? 00:06:13.699 --> 00:06:17.720 When we are producing political or news -related 00:06:17.720 --> 00:06:22.459 content, especially when it's new shows that 00:06:22.459 --> 00:06:26.300 we are launching, we listen to the content before 00:06:26.300 --> 00:06:30.220 it launches. And part of the reason why we haven't 00:06:30.220 --> 00:06:33.860 leaned into news and politics quite yet is because... 00:06:34.360 --> 00:06:38.620 I think there still needs to be human review 00:06:38.620 --> 00:06:43.279 and not even going to say editorializing, but 00:06:43.279 --> 00:06:47.079 still AI is not so great at always capturing 00:06:47.079 --> 00:06:53.120 the tone of the moment. Yeah. And when you are 00:06:53.120 --> 00:06:57.079 crafting personalities that you want to be complex, 00:06:57.279 --> 00:07:00.920 you have... into these personalities, things 00:07:00.920 --> 00:07:05.699 like senses of humor and flaws and weaknesses 00:07:05.699 --> 00:07:08.879 and things. And so sometimes they don't always 00:07:08.879 --> 00:07:12.519 treat sensitive topics with enough sensitivity. 00:07:12.879 --> 00:07:16.879 They don't strike exactly the right chord. And 00:07:16.879 --> 00:07:21.620 so as we think about news and politics, we are 00:07:21.620 --> 00:07:25.699 figuring out how do we build the right... Infrastructure 00:07:25.699 --> 00:07:27.920 that still leverages AI and still is consistent 00:07:27.920 --> 00:07:32.439 with our model, but mitigates the risk of getting 00:07:32.439 --> 00:07:35.740 it wrong. And that probably means that we need 00:07:35.740 --> 00:07:45.540 to have a well -known news person who is reviewing 00:07:45.540 --> 00:07:48.800 that content before it was released and is making 00:07:48.800 --> 00:07:53.160 sure that we're striking. that we're getting 00:07:53.160 --> 00:07:55.180 it right and striking the right tone. Yeah. So 00:07:55.180 --> 00:07:58.480 human review there is important, but human review 00:07:58.480 --> 00:08:01.980 for a podcast about whether it's knitting or 00:08:01.980 --> 00:08:04.759 about gardening isn't necessarily as important. 00:08:05.060 --> 00:08:08.160 Yeah. I mean, there's all sorts of levels of 00:08:08.160 --> 00:08:11.180 risk, right, associated with the different content. 00:08:11.319 --> 00:08:17.240 I mean, so we think about which ones need to 00:08:17.240 --> 00:08:19.939 be. much more curated and where we need to be 00:08:19.939 --> 00:08:23.319 much more careful. And then we also really examine 00:08:23.319 --> 00:08:25.879 and stay on top of what are the capabilities 00:08:25.879 --> 00:08:32.620 of the tools, right? AI still struggles with 00:08:32.620 --> 00:08:35.779 numbers, right? I would not rely on AI to do 00:08:35.779 --> 00:08:39.440 your math homework, right? So when we're thinking 00:08:39.440 --> 00:08:43.519 about current, like what are the topics that 00:08:43.519 --> 00:08:45.379 we want to cover, we lean into the things that 00:08:45.379 --> 00:08:49.549 we know that AI is really good at. for which 00:08:49.549 --> 00:08:54.750 we are less concerned about any potential downside 00:08:54.750 --> 00:08:59.190 risk if it's giving health and wellness advice. 00:08:59.350 --> 00:09:01.990 We're going to want to make sure that we're doing 00:09:01.990 --> 00:09:06.919 more to... pull from multiple different sources 00:09:06.919 --> 00:09:10.580 and mitigate against hallucinations and that 00:09:10.580 --> 00:09:13.879 somebody is involved in listening and reviewing 00:09:13.879 --> 00:09:16.220 the content before it's released. But yeah, I 00:09:16.220 --> 00:09:18.440 mean, I'm releasing 3 ,000 episodes a week and 00:09:18.440 --> 00:09:20.419 I have eight people on my team. There's no way 00:09:20.419 --> 00:09:21.960 we're listening to all the content. There's no 00:09:21.960 --> 00:09:24.200 way we're listening to the overwhelming majority 00:09:24.200 --> 00:09:26.700 of our content before it's released. There's 00:09:26.700 --> 00:09:29.539 one podcast directory out there removing AI -generated 00:09:29.539 --> 00:09:32.220 podcasts from its directory. What's your thought 00:09:32.220 --> 00:09:35.990 on that? I think that it's an antiquated thinking. 00:09:36.190 --> 00:09:39.730 I think that in a very short amount of time, 00:09:39.789 --> 00:09:42.230 if I can make yet another bold prediction, that 00:09:42.230 --> 00:09:45.769 will probably cause stir. You're not going to 00:09:45.769 --> 00:09:49.230 upset more people, are you? That seems to be 00:09:49.230 --> 00:09:53.669 my thing lately. I think that in a very short 00:09:53.669 --> 00:09:56.549 amount of time, and I would predict certainly 00:09:56.549 --> 00:10:00.129 less than 24 months and probably less than 12. 00:10:01.789 --> 00:10:05.730 AI will become the default tool for creating 00:10:05.730 --> 00:10:11.769 content. And this conversation around is it AI 00:10:11.769 --> 00:10:15.429 or not, transparency or not, will go away the 00:10:15.429 --> 00:10:18.870 same way that this happened with, you know, digital 00:10:18.870 --> 00:10:21.029 photography and photo editing, right? I mean, 00:10:21.049 --> 00:10:23.690 now nobody makes a disclosure like this picture 00:10:23.690 --> 00:10:26.629 was taken with a digital camera and was edited 00:10:26.629 --> 00:10:29.919 after the fact, right? Everybody knows. People 00:10:29.919 --> 00:10:32.519 use digital photography and photo editing. And 00:10:32.519 --> 00:10:35.919 in fact, nowadays you make a disclosure when 00:10:35.919 --> 00:10:40.379 something is not captured. You would say this 00:10:40.379 --> 00:10:43.580 is made on traditional 35 millimeter and is not 00:10:43.580 --> 00:10:45.899 edited, right? Because that's been the thing 00:10:45.899 --> 00:10:48.100 that's special about it. Already there was a 00:10:48.100 --> 00:10:52.639 survey on YouTube where 92 % of creators said 00:10:52.639 --> 00:10:56.419 that they're using AI in their creative process. 00:10:57.039 --> 00:10:58.940 And I think there's a difference between using 00:10:58.940 --> 00:11:02.500 AI as part of your creative process and a difference 00:11:02.500 --> 00:11:06.039 between using AI as the only part of your creative 00:11:06.039 --> 00:11:09.340 process. We're not using AI as the only part 00:11:09.340 --> 00:11:11.379 of our creative process. And I think that is 00:11:11.379 --> 00:11:13.460 a very, very nuanced conversation that people 00:11:13.460 --> 00:11:16.100 don't realize. So right now, the standard is, 00:11:16.120 --> 00:11:19.019 you know, it's been relayed to us and we're kind 00:11:19.019 --> 00:11:21.000 of following as the community guidelines. If 00:11:21.000 --> 00:11:24.940 you use AI. It is a material portion of your 00:11:24.940 --> 00:11:27.779 content you need to include a disclosure. So 00:11:27.779 --> 00:11:30.360 we include, we're AI transparent. We include 00:11:30.360 --> 00:11:32.259 disclosure in all of our content. I know you 00:11:32.259 --> 00:11:35.279 reference that. You didn't do that at the beginning 00:11:35.279 --> 00:11:39.779 and now you do. Why did you change your view 00:11:39.779 --> 00:11:48.320 there? We were limited because we have so many 00:11:48.320 --> 00:11:52.120 episodes. We were limited by the API call capacity. 00:11:53.379 --> 00:11:57.360 to be able to dynamically update all of our content. 00:11:58.580 --> 00:12:03.799 So we were not able to go back quickly to do 00:12:03.799 --> 00:12:07.679 it with all of our back catalog. But now we were 00:12:07.679 --> 00:12:09.779 able to build out the process in order to do 00:12:09.779 --> 00:12:11.700 that. So now if you go through and you see all 00:12:11.700 --> 00:12:14.460 of our show notes, it's now in all of our content. 00:12:14.740 --> 00:12:17.580 Yeah, but you didn't when you started, I guess 00:12:17.580 --> 00:12:20.759 is the question. So why did you change your mind 00:12:20.759 --> 00:12:24.000 and start? disclosing that it was ai generated 00:12:24.000 --> 00:12:27.139 you mean like it's two years ago when people 00:12:27.139 --> 00:12:29.360 first started making content well i mean to be 00:12:29.360 --> 00:12:31.559 clear not all of the content that you hear from 00:12:31.559 --> 00:12:35.259 two years ago is ai generated many of it is in 00:12:35.259 --> 00:12:40.080 the voices of uh people who are on our team um 00:12:40.080 --> 00:12:43.059 some of it we went back and then replaced with 00:12:43.059 --> 00:12:45.860 synthetic voices and personalities that we designed 00:12:45.860 --> 00:12:49.740 um you know i mean at the time i don't even think 00:12:49.740 --> 00:12:52.490 people were thinking about how would you make 00:12:52.490 --> 00:12:58.330 an AI disclosure or not? So, I mean, we've adopted 00:12:58.330 --> 00:13:02.269 and evolved as the industry evolved. And now 00:13:02.269 --> 00:13:05.009 I think we're really proud of the way that we're 00:13:05.009 --> 00:13:09.730 leading the way in transparency. All of our personalities, 00:13:09.789 --> 00:13:13.090 we have some clever prompt engineering around 00:13:13.090 --> 00:13:16.929 how we have them do their AI disclosure so that 00:13:16.929 --> 00:13:22.139 they say that they're AI, but they... do it in 00:13:22.139 --> 00:13:25.000 a way that sells the benefit of having a host 00:13:25.000 --> 00:13:28.639 that is an AI personality. And from our data, 00:13:28.700 --> 00:13:32.940 we find very little, sometimes immeasurable drop 00:13:32.940 --> 00:13:36.259 -off after our AI disclosure. So our read is 00:13:36.259 --> 00:13:39.100 that... If people like the personality and they 00:13:39.100 --> 00:13:41.360 like the content, they don't care that it's AI. 00:13:41.600 --> 00:13:44.779 In the same way as there's an explicit tag in 00:13:44.779 --> 00:13:49.399 podcasting where you can mark if you use saucy 00:13:49.399 --> 00:13:53.100 language or adult discussions, some people have 00:13:53.100 --> 00:13:57.399 been calling for an AI tag to help audiences 00:13:57.399 --> 00:14:00.559 or advertisers programmatically filter out majority 00:14:00.559 --> 00:14:06.539 AI content. What's your view on that? I think 00:14:06.539 --> 00:14:09.980 that we're in a moment. Well, I mean, first of 00:14:09.980 --> 00:14:13.059 all, I think that some people are trying to capitalize 00:14:13.059 --> 00:14:16.059 on the fear around this conversation to make 00:14:16.059 --> 00:14:20.059 news for themselves. The conversation is, if 00:14:20.059 --> 00:14:22.100 you have people who are really thoughtfully engaging 00:14:22.100 --> 00:14:25.080 in this conversation, I think the conversation 00:14:25.080 --> 00:14:28.759 is much more complicated, right? Because if you're 00:14:28.759 --> 00:14:33.480 using, you know, ChatGPT to make a script, but 00:14:33.480 --> 00:14:36.480 you're... actually reading that into the mic, 00:14:36.580 --> 00:14:40.559 is that materially AI generated? If you're, you 00:14:40.559 --> 00:14:43.100 know, actually write the script, but you use 00:14:43.100 --> 00:14:45.840 a synthetic voice, is that materially AI generated? 00:14:45.899 --> 00:14:48.059 If you're doing an interview and it's a real 00:14:48.059 --> 00:14:50.360 person that's interviewing a digital person, 00:14:50.440 --> 00:14:54.440 is that AI generated, materially AI generated? 00:14:55.279 --> 00:14:59.070 It's getting to the point now where... It's so 00:14:59.070 --> 00:15:01.929 nuanced. I mean, I've actually had people reach 00:15:01.929 --> 00:15:03.629 out to me that some of these people who are in 00:15:03.629 --> 00:15:06.610 this space making commentary that I can't believe 00:15:06.610 --> 00:15:08.490 they're doing this or they need to have disclosure 00:15:08.490 --> 00:15:11.730 or, you know, more disclosure. Advertisers aren't 00:15:11.730 --> 00:15:14.169 going to want this. Right. I have people behind 00:15:14.169 --> 00:15:16.710 the scenes emailing me saying, I work at this 00:15:16.710 --> 00:15:19.250 company and we're using AI in our process. Right. 00:15:19.570 --> 00:15:23.190 And basically every podcast that I'm doing, some 00:15:23.190 --> 00:15:25.710 people, you know, going for the jugular, asking 00:15:25.710 --> 00:15:28.320 me the tough questions about. controversy in 00:15:28.320 --> 00:15:30.600 this space and then we get off the call and then 00:15:30.600 --> 00:15:32.899 they say how do you think that i could better 00:15:32.899 --> 00:15:35.519 use ai in my process to make my process more 00:15:35.519 --> 00:15:39.639 efficient or could you all recreate my voice 00:15:39.639 --> 00:15:41.340 and what would that sound like and maybe i could 00:15:41.340 --> 00:15:44.379 actually take a vacation sometime right there's 00:15:44.379 --> 00:15:48.799 the creators uh are going to be this is the most 00:15:48.799 --> 00:15:51.200 powerful creative tool that the world has ever 00:15:51.200 --> 00:15:54.480 seen and creators are going to be leaning in 00:15:55.120 --> 00:15:59.379 to this technology very quickly. And this conversation 00:15:59.379 --> 00:16:03.320 is going to go by the wayside. But in the meantime, 00:16:03.500 --> 00:16:09.980 as people are interested in, we will follow the 00:16:09.980 --> 00:16:14.659 guidelines as to as much disclosure as you want 00:16:14.659 --> 00:16:18.899 us to have. The reality is we've not heard any 00:16:18.899 --> 00:16:23.500 pushback from any advertiser ever. And we... 00:16:23.879 --> 00:16:27.460 are not finding that we're missing out on audiences.