WEBVTT 00:00:00.000 --> 00:00:06.240 You're recording? Oh, yeah. I would ask that and I was like, yeah, I have to double check. 00:00:06.240 --> 00:00:09.800 My glasses. It's OK. 00:00:09.800 --> 00:00:13.840 We'll make it work. All right. I am here with? Jonathan Horst. 00:00:13.840 --> 00:00:20.040 And what do you do for the company? I write, host, produce, American Dress. 00:00:20.040 --> 00:00:23.080 The Apple Focus channel. We love that channel. 00:00:23.080 --> 00:00:29.520 Oh, I think. I like it, actually. I feel like I watched the, I watched, I think, Hoffman edited it. 00:00:29.560 --> 00:00:32.760 They're the most recent one. Yeah. That was really good. I liked it. I watched the whole thing. 00:00:32.760 --> 00:00:36.800 I was like, interesting. You'll be so lucky. 00:00:36.800 --> 00:00:40.320 No. I'm not. I'm not that kind of editor. 00:00:40.320 --> 00:00:44.280 You sure? Yeah. It's not my strength. 00:00:44.280 --> 00:00:47.880 All right. I mean, I was like, maybe it is. We'll find out. 00:00:47.880 --> 00:00:53.840 What's that? All right. So I'm just asking everyone in the company what their thoughts on AI it is. 00:00:53.840 --> 00:00:54.800 Oh, my God. 00:00:57.880 --> 00:01:01.240 Over-hyped? Over-hyped? Maybe incorrectly named. 00:01:05.840 --> 00:01:09.560 Annoying? I don't know. 00:01:09.560 --> 00:01:13.200 What is there to know? I don't even feel like I know enough about it, 00:01:13.200 --> 00:01:16.920 but I'm annoyed by the hype around it. 00:01:16.920 --> 00:01:21.560 Let's wait till the door closes. It don't work. 00:01:21.560 --> 00:01:24.720 Yeah, I'm not a lot of the bad idea, but we're committed. 00:01:24.720 --> 00:01:28.480 We committed to this location, and it's our fault. 00:01:29.480 --> 00:01:33.400 Yeah, just. Yeah, go ahead. It's OK. The bid is committed. 00:01:33.400 --> 00:01:37.360 It's committed. It's footprint. Yeah. 00:01:37.360 --> 00:01:41.720 Yeah. OK, so it sounds like you don't like it at all. 00:01:41.720 --> 00:01:47.120 I mean, OK, so here's the problem. Especially this past year, with the release of chat 00:01:47.120 --> 00:01:53.800 GPT and the mid-journey bot and all this weird stuff, everyone started attributing all this amazing capability 00:01:53.800 --> 00:01:57.840 to something that really hasn't been proven. 00:01:57.840 --> 00:02:03.200 And so I think we got a little fooled by what it was doing and thought it could do more. 00:02:03.200 --> 00:02:07.200 And think it can do more than it can. And we're thinking of it as like, 00:02:07.200 --> 00:02:13.200 it's going to just do the thing we want it to do when it's probably just going to be yet another tool in the toolbox 00:02:13.200 --> 00:02:16.520 to help. You still have to do the thing, though. You know what I mean? 00:02:16.520 --> 00:02:20.560 So when people are like, writers are going to be out of jobs 00:02:20.560 --> 00:02:23.560 because you could just use chat GPT, it's like. 00:02:23.560 --> 00:02:26.520 And I'm not sure about that, because it's just 00:02:26.560 --> 00:02:30.440 a summary of all the writing it's ever looked at. 00:02:30.440 --> 00:02:35.280 So how are you going to get any new ideas or new ways of thinking about things or new thinking? 00:02:35.280 --> 00:02:39.560 Because that's what writing is. From that, I don't know. 00:02:39.560 --> 00:02:44.160 OK, well, surprisingly, you're the first person I interviewed today who is really against it. 00:02:44.160 --> 00:02:49.160 So against it. But what's one thing you like about AI, then? 00:02:52.600 --> 00:02:59.160 I don't know. I haven't really, ah, AI is such a broad term. 00:02:59.160 --> 00:03:04.240 I guess AI in the current state. Or in the current culture. 00:03:04.240 --> 00:03:04.740 Exactly. 00:03:16.880 --> 00:03:21.160 It's impressive, but I'm not impressed, if that makes any sense. 00:03:21.160 --> 00:03:24.600 It's like, I don't know, I dabbled with mid-journey. 00:03:24.600 --> 00:03:28.800 And I was like, if you have broad strokes, 00:03:28.800 --> 00:03:34.560 you'll get an interesting image. But then if you want anything specific from the real world 00:03:34.560 --> 00:03:37.560 or something like that, it just starts making stuff up. 00:03:37.560 --> 00:03:44.720 And you're like, ah, it's a bit of a fever dream still. And it's like, I don't know. 00:03:44.720 --> 00:03:50.000 It can only get you so far, it feels like. And then hearing about people writing emails with chat GPT 00:03:50.000 --> 00:03:55.600 or something like that, it becomes this weird question of, well, what's the point? 00:03:55.600 --> 00:04:00.480 If you're not willing to put effort into saying 00:04:00.480 --> 00:04:04.840 what you want to convey, then are you saying anything? 00:04:04.840 --> 00:04:11.240 I don't know. So it just makes everything so much harder to trust. 00:04:11.240 --> 00:04:18.440 And so I find these concerns way bigger than the benefits. 00:04:18.440 --> 00:04:22.080 And so I'm just like, I just haven't 00:04:22.080 --> 00:04:25.920 bothered to really play around too much with it. 00:04:25.920 --> 00:04:31.040 So I'm not the most experienced, I could say. But when I have dabbled with it, I've been like, oh, 00:04:31.040 --> 00:04:34.280 this is kind of neat. But and maybe this could be a tool, 00:04:34.280 --> 00:04:40.680 like I guess Adobe's generative fill is kind of interesting. 00:04:40.680 --> 00:04:46.600 But it's like, they're demonstrating it. And look, we could make this image so much wider. 00:04:46.720 --> 00:04:52.680 We can fill it just from this tiny little portrait. We could turn it into a whole landscape thing. 00:04:52.680 --> 00:04:56.720 And it's like, yeah, I guess that's cool. But it's also like, what is the picture then anymore? 00:04:56.720 --> 00:05:01.160 Whereas it's probably a really great tool when you're trying to do a cool effect 00:05:01.160 --> 00:05:04.280 or you take the subject foreground or the background. 00:05:04.280 --> 00:05:08.360 But when you do that, the background has that thing cut out of it. 00:05:08.360 --> 00:05:12.360 So you could fill it with the generative fill, that little space. And that would be really great. 00:05:12.360 --> 00:05:17.440 You know what I mean? It's great for small things. It got marketed to us as this big thing. 00:05:17.440 --> 00:05:23.200 And so I think then it becomes really hard to trust it or understand it within those limitations 00:05:23.200 --> 00:05:30.080 and then be able to take advantage of it there. Instead, we're being hyped on this new future that actually 00:05:30.080 --> 00:05:36.760 sounds worse. You know what I mean? Yeah, no. I always talk about AI as like, oh, I know, we've always had it. 00:05:36.760 --> 00:05:40.880 It's like in a small scale, like you said. But it's like, now that it's this big scale thing, 00:05:40.880 --> 00:05:45.920 like, oh, it can take over jobs. So I thought that's like, oh, OK, that's not really. 00:05:45.920 --> 00:05:50.320 What's being said is that. I don't buy that. Yeah, near do I, yeah. 00:05:50.320 --> 00:05:54.520 Yeah, yeah. So yeah, so it's annoying. 00:05:54.520 --> 00:05:58.200 It's really annoying. I feel like, and then I get an argument with people about it. 00:05:58.200 --> 00:06:02.880 I don't know what to say. Have you written something? It's going to be really satisfying when you figure out 00:06:02.880 --> 00:06:05.920 how to write. I hate writing, by the way. 00:06:05.920 --> 00:06:10.600 But when you figure out how to write something really cohesively, that's really impressive. 00:06:10.600 --> 00:06:14.800 And I just don't know how a prompt is going to lead to that. 00:06:14.800 --> 00:06:16.680 As a writer, you hate writing? I do. 00:06:20.920 --> 00:06:26.080 All right, let me look at my next few questions here. OK, so you kind of touched on it. 00:06:26.080 --> 00:06:29.200 But you said that like, there's always 00:06:29.200 --> 00:06:34.600 like the idea of that AI is going to take over jobs. So you don't believe in that, right? 00:06:34.600 --> 00:06:40.320 It's like, I mean, you could say any sort of technology 00:06:40.320 --> 00:06:44.280 takes over jobs. But in some ways, the jobs shift, perhaps. 00:06:44.280 --> 00:06:49.760 And so like, I don't know, like you would have trying to think of an example. 00:06:49.760 --> 00:06:54.320 Like you could have a really specific job for like a really menial task. 00:06:54.320 --> 00:06:57.400 And like, and then they get a robot to do that. 00:06:57.400 --> 00:07:01.440 I think like, so I'm interested in cars. So in the car industry, like they tried, 00:07:01.440 --> 00:07:07.120 there was an attempt to roboticize the whole production 00:07:07.120 --> 00:07:11.600 process in the 80s. General Motors tried it, and it was a massive failure. 00:07:11.600 --> 00:07:15.640 And like, I think there are certain tasks that like are really great where somebody 00:07:15.640 --> 00:07:19.480 could get out of a job. So like, I don't know, like welding, you know, 00:07:19.480 --> 00:07:23.480 stamping and welding, the like metal panels of a car. 00:07:23.480 --> 00:07:29.360 You know, instead of having someone like manually doing that, you have the system because every piece is exactly the same. 00:07:29.360 --> 00:07:32.720 Smush and bush and do the thing. And it's like, OK, so the person who 00:07:32.720 --> 00:07:37.400 had to do that manual menial task, I guess it did replace their job. 00:07:37.440 --> 00:07:41.880 But like, they can do something else. I think when we get up to this weird, like, 00:07:41.880 --> 00:07:45.440 broad creative space, it's puzzling to me 00:07:45.440 --> 00:07:49.240 that you would say, oh, this tool is going 00:07:49.240 --> 00:07:55.680 to take over a creative job instead of like making the creative job easier or allowing the creative job 00:07:55.680 --> 00:07:59.760 to be the boundaries to be pushed or something like that. Those are two like very different things. 00:07:59.760 --> 00:08:03.080 And I just think it's more the last one. So I just want to go. Do you need ice? You can get ice if you want. 00:08:03.080 --> 00:08:07.120 That's good, that's good. OK, so this is a bad idea. 00:08:07.120 --> 00:08:11.200 I don't have any ice in my water, so it's a possibility. 00:08:11.200 --> 00:08:16.480 Yeah, so it's probably more like, oh, maybe it'll push the boundaries of what the job looks like 00:08:16.480 --> 00:08:19.600 or maybe how many people are involved in it. 00:08:19.600 --> 00:08:25.520 Maybe, perhaps. Yeah, OK, I guess that's true. But it's not like a one-to-one replacement. 00:08:25.520 --> 00:08:29.200 But I guess what people do shifts all the time anyway, right? 00:08:29.200 --> 00:08:32.360 Exactly, exactly. All right, so and you kind of dabble on it, 00:08:32.360 --> 00:08:37.360 but you've used AI. How do you feel about it? 00:08:37.360 --> 00:08:41.480 You said you were OK about it, weird about it, or? 00:08:41.480 --> 00:08:45.680 Oh, I was like, oh, this is pretty neat. And I tried some image prompts. 00:08:45.680 --> 00:08:50.960 And I was like, oh, and weirdly, I thought the broader it was, the better it was, 00:08:50.960 --> 00:08:58.640 which I thought was interesting. And you can sort of see the years of going through image 00:08:58.640 --> 00:09:03.880 search or scrolling through Pinterest or whatever. You can sort of see how it all averages out 00:09:03.880 --> 00:09:10.360 and how it starts to be its common thing. So I was like, oh, OK, this is like a weird kind of take 00:09:10.360 --> 00:09:14.160 on all these pictures that I could picture having seen 00:09:14.160 --> 00:09:19.120 or whatever, right? So I don't know. So in that case, it's like, that's neat. 00:09:19.120 --> 00:09:22.280 But also, it has no sense of the physical world. 00:09:22.280 --> 00:09:26.760 It has no understanding of anything. 00:09:26.760 --> 00:09:30.160 It's literally just like, I know that this pixel is most 00:09:30.160 --> 00:09:33.320 likely to be next to this pixel when these pixels clump 00:09:33.760 --> 00:09:38.720 together like that. And I mean, I'm probably reducing it. I'm sure some AI people are going to be so mad, and I don't care. 00:09:38.720 --> 00:09:42.840 But it doesn't know. 00:09:42.840 --> 00:09:46.280 Have you seen the AI-generated commercials? 00:09:46.280 --> 00:09:50.760 Yeah, I've seen a lot. They're like, whoa. Some of them are really like, some of them are like, cool, 00:09:50.760 --> 00:09:56.400 but you can tell that it's very AI. Well, I mean, when you get to motion, 00:09:56.400 --> 00:10:00.200 then even the animation, you're replicating 00:10:00.200 --> 00:10:04.760 the physical reality, right? So animators in Disney in the past 00:10:04.760 --> 00:10:08.040 would study how a bird flies. And then they would draw that out. 00:10:08.040 --> 00:10:11.120 And it was an acknowledgement of how something actually 00:10:11.120 --> 00:10:15.080 physically happens. And then you watch this. There's this really funny beer commercial. 00:10:15.080 --> 00:10:20.760 And no one's drinking the beer. They're holding the beer bottle, and it's near their mouth. 00:10:20.760 --> 00:10:24.120 And then the person's sucking on. 00:10:24.120 --> 00:10:27.120 But they're making a sucking face. 00:10:27.120 --> 00:10:30.120 And it's like, there's no liquid. There's nothing. 00:10:30.560 --> 00:10:34.080 It's such an indicator of, oh, these things just 00:10:34.080 --> 00:10:39.040 don't get the physicals. And so what are you going to do then? 00:10:39.040 --> 00:10:44.240 Are you going to now create an AI that gets the laws of physics calculated and now has 00:10:44.240 --> 00:10:48.560 to figure out how to render that? It's just going to be more and more complicated. 00:10:48.560 --> 00:10:52.600 OK. So then as a tech-slash-creator company that we are, 00:10:52.600 --> 00:10:55.640 do you think that we should be using AI in any capacity? 00:10:55.640 --> 00:10:59.160 Or? 00:10:59.160 --> 00:11:03.120 I mean, again, from what little I know of what's out there, 00:11:03.120 --> 00:11:07.880 I'm sure there might be some really basic things where 00:11:07.880 --> 00:11:12.360 it could be beneficial. Like again, like I said, the gender to fill, 00:11:12.360 --> 00:11:15.920 if you want to animate a photo in our space. 00:11:15.920 --> 00:11:19.600 Sorry, can we start again? So like, start again, yeah. 00:11:19.600 --> 00:11:23.080 I'll start from the top. Yeah, yeah. Oh, yeah, OK. This is a bad picture. 00:11:23.080 --> 00:11:32.560 I'm so sorry. Don't worry about it. So in the case of here, using the tools really broadly, 00:11:32.560 --> 00:11:38.160 saying chat GPT will replace a writer or that is a bit rough. 00:11:38.160 --> 00:11:42.280 Maybe getting a good background image for a thumbnail 00:11:42.280 --> 00:11:46.640 with mid-journey might work. I could see the gender to fill being really great 00:11:46.640 --> 00:11:49.880 if we wanted to animate a photo and give it 00:11:49.880 --> 00:11:56.760 this parallax effect. I don't know if I should have stopped because it's just sounds. 00:11:56.760 --> 00:11:59.920 They're just backgrounds. But yeah, you get this parallax effect. 00:11:59.920 --> 00:12:03.160 If you get the parallax effect, then you could have this really cool animation. 00:12:03.160 --> 00:12:09.040 You fill it with the gender to fill. That sounds great. So again, really specific use cases 00:12:09.040 --> 00:12:14.280 where it's going to be really amazing. But in terms of broad strokes, no. 00:12:14.280 --> 00:12:20.600 It's going to help just like that tiny little step, like making one vision you have a bit easier 00:12:21.160 --> 00:12:25.320 to execute or making it better. That sounds great. 00:12:25.320 --> 00:12:29.200 So basically, we use it as a starting point, but then it won't be some new use for it. 00:12:29.200 --> 00:12:33.480 We should use it as a starting point, but we shouldn't use it for the final product. 00:12:33.480 --> 00:12:39.920 As a tool. We should be using it as a tool. Like you'd use a computer as a tool or a camera as a tool. 00:12:39.920 --> 00:12:46.800 That's what it's supposed to be, and that's what it is. And thinking of it as anything but that is annoying. 00:12:46.800 --> 00:12:50.320 Yeah. I think what everyone talked to, it's kind of like it is a tool, 00:12:50.320 --> 00:12:54.080 but it shouldn't be something we really rely on, basically. 00:12:54.080 --> 00:12:58.000 Yeah, no. No. 00:12:58.000 --> 00:13:01.720 What's his name? Dax Flame has been trying to do. 00:13:01.720 --> 00:13:08.880 Have you ever heard of him? He's a YouTuber who's been using AI to guide his videos along. 00:13:08.880 --> 00:13:13.000 And sometimes I'm like, man, you have great ideas. 00:13:13.000 --> 00:13:16.000 You just need to be confident in yourself. 00:13:16.000 --> 00:13:19.000 Just do that. Yeah. Yeah. 00:13:19.000 --> 00:13:22.840 All right, we're going to wrap up here quickly. But is there anything that we haven't mentioned 00:13:22.840 --> 00:13:28.720 that you want to discuss? No, I'm so glad to get this off my chest. 00:13:28.720 --> 00:13:32.760 All right, well, thank you. Or saying goodbye to the float planers now. 00:13:32.760 --> 00:13:35.760 Goodbye float planers. Yeah, that's the name I've coined now. 00:13:35.760 --> 00:13:38.760 I don't know if it's official. Is it one video or just each separate? 00:13:38.760 --> 00:13:42.960 It's going to be like 20 uncuts. And then I got cut down to like a five minute video. 00:13:42.960 --> 00:13:45.000 Oh yeah, OK.