WEBVTT

1
00:00:00.100 --> 00:00:05.900
[upbeat music] Welcome to the Rebooting Show. I'm Brian Morrissey.

2
00:00:06.120 --> 00:00:15.160
This week I have a spotlight episode with Josh Brandau, the CEO of the Rebooting's partner, Noda, which is a company that offers AI tools to improve workflow in newsrooms.

3
00:00:15.640 --> 00:00:22.880
We're very excited to partner with Noda on an upcoming online forum where we'll discuss how AI can be a force for good in a more with less era.

4
00:00:23.360 --> 00:00:31.240
You know, the other week I wrote a piece on the Rebooting about the AI adoption curve. I compared it to the Kubler-Ross stages of grieving, because I love to do that.

5
00:00:31.760 --> 00:00:40.180
There's, of course, the anger and, and bargaining phases, which we're still in with getting to the deals that many publishers are striking. And Josh and I discussed these.

6
00:00:40.660 --> 00:00:52.560
But I think a key phase of this adoption curve is the efficiency phase. I mean, companies of all kinds are adopting AI initially in the hopes of becoming more efficient and therefore more productive.

7
00:00:52.860 --> 00:00:58.420
This is critical, right? So publishing is no different, and if anything, it's under unique pressures.

8
00:00:58.620 --> 00:01:09.539
And, you know, the Rebooting's recent survey of news publishers found that efficiency was by far the biggest opportunity that they saw in AI. Josh is no stranger to this. He's a publishing veteran.

9
00:01:09.620 --> 00:01:12.620
He was the CRO and CMO at The Los Angeles Times.

10
00:01:13.020 --> 00:01:25.080
That greatly informed his decision to start Noda, since like other publishers, you know, he had saw at the times, you know, legacy media that was struggling to adopt technologies that inevitably underpin sustainable businesses.

11
00:01:25.360 --> 00:01:42.380
We discussed those inefficiencies that are inherent in a lot of newsrooms and end up taking away scarce resources from the actual news reporting, and also how tasks like versioning, content optimization, SEO and tagging, which can sound all boring, but that they can be sped along with an assist from AI.

12
00:01:42.900 --> 00:01:48.890
And this... People inevitably go to job loss, right? And I think it's real. But I also think that it's awesome.

13
00:01:49.040 --> 00:01:56.640
It's really inevitable that the future of these organizations are both slimmer and more productive, and, and the market just necessitates that.

14
00:01:57.140 --> 00:02:02.480
Josh and I also took a big picture view of where journalism goes in this kinda AI world, assuming we're going into it, right?

15
00:02:02.980 --> 00:02:10.940
And we both see that licensing is gonna be a growing revenue source, and there's a lot of, there's a lot of challenges to overcome to get there.

16
00:02:10.960 --> 00:02:21.440
But we also talk about how AI could create new revenue streams as publishers move beyond efficiency and begin to create new products that improve the customer experience. Imagine that. So I hope you enjoy this episode.

17
00:02:21.520 --> 00:02:37.620
Thanks again to Josh and Noda for their support. [upbeat music] Josh, thanks so much for, uh, joining me for the podcast. Really appreciate you having me.

18
00:02:37.660 --> 00:02:46.760
I'm-- I've been a huge fan for quite some time, so- All right... it's a little surreal to talk to you- All right. Please, please... in this professional capacity. This'll-- This is gonna get you nowhere, Josh. All right.

19
00:02:46.900 --> 00:02:59.480
No, thank you. I appreciate it. So tell me about your experience, you know, at a quote unquote "legacy publisher." What-- How did it inform, you know, going off to build, build Noda? Yeah.

20
00:02:59.540 --> 00:03:08.500
I joined the Los Angeles Times as the CRO and, and the CMO at a really interesting time in its life cycle. So Patrick Soon-Shiong had bought it from Tribune.

21
00:03:08.800 --> 00:03:13.440
He'd brought in a bunch of folks who had not been from the industry. I was a consultant.

22
00:03:13.520 --> 00:03:26.200
I was in media advertising, media planning and buying my entire career before that, and was really helping the sales team understand how to sell to advertisers, and that kind of evolved into the position I took at the LA Times.

23
00:03:26.540 --> 00:03:36.480
It, it was exciting in the reality that we were doing something that was super meaningful for the community and, and the nation at large, and that really invigorated me in what I was doing.

24
00:03:36.560 --> 00:04:00.740
But it was also really telling to see the technologies that apply to the legacy media space, as we call it, and how woefully weak they are, how there's not really an ecosystem of technology partners that supports it from a, a grander scale, and how porting over a lot of the technologies from similarly, similarly related media companies in other spaces doesn't really work very well for the use cases of media.

25
00:04:00.829 --> 00:04:01.470
So- Mm-hmm...

26
00:04:01.520 --> 00:04:16.340
it was really instructive in, in a lot of capacities, and it, it m-made me and the CTO at the time, Ben Gerst, who's my co-founder now at this company, Noda, think that, hey, if there's ever real an opportunity to support legacy media overall, we, we wanna give it a try.

27
00:04:16.680 --> 00:04:21.480
Yeah. But what specifically with-within... Where, where does it break with the tech?

28
00:04:21.660 --> 00:04:32.560
And I mean, a lot of it is, you know, particularly you gotta break it, but it's like newspapers have a, a certain problem that they're operating a legacy publishing system that is complicated. I mean, it's...

29
00:04:32.630 --> 00:04:42.260
Imagine, I mean, you're like there's delivery trucks involved and whatnot. [chuckles] Like- Mm-hmm... and that's non-trivial. There's printing plants. And that's like a particular issue.

30
00:04:42.660 --> 00:04:53.480
But i-is there anything that you saw within there and within the ecosystem that sort of said, "You know, there's, there's an opening here that, that we can, that we can help these publishers.

31
00:04:53.980 --> 00:05:02.800
They're obviously playing catch up, but to actually catch up"? That's a really good point from a CMS construction perspective. There's a lot of use cases that are very...

32
00:05:02.880 --> 00:05:08.120
Actually, I think at this point, pretty completely unique to newspaper delivery on a daily basis.

33
00:05:08.160 --> 00:05:19.730
But outside of that, which does keep some of the legacy systems truly legacy because a lot of those companies aren't really innovating to the degree that you'd want, what we saw were a few pain points that were, are very obvious.

34
00:05:19.770 --> 00:05:25.040
So one would be also the nature of the industry, breaking news, so investigative reporting.

35
00:05:25.240 --> 00:05:43.280
By definition, I had no idea that we were about to, to, to break something huge, but once we do, it's, you know, my teams, the audience teams, consumer marketing teams, content teams, everybody's job to try to get as many eyes on that as possible, and you're really playing catch up while that story is being told in the marketplace.

36
00:05:43.720 --> 00:05:53.260
So how do we get to a, how do we get to a system that allows us to create assets a lot faster than the traditional models of studios? So that was one area that we thought we could improve upon.

37
00:05:53.710 --> 00:05:55.620
Another area was newsletter creation.

38
00:05:55.720 --> 00:06:18.750
So how do we take all of this gigantic amount of sub-segmentation data across our customer base and really apply rigor to it so that we can provide you as a, a, you know, a customer with newsletters that you really wanna see?Or why don't we provide you with newsletters full of engaging content that we know you didn't see on site, for example, when you're logged in, to show the value of what we're providing you?

39
00:06:19.310 --> 00:06:28.190
And then y- honestly, something as simple as I had a channel in Slack called Headlines that was just a debate about what's a good headline for my story.

40
00:06:28.270 --> 00:06:41.270
That really has you questioning, you know, how much time needs to be spent on something that isn't necessarily for your audience, but for, you know, a, a search functionality that's necessary for an editor or an, uh, a journalist to think about today.

41
00:06:41.310 --> 00:06:51.110
So there's just like 15 to 20 pretty massive pain points that we think could be solved with, or we thought could be solved with technology and have proven over the last few years that they can be. Yeah.

42
00:06:51.190 --> 00:07:01.350
So obviously when, when we're talking about solving pain points and problems with technology these days, we're talking about AI, right? And explain the opportunity that you see.

43
00:07:01.430 --> 00:07:14.190
I, I wanna get into the threat stuff, but we'll actually try to turn it around and talk a little bit about the opportunity, because it, it's there, and I think publishers I think are starting to see it, if not have already seen it.

44
00:07:14.270 --> 00:07:23.020
I mean, we did a, a survey just recently. We just came out with a report of news publishers, and we asked how do you expect AI to impact your business?

45
00:07:23.530 --> 00:07:38.430
Now, 44% took the safe neutral, but 28% said very positively, and 22% said, you know, absolutely negatively and all, and then 6% said too soon to tell. What, what are you seeing as far as the...

46
00:07:38.450 --> 00:07:44.430
And what did you see as the opportunities to use AI in order to make these businesses work better?

47
00:07:44.950 --> 00:07:58.760
I think that that's a pretty representative understanding of where the AI technology can lead us now that everyone's not operating from a, a, you know, a one or zero or everything or nothing kind of position, where it's very clear.

48
00:07:58.830 --> 00:08:08.790
I think if we start from a, the, the happiest version of what AI can bring to the industry, which is a little unlike this podcast, no offense, right? We can just be a little- [laughs]... a little rosy- No take olds...

49
00:08:08.800 --> 00:08:18.910
about this. [laughs] That's why I started with the positive. What do you want from me? Yeah, I love that. [laughs] It's, it's, uh, it's not a novel technology for our industry. That's really the best news possible.

50
00:08:19.070 --> 00:08:32.530
So the application of a lot of this stuff really accelerates the ability for you to tell your story to a larger audience a lot faster, and a- through a lot of different mediums that you couldn't do that before, at least without spending an insane amount of time trying to make that happen.

51
00:08:33.030 --> 00:08:44.250
It's not a novel promise of, you know, what, what AI could bring to this tomorrow. It's a tool structure that exists that we've created, and a whole bunch of other folks have created, that allows you to do that today.

52
00:08:44.470 --> 00:08:46.430
So I think that's the really good news.

53
00:08:46.630 --> 00:09:01.290
If we get back to the idea of what is the purpose of this entire industry, it's, you know, expressing stories about the world around us to each other so that we can understand and really appreciate the communities we live in and, and, and the broader spectrum of what's happening with humanity.

54
00:09:01.870 --> 00:09:08.830
I think that this is amazing because it doesn't make you choose, you know, the one format that you're really good at telling that story in.

55
00:09:09.130 --> 00:09:15.309
It allows you to tell that story in any format you possibly can, and I think that's... I think that there's a lot to that.

56
00:09:15.430 --> 00:09:26.470
But I think broadly, it's the real promise of what this technology brings that this industry has been waiting for for 20 years as it's been trying to evolve past, you know, being very, very good at three things.

57
00:09:26.730 --> 00:09:34.560
So structurally being very good at writing articles- Mm-hmm... or structurally being very good at writing or creating a news segment or, or a radio segment.

58
00:09:34.730 --> 00:09:39.510
Those are really the three major areas of optimization for these gigantic structures.

59
00:09:39.800 --> 00:09:57.390
Now you can utilize that same ability to tell a story in a singular format and apply it to every other format, which serves not only the purposes of why you do anything, which is telling that story, but also applies it to way more people in the way in which they want it and expect it from every other part of their lives at this point.

60
00:09:57.750 --> 00:10:07.370
Yeah. So I think it's awesome. So bucket out the opportunities between... And these are crude, 'cause I'm sure that they'll, they'll bleed. But like one is like efficiency, right?

61
00:10:07.490 --> 00:10:26.030
I mean, I think we see this overall, and it's interesting because AI is an interesting technology paradigm 'cause I haven't-- I can't think of one that has arrived with as much consternation, and mostly because I think the way it's been packaged has been as a, a people replacer, right?

62
00:10:26.090 --> 00:10:32.450
It's been, it's been viewed as an either/or thing. Is AI gonna take your job, et cetera?

63
00:10:32.930 --> 00:10:42.970
And I think that was maybe a tough brand thing, 'cause I, I, from what I've seen, you know, there's a lot of trepidation in a lot of the consumer surveys about AI for that, that very reason.

64
00:10:42.990 --> 00:10:51.130
But there's efficiency, right? Which doesn't necessarily have to mean that jobs are going away. They're gonna change. Some will go away, I'm sure.

65
00:10:51.150 --> 00:11:03.970
But like the reality is humans, I'm always gonna bet on humans finding other, finding, finding [chuckles] finding work, finding jobs. They're just always going to. So there's that bucket, right? Then there's the- Yeah...

66
00:11:04.050 --> 00:11:13.060
using AI to develop, say, new products, right? I mean, we just saw Politico is... Or enhance, uh, existing products, but with new features.

67
00:11:13.150 --> 00:11:19.310
You know, they just hooked up with an AI firm out of, uh, Y Combinator in order to...

68
00:11:19.330 --> 00:11:28.910
Which makes total sense, I think, for their model, to give Politico Pro subscribers the ability to comb through tons of data and get, you know, just the, just the bullet points they needed.

69
00:11:28.950 --> 00:11:36.890
It's just a very Washington thing. Nobody can read anything that's not like five bullet points, I, I think, in Washington. That's what I learned in my early career there.

70
00:11:36.900 --> 00:11:41.030
[chuckles] And I think that's why Axios actually went with the smart brevity format. That's my conspiracy theory.

71
00:11:41.480 --> 00:11:54.850
And, and then there is, you know, is this gonna be like a tool for, you know, better ad revenue or, uh, of some, of some sorts? Or is it, you know, to, to create like entirely new experiences and products?

72
00:11:55.050 --> 00:11:57.790
But let, let's, let's start. Like one is the efficiency, right?

73
00:11:57.960 --> 00:12:08.350
And I think my thing is, has always been, not as always, but h- has recently been that, you know, publishing is going into a more with less era, or if it's not already there.

74
00:12:08.510 --> 00:12:17.706
And just the reality is a lot of the structuresOf these businesses need to slim down even as they do more things. And so [chuckles] you've got a problem there.

75
00:12:17.886 --> 00:12:29.326
It's just like you have to do a bunch of different things, but you also have to cut down on the size of your infrastructure. Wh- where... Is that the biggest opportunity for AI? Yeah. That's a really great way to put it.

76
00:12:29.446 --> 00:12:40.286
I think the downward pressure on the industry overall has been for quite some time, right, the expectation that we're just gonna eliminate a lot of positions that did this one thing really well. So now you take it on.

77
00:12:40.346 --> 00:12:49.346
So just report the story, edit the story, also optimize it for all the social channels we still wanna be on, even though that's changing on a daily basis. If you have time, could you make a video out of this story?

78
00:12:49.366 --> 00:12:52.246
If you have time, could you make a newsletter to tell everyone about it?

79
00:12:52.286 --> 00:13:03.406
Could you make this into, you know, could you look at all of our, to your point, content libraries and try to find other recirculation opportunities here for something related to the article you're writing on?

80
00:13:03.466 --> 00:13:11.886
All that stuff now falls onto a smaller and smaller and smaller team, but we know in order to get more engagement to make the business viable, you have to do a lot of that stuff.

81
00:13:12.006 --> 00:13:21.886
So would you either create less net new information for the world and everyone loses, or you find these tools that can help you do that without...

82
00:13:21.906 --> 00:13:31.306
With ideally, like optimizing your day so that you can tell more stories as opposed to, you know, a larger media entity saying like, "Great, I can have less people again tell the same amount of stories."

83
00:13:31.706 --> 00:13:42.246
Well, I think that the challenge there is, you know, when... I, I always saw this within, you know, small media organizations because you have to do more with less, right?

84
00:13:42.286 --> 00:13:48.896
So you, you go through these periods where you have people doing four different things, right? And you... There's total trade-offs, right? Because they don't... They're not...

85
00:13:48.966 --> 00:13:58.566
Nobody's good at, at four different [chuckles] things. Rarely- Right... are they. But, you know, using the, the... And you have real business needs, but that's the real challenge.

86
00:13:58.626 --> 00:14:13.026
And but using the tools would seem like a easy win, not easy, but it's a win for publishers in that, you know, you see it all the time with I always sort of ruefully joke the copy editors go first, right?

87
00:14:13.086 --> 00:14:24.706
I mean, it's still ne- y- copy still needs to be edited, and I think a lot of times people go to replacement when these technologies are their best, at least the more I use them, like they augment, you know.

88
00:14:24.766 --> 00:14:37.526
They're, they're an augmentation to me because I'm the writer. I... This is a unique [chuckles] I'm like the writer, I'm the editor, I'm the art director. I, I'm like, you know- Yeah... I gotta do it all.

89
00:14:37.786 --> 00:14:47.906
So I love AI [laughs] if it can give me the ability to do more with less and scale myself. Yeah, I, I think that's typically, I think, where we're...

90
00:14:47.946 --> 00:14:52.966
Yeah, where we're at in terms of the life cycle of understanding a lot of the tech that's out there.

91
00:14:53.086 --> 00:15:03.706
One, I would say early innings for AI in general in the construction of what, you know, what it's used, what it's good for, even for our own industry. It's simply not gonna replace a human being yet.

92
00:15:03.726 --> 00:15:11.815
It's not good enough, even at the most finely nuanced version of it to do that, which is a great thing. We believe that humans should never be replaced.

93
00:15:12.186 --> 00:15:21.016
But it's certainly good enough to get you examples of how you need to optimize that article for different social platforms, right? It's certainly good enough to get you to a place where- Yeah...

94
00:15:21.026 --> 00:15:30.826
it can provide you with a starting point summarization of your articles for your newsletter. It can certainly bullet point out longer form pieces of content. Yeah. It can do a lot of things.

95
00:15:31.366 --> 00:15:39.586
It cer- and it definitely can do a lot of SEO optimization suggestions, which I like to say is kind of robots talking to robots, really, if you think about it. Right.

96
00:15:39.886 --> 00:15:49.676
[laughs] It's better to leave it to the robots to sort it out. Yeah. No, I think that's a great point, I mean, because we always go to all or nothing, and it's...

97
00:15:49.706 --> 00:15:54.625
Things are rarely like that, and I mean, I, I use AI, like, all the time.

98
00:15:54.746 --> 00:16:04.506
Like, last week's podcast, afterwards, I, like, ran the transcript through Claude and was like, "I'm gonna write an intro to this, and I want you to, like, give me bullet points of your...

99
00:16:04.566 --> 00:16:16.146
what you think w- were the biggest and most important takeaways. Focus on this, this, and this particular, no more than 750 word," you know? Mm-hmm. And did I read that? No. Of course I didn't.

100
00:16:16.666 --> 00:16:30.316
But did it, like, spur connections that, you know, ultimately got me instead of going back and re-listening to the whole podcast to, to understand? Like, that saved me, I don't know, two hours, right?

101
00:16:30.506 --> 00:16:40.436
And I think the result was actually more comprehensive. I mean, this is a very small and basic example, but when you start to, when you start to do that across everything- Yeah...

102
00:16:40.436 --> 00:16:53.486
you start to add up, you know, some real gains. Yeah. What if you're a newsroom doing 10 or 50 or 600, like the LA Times was when I was there, you know? That's- Yeah... two hours of savings per, per editor per day.

103
00:16:53.766 --> 00:17:05.226
That's an incredible amount of savings, and ideally is utilized to then find more stories or help tell those stories in a better capacity, uh, moving forward. Okay. So which specific areas?

104
00:17:05.286 --> 00:17:19.926
You've mentioned, like, areas like SEO, tagging. There's, like, a lot of... 'Cause AI, I think originally or just initially gets pointed at, quote unquote, "cognitive manual labor." And I don't mean it in a bad way.

105
00:17:19.986 --> 00:17:27.606
We all do cognitive manual labor. Sometimes, like, a good part of my day [chuckles] I, I feel is cognitive manual labor. But do you know what I mean by that?

106
00:17:27.646 --> 00:17:36.806
Like, it's usually pointed at these kinds of use cases, and for anyone who's not operated within, like, a newsroom, there's a lot of cognitive manual labor.

107
00:17:36.826 --> 00:17:52.206
[laughs] I think where I saw this in my career being frustrating is even something as simple as excitement from an editorial perspective means that they're not adding the right tagging structures to the story before they push it to go live because they broke it, and they're really excited about it.

108
00:17:52.786 --> 00:17:57.076
And then, you know, because they didn't do that, they didn't take the extra few minutes to figure that out- Yeah...

109
00:17:57.076 --> 00:18:05.456
they start immediately getting down ranked by everyone who had to follow that breaking story with- Yeah... a regurgitation of their own story functionally. So I think that's a good use case.

110
00:18:05.456 --> 00:18:14.766
I, like, wouldn't have tagging, I wouldn't have tagging for our articles because I knew that the reporters wouldn't do it, and I would be spending, like, the rest of my days there bugging people to, like, tag stories.

111
00:18:14.986 --> 00:18:24.126
And I was like- Yeah... "Ugh, I'm just giving up." I wouldn't have to give up if, if [chuckles] if AI was doing it for them, 'cause then it wouldn't be s- such a big chore. So that's the...

112
00:18:24.256 --> 00:18:35.978
There's three pieces of that, what you just said, that makes my company viable, and that's exactly-These are points of friction, and everyone's kind of everyday- Reporters will never do the extra step They'll never do- [laughs]

113
00:18:36.018 --> 00:18:43.778
Even w- even with the best of their intentions, and frankly, that's because they shouldn't be charged with having to figure out how to do that. I know. It drove them crazy.

114
00:18:43.838 --> 00:18:53.978
I'm sorry to any former, former reporters of mine who are listening to this, 'cause I would make them do their own art and stuff like this, and every single time I would ask where a story was, they're like, "I'm just looking for art."

115
00:18:54.038 --> 00:19:03.078
[laughs] I was like, "If I'd asked you this two hours ago, I feel like this would be the same answer." Yes. But what about like going beyond that of creating new, new experiences or new products?

116
00:19:03.108 --> 00:19:15.248
'Cause one of the things that I am, I'm really on is the more I use Perplexity and these, these various tools, and yes, it's very early, disclaimer, disclaimer, disclaimer, et cetera. Mm-hmm.

117
00:19:15.638 --> 00:19:29.578
The more I think about how expectations will change, and I just cannot, you know... I think that the, the expectations have a- already changed. I go back to, you know, same-day shipping, right?

118
00:19:30.218 --> 00:19:40.708
When, when people got same-day shipping from Amazon, right, or one-click, any of the, you know... It just becomes like the expectations. You go and someone's like, "It's gonna be five days." I'm like, "What?

119
00:19:41.158 --> 00:19:50.798
What are you talking about?" Right. Like, no. Like [laughs] we get things and then we just complain about them. It's oh, this Uber said it was gonna be seven minutes away, and it's nine minutes, and it's...

120
00:19:50.838 --> 00:19:56.038
You're ordering a car that's gonna show up like off your phone, which is kinda nuts. And I- Yeah...

121
00:19:56.758 --> 00:20:08.958
so I wonder about how publishers can-- We'll get to some of the defense stuff, but like how they can use this to keep ahead of consumer expectations, if you know what I mean.

122
00:20:08.998 --> 00:20:11.788
So let's quickly hit on internal expectations. Okay.

123
00:20:11.818 --> 00:20:34.058
Because that, I think, to your point, has gone from fear they're gonna take our jobs to, you know, I, I think the life cycle, it's fear they're gonna take our jobs to we're trying some stuff internally to, oh, this really works, to now almost everyone is using it in some capacity individually, but not really talking about it, to, okay, there's some enterprise grade tools that are specific for our use cases, let's start incorporating those, which are kind of like ours.

124
00:20:34.398 --> 00:20:44.018
And now we're at a place where we just launched a thing called Tone Builder, which we can get you down to the specific tone of voice of a journalist as opposed to just the, you know, the publication at large.

125
00:20:44.498 --> 00:20:50.568
Because that was a, a massive request. Oh, wait. How does this work? I like this. I wanna use this, Josh. Yeah. Well, let's do it. It's just...

126
00:20:50.627 --> 00:21:01.358
It's, it's like a lot of when we say AI, I think we will have more nuanced terminology that we can deploy in the future, but it's just, just taking a corpus of data that's specific to- Yeah... you, Brian, and your...

127
00:21:01.388 --> 00:21:13.958
And ingesting that into our model and then outputting a version of that tone that you and I can talk about. Is this you? Are you more bombastic? Are you more regimented in your, you know, word use of adverbs?

128
00:21:14.338 --> 00:21:20.438
That kind of thing. And we get to a place where all the outputs of all of our tool sets are then in your structural tone. Yeah.

129
00:21:20.498 --> 00:21:30.778
I mean, that's like the scaling- So it gets you from 80% of the way there to 90% of the way there, and we already get people who say, "How do we get [laughs] like 99% of the way there or 100%?" Yeah. And we say, "What?

130
00:21:30.798 --> 00:21:38.298
Remember two years ago when you were afraid of them taking your jobs? Now you're functionally telling me that you wanna train them to do 100% of your tone of voice?" Yeah.

131
00:21:38.318 --> 00:21:45.608
Well, it's about sca- I think everyone needs to scale themselves to some degree, I think, in this world. And it's...

132
00:21:45.658 --> 00:21:56.458
A lot of it is, to me at least, the ones that I, m- I'm using now, a lot of it is to get things 80%, and they're not quite at 80% there. But I'm like, "I'll take 60%." I mean, we're talking about 0% otherwise.

133
00:21:56.798 --> 00:22:00.458
So I'm like, I will take them particularly for certain tasks.

134
00:22:00.518 --> 00:22:21.248
And I feel, like you said, it's almost like people don't wanna talk about how they're [laughs] using AI tools because of some of the ham-handed, I don't know what other word to use, of the examples early on, which were, oh, let's replace, let's, let's pass off some robot, you know, writing something- Yeah...

135
00:22:21.258 --> 00:22:29.558
and let's gin up some, like, fake profiles. It's... Yeah, I mean, obviously that's not the right way to use AI. I mean, must we say this? [laughs] Apparently we must.

136
00:22:30.038 --> 00:22:40.858
I mean, how much- I don't care what the technology is, lying to your audience, generally bad. [laughs] Never a good idea. Never ever. I mean, that's, that's also like a, a Rubicon, yeah, we have not crossed, as we say.

137
00:22:41.218 --> 00:22:49.257
Everything that we create, a human has to be involved in pretty dramatically. Right. And we feel as if that's the way that you need to be. Always we're just being assistive.

138
00:22:49.748 --> 00:22:58.818
That's internally, that's just been interesting to get to that point, right? Where it's they're taking our jobs all the way to, like, trying to ask us how they can get closer to taking [laughs] jobs with outputs.

139
00:22:59.178 --> 00:23:05.418
And then externally, I think the big opportunity is I still see it as being multimedia for the first real time.

140
00:23:05.938 --> 00:23:18.678
And I mean multimedia now by saying, like, I create a story in the function that I'm best suited to create that story in, and now I put it into this tool s- structure, and it can become anything else you need it to be with the same story core at its heart.

141
00:23:19.098 --> 00:23:23.458
Oh. The same- Yeah, well, that's-... yeah, core story. Yeah, that's right... yeah, that's the magic of media, right?

142
00:23:23.638 --> 00:23:33.988
It's like you create one, one thing, and then you create more value out of it, either in different windows or in different formats, and, and that's when you start to unlock value where...

143
00:23:33.998 --> 00:23:44.458
And that's what I talk about scaling yourself, right? I'm always trying to do-- You know, I do two podcasts and I also write two newsletters. I try to link them together in order to, like, not...

144
00:23:44.518 --> 00:23:48.118
I mean, there's only so much I'm gonna do [laughs] throughout, throughout the week.

145
00:23:48.608 --> 00:23:56.657
And trying to wring new value out of, 'cause there's people who listen to these podcasts that don't read the newsletter or, you know, vice versa.

146
00:23:56.718 --> 00:24:11.958
And so you're not, you're not trying, I'm not trying to, you know, just be duplicative, but I'm trying to create new value in, by, by, just by transferring a piece of intellectual property, in quotes, into a new format.

147
00:24:12.078 --> 00:24:19.108
I mean, it just makes sense. Right. And that's gonna get wilder. I mean, if we wanna get- How wild?... with an articulation- [laughs] Let's go wild. How wild? Or what are you talking about?

148
00:24:19.108 --> 00:24:28.908
My favorite example is, like, I want a cartoon dinosaur to sing me the top 10 stories of the morning in K- like a K-pop tone of voice. I think that's stuff- Which is- Do you think that stuff is really gonna take off?

149
00:24:29.438 --> 00:24:39.548
It feels novelty, doesn't it? It does seem super novel, but it also seems like we're the, we're in the first inning, right? I think-Expectations shift and grow- Yeah... over time. It's true.

150
00:24:39.578 --> 00:24:48.218
I certainly know that that, what I just described is possible today. I know. I know. We can create it today. We did something with People Vs. Algorithms, this other podcast that I do.

151
00:24:48.318 --> 00:25:00.758
Alex, my co-host, downloaded all of the transcripts of our podcast for the last year and a half or something, and uploaded it to NotebookLM, which is Google's AI tool. It's not Gemini.

152
00:25:00.818 --> 00:25:13.098
I mean, they've got, like, so many things going on. I actually li- like NotebookLM a lot. And it created a podcast about our podcast that summarized, like, the views and everything, and it was, it was fine.

153
00:25:13.138 --> 00:25:20.978
It was interesting. It got some things wrong. It said Alex was a former design head at, uh, Facebook, and it was Airbnb. But it was interesting. Like, I'm not...

154
00:25:21.038 --> 00:25:40.838
You know, it's one of those use cases where I'm like, I'm not totally into, but then I was like, wait a second, if you could train this on my tone of voice, like, I've produced an, an unholy amount of content, like, [laughs] for many years, and, you know, I'm always trying to figure out ways to...

155
00:25:41.058 --> 00:25:51.338
The, the secret of any publication, if there is it, is just consistency, right? And consistency of voice. You want the advertising to sound like, you know, the same.

156
00:25:51.438 --> 00:26:01.498
I think, you know, the best and maybe the worst at sometimes parts of The Economist are like, it, it reads like, you know, one, one perfect, you know, person who's, who's writing it. Mm-hmm.

157
00:26:01.578 --> 00:26:13.098
And I don't know, I think things like that can, can become, you know, really beneficial, particularly, you know, the, the, the reality is it's, as I said, with more and with less, you're just gonna have to.

158
00:26:13.158 --> 00:26:19.838
There's not gonna be a choice. And it's not like a morality play. Don't, don't fool your audience, right? Right.

159
00:26:19.878 --> 00:26:26.818
If you, if the AI creates terrible stuff, then, like, the audience is gonna see that it's gonna be terrible, right?

160
00:26:27.378 --> 00:26:35.778
I don't see the benefit right now of anyone saying, "We're gonna have the AI do, do, do the work of reporters."

161
00:26:35.938 --> 00:26:44.018
And if they are, then that reporting work wasn't that valuable, 'cause from what I see from, from a lot of these AI engines. I think that's right. That's a good distinction for right now. I, I...

162
00:26:44.298 --> 00:26:54.338
Maybe in the future they'll be good enough to do that, and I don't know what that looks like. But right now, that's what the value of all this is, is we're, we're telling human-authored stories to other humans. Yeah.

163
00:26:54.398 --> 00:27:00.788
I think the format in which those, those people wanna consume that is the benefit or the, the- Yeah... opportunity that we're faced with.

164
00:27:01.338 --> 00:27:12.558
There's gonna be some grotesque union battles o- on this because it's like if you're training, if you're training the AI on, on, on people's style and approach, like, what is your...

165
00:27:12.938 --> 00:27:22.207
You know, I mean, it's a good segue, actually, on, on the sort of initial battle, 'cause I know you get the, you get the questions all the time about the AI- Yeah... crawlers, right?

166
00:27:22.758 --> 00:27:24.478
Is that the first question that you get? Yeah. I mean, we get all the...

167
00:27:25.058 --> 00:27:37.398
Now, the first question we get, if it's combative or, you know, interested in something, is probably how, what's the value of our archives to foundational model companies?

168
00:27:37.878 --> 00:27:47.408
And the second one would be, yeah, like, what, what data do you use and what data do you keep, and how do you, you know, disseminate between those two? Those are very normal questions that we get.

169
00:27:47.408 --> 00:27:52.038
You mean like Nota, what data do you guys keep? Yeah. What- Yeah... what, for us, for our tools.

170
00:27:52.098 --> 00:28:00.778
So you're running all of your, your editorial through our tools, pretty much every story once you start getting the hang of this. Right. Do we keep all that? Do we not keep all that? What do we do with it?

171
00:28:01.318 --> 00:28:06.998
So those are pretty normal conversations. Okay, well let's talk about that, 'cause they're both, like, kind of related, right? Yeah. I think- Very much...

172
00:28:07.018 --> 00:28:18.398
a lot of, a lot of the initial, you know, the, the sort of, you know, negative, as you said, we put the positive first. A lot of the sort of negativity around this is, you know, it's, I think it's warranted, right?

173
00:28:18.658 --> 00:28:26.838
Is this is sucking up a bunch of our content. We never consented to this. This is not maybe, maybe legal. I don't know. We'll see.

174
00:28:26.938 --> 00:28:32.858
But like, you know, robots.txt was supposed to, you know, they, they just ignored that anyway.

175
00:28:32.978 --> 00:28:46.458
So, you know, like in our, in our survey, almost a third of people said that they were blocking some kind of LLM from, from crawling their content. And this is something that comes up in any of the dinners that we do.

176
00:28:47.018 --> 00:29:00.718
It comes up about what to do about, about these, these AI companies. And most publishers are cutting deals, you know, 'cause they don't have a lot of leverage. That's my sort of takeaway.

177
00:29:00.858 --> 00:29:10.078
You might not think it's right, but it's, I don't see how certainly individual publishers can, can have much leverage in any negotiation.

178
00:29:10.618 --> 00:29:21.358
And then on top of that, you just see that the industry is not going to have some sort of united front for a whole bunch of different reasons. You know, some antitrust, some just the structure of the industry.

179
00:29:21.738 --> 00:29:33.178
And where do I even start there? Yes, I, I'd say that they ask these questions of us in terms of have we seen the deal structure of anyone else broadly, you know, not specifically obviously.

180
00:29:33.198 --> 00:29:48.778
But I think the narrative that we always give back to everyone, and we're hearing this more and more, is, you know, this is probably the fifth version of this in the last 15, 20 years, where it's, "Hey, tech, big tech says we're creating a new technology and a new platform that's gonna be robust.

181
00:29:48.808 --> 00:29:58.438
We haven't figured out the marketplace for it yet, so we need your data. And once we get the data and get the marketplace figured out, we'll, you know, we'll figure out monetarily how this works with you."

182
00:29:58.738 --> 00:30:03.208
And it inevitably does not work very well- Yeah... for media, and it works very well for them.

183
00:30:03.738 --> 00:30:13.018
So I think the opportunity here is to not sit on the sidelines actually, but to start trying to set the marketplace value of your data with these companies.

184
00:30:13.378 --> 00:30:34.398
And I think that if you, you could probably argue in a mid to long-term scenario that this forces some of the larger players that have not been very good at paying Google specifically for data to start paying, because their competitive set that's coming up, whether that be OpenAI search product that they're releasing, Perplexity, you know, or another one, maybe Pro Rata, for example.

185
00:30:34.948 --> 00:30:34.948
Yeah.

186
00:30:34.978 --> 00:30:49.246
Because they have that RAG structure, they're gonna need local market, very specific news data to prove out their search products.Explain, explain RAG for those who don't know, 'cause it's actually a key term in all this Yeah, retrieval augmented generation.

187
00:30:49.326 --> 00:31:00.286
So functionally, it's making sure that it's, that the output of the AI is based on something that it has the ability to retrieve and augment that output with.

188
00:31:00.366 --> 00:31:26.266
So for an example here would be a news story if they have access to all of News Corp's data, so they would have access to all of their publications in terms of digital sites, and you're searching for something in a Fox local affiliate, they've told a story about it, that search will then augment, augment the retrieval, but basically augment the output of what it's trying to tell you based on the data of, of Fox.

189
00:31:26.666 --> 00:31:40.906
Hmm. I have a couple of axioms in life. One is that most parents overrate the, uh, talents of their children, and the other one is that publishers overrate the value of their data [laughs]. That's- [laughs] Yeah.

190
00:31:41.046 --> 00:31:48.946
I, I think that everyone thinks that an archival deal is probably a conversation if they have not had these with foundational mo- model companies.

191
00:31:49.346 --> 00:31:58.166
They're thinking they might at some point, and they wanna know how much they get paid for it. But what we're hearing is, you know, until the laws, to your point, are passed, it's just a polite...

192
00:31:58.726 --> 00:32:03.896
The robots file is just a polite way of saying, "Please don't take this." It's not legally binding- Right... is the problem.

193
00:32:04.136 --> 00:32:11.906
So these companies are just gonna say whatever they're gonna say, but they're, as we've seen in anyone who does these tests, they're still taking the data, not stopping them.

194
00:32:11.966 --> 00:32:18.526
It's just a, a polite suggestion at this point. Right. Yeah. And, and, and they've raised a lot of money.

195
00:32:18.586 --> 00:32:26.466
They have a lot of money, so- A lot of money, and they have an ethos that's very spec- I know everyone's talking about the Eric Schmidt, like what he said, "Steal it until you can't," basically. Yeah.

196
00:32:26.926 --> 00:32:36.266
But that is- Well, they have a track record... the ethos of technology companies. They have a track record. I mean, I was running that Charlie Brown trying to kick the football GIF for a decade [laughs]

197
00:32:36.506 --> 00:32:40.206
from any of the stories about publishers dealing with these same companies.

198
00:32:40.286 --> 00:32:47.786
So I mean, at some point when you come around, "No, no, no, this time, this time it's gonna be completely different, and we're in this as business partners," so...

199
00:32:48.086 --> 00:32:55.266
Uh, it'll be interesting because, look, the reality is, you know, there's overall like an ethos in a lot of these companies that they don't need publishers.

200
00:32:55.346 --> 00:33:06.046
They've sort of walked away from news at Google and at Facebook, and, you know, Mark Zuckerberg was just out there just recently saying, you know, "Publishers don't want, that's fine.

201
00:33:06.086 --> 00:33:14.566
Like, I, we don't really, you know, we don't really need publishers at all for LLaMA." So, you know, a lot of this is just the marketplace will sort it out.

202
00:33:14.586 --> 00:33:26.426
But to me, like it would, it, it might actually be a weird situation, 'cause I look at it, it's like the most valuable content the publishers have been creating in, in the previous generation was evergreen content.

203
00:33:26.486 --> 00:33:39.726
It was library content because the, the, the structure of search and how search was the front door, and news, the value was never... Uh, uh, the value just went down, right, when it got published, right?

204
00:33:39.826 --> 00:33:52.366
If it's news- Mm-hmm... there's new, right? Whereas in this, it seems like, you know, the value of news could actually go up in some ways because there's a ton of archival content out there, right?

205
00:33:52.436 --> 00:33:58.466
But like the newest, freshest stuff, won't these models need that? It depends on what they're using that for, right?

206
00:33:58.506 --> 00:34:07.966
So if they have a search product, then absolutely they have to have it, because the only benefit of new news, to your point, is for search.

207
00:34:08.026 --> 00:34:21.706
Like, people wanna know what's happening right now, and if they don't have, if these companies don't have the ability to get that information from you because you figured out a way to block them from doing so or legally they're not allowed, then that's not a very good product.

208
00:34:21.946 --> 00:34:25.366
So they won't, they'll, they'll fail compared to their competitive set. Yeah.

209
00:34:25.706 --> 00:34:38.366
So give me like the optimistic take for, you know, what a grand bargain would, would look like that leaves publishers like in a, i- in as best of a possible position. Wow. Okay. So

210
00:34:39.486 --> 00:34:50.996
if I'm a, let's say I'm a publisher then, and my hope would be that regulation happens where it's, it's deemed to be you're unable as a technology company to just take any data that I put out there.

211
00:34:51.086 --> 00:34:59.426
I don't know what the mechanism for that would specifically be. I don't think anyone really has the perfect idea of it, but you have to pay me for it, at least from a percentage perspective.

212
00:34:59.666 --> 00:35:10.926
So augment or help me achieve, you know, the creation of that actual labor that goes into telling that news story, give me some version of that back in terms of revenue. Okay. Right? And- So like licensing...

213
00:35:10.986 --> 00:35:14.246
do that at scale, do that consistently, and do that pretty transparently.

214
00:35:14.346 --> 00:35:24.806
But I think what is also an opportunity in this marketplace that no one's really talking about as much anymore, or yet rather, is not just foundational models, but all the applications building- Yeah...

215
00:35:24.866 --> 00:35:35.286
all this other stuff that needs that data to be successful. I think- Yeah... an example that someone gave me last month that really resonated was American Express Travel, I believe.

216
00:35:35.366 --> 00:35:44.206
You know, they have a chatbot, they have a whole travel experience, user points here, but how do they t- what if I'm asking, you know, "What's the best time to go to Paris?" Or, "Where should I stay there?"

217
00:35:44.386 --> 00:35:52.126
So Conde Nast Traveler, wouldn't that not be, that archive would be super valuable for that chatbot. Yeah. There's thousands and thousands of those examples.

218
00:35:52.626 --> 00:35:58.976
Yeah, I think it's like a lot of the focus ends up going just on the, like, the LLMs, you know? But there's- Yeah...

219
00:35:58.986 --> 00:36:11.086
there's a lot that, that goes on and, you know, you could see a situation where the, the licensing line and publisher balance sheets just go, it, it gets bigger.

220
00:36:11.266 --> 00:36:24.446
And it, and you know, the, the deals on themselves might not be massive, right? But if you're cutting a lot of them, it can add up, you know? But I think those- Yeah... mechanisms have to get in place.

221
00:36:24.506 --> 00:36:34.126
It's really interesting also from a use case perspective, the more nuanced and individualized that the publisher data is, the better they're suited for stuff like that.

222
00:36:34.226 --> 00:36:49.198
So if you're a medical journal, all of a sudden you have a lot more interest from a, a wide variety of applications than I think you would have had a year ago.And then from a data marketplace perspective, those four or five companies that are emerging are trying to do that.

223
00:36:49.278 --> 00:37:01.758
I think provide transparency from a, a basic CPM model converted to per word. That's really what we're talking about. Oh my God. Price per word's coming back. [laughs] Who knew? [laughs] Everything old.

224
00:37:01.778 --> 00:37:07.798
Future looks like the past. I still get people sometimes telling me, it's, "Well, how much do you pay per word?" I was like, "Oh, wow, we're still doing per word."

225
00:37:08.098 --> 00:37:16.118
I don't like the incentives there, so I'm like, "Won't you just write really long?" [laughs] That's what I would do. Like, you're gonna get- Yeah... 3,000 words whether you want them or not.

226
00:37:16.538 --> 00:37:27.658
So we might have, like, much longer content. What, what... Do you, do you see, like, internal resistance to, you know, within editorial teams to, to using Notta? Not anymore.

227
00:37:27.718 --> 00:37:34.018
Mostly because we've been very cautious about how we go to market in terms of staying away from anything that isn't just assistive.

228
00:37:34.438 --> 00:37:40.638
So honestly, the places we're going next are more back of house than front of house editorial work.

229
00:37:40.758 --> 00:37:50.208
So archival retrieval is a problem that I've never, I've never met a, a company that had great structured archives, so that's a, that's an area of improvement.

230
00:37:50.758 --> 00:38:02.598
That also gets back to, you know, the, the possibility of IP. I think creating packages for IP based on your archival retrieval is a huge opportunity for the market. You can put per...

231
00:38:03.138 --> 00:38:16.118
Imagine, imagine putting parameters as easy as, you know, this had to be a murder, this had to run for a month consistently, we had to have at least 100 images, we had whatever you want, and it just, like, goes through your archives and says, "Here's a story that's all packaged up.

232
00:38:16.338 --> 00:38:26.058
This could be really compelling for the next great documentary or whatever you wanna sell it for." That's exciting, but also the real time nature of that in terms of augmenting what you're already working on.

233
00:38:26.158 --> 00:38:36.758
So if you're writing in a CMS, it should be able to pull up everything that's been written about that subject whenever, in whatever capacity you want. Like, I want the, the last three stories written about this subject.

234
00:38:37.158 --> 00:38:49.448
I want all the images related to the subject that we've already published. Just save a ton of time and then add a lot more value. Okay. So you're, like, on top of the CMS? Yeah, we integrate within all the CMSs. Okay.

235
00:38:49.448 --> 00:38:59.358
Or not all of them, but a lot of them. And they won't build these tools themselves, or they'll try to, but you're a specialist, so your tools will be better, right? Yes. [laughs] I believe that.

236
00:38:59.738 --> 00:39:06.758
Yeah, I mean- Well, no, it's always the question. It's like, well, wait, if you're on top of it, won't the underlying technology just add that to their, their own stack and...

237
00:39:07.258 --> 00:39:12.558
I j- I feel like I've asked that question- They will add versions of this... a lot over the years. [laughs] I think they will add versions of this.

238
00:39:12.698 --> 00:39:22.567
And, you know, part of what I have to do as the CEO here is incentivize them to find value in just having our built- Yeah, yeah... and operated tools- Like-... as opposed to trying to build them themselves...

239
00:39:22.578 --> 00:39:27.437
I was gonna say, like, I've asked this question a million times. I don't know how many times I asked companies over the years. I'm like, "Why won't Google just do this?"

240
00:39:27.938 --> 00:39:37.578
And they're like, "Do you know how many things Google is doing?" [laughs] They're like, "This is all we do." I remember, like, Balance Exchange. I used to love Balance Exchange. They, they're like Wunderkind now.

241
00:39:37.638 --> 00:39:44.658
They like... When they were, the only thing they did was, like, exit intent, was get people to give you their email address. Mm-hmm. They were really good at it.

242
00:39:45.238 --> 00:39:56.368
So many other, like, platforms that we use would be like, "No, we can do that too." I'm like, "No, but this is all they do. They're really good at it." [laughs] Yeah. That's it. Okay, so what about the data question?

243
00:39:56.438 --> 00:40:05.108
How do you... I assume you're not stealing, uh, people's data and then selling it out the back door. No. [laughs] This... Yeah, this is where I wanna... I'll, I'll break that news. [laughs] That's what I'm doing.

244
00:40:05.278 --> 00:40:13.038
Oh, shit. [laughs] Uh, no. We're... I mean, I think that's another area that we get a lot of questions and we're very conscious of. Yeah. And I spent a frightening amount of money with lawyers.

245
00:40:13.078 --> 00:40:17.698
That's one of my maxims, I guess, in life is like- [laughs]... lawyers win. Yeah, no, lawyers always win.

246
00:40:17.858 --> 00:40:26.618
[laughs] The lawyers are to- Although this one, you could see it being at least a draw because you can, you can, you know, the legal bot can, can get you, you know.

247
00:40:26.658 --> 00:40:36.038
You might not need as many lawyers, but the lawyer, the lawyers that have remained will win. Yeah. So the idea just there being we wanna be very straightforward and transparent about use of data.

248
00:40:36.448 --> 00:40:42.768
And there's benefits to giving us more of your data, and there's benefits to not giving us any. It really depends on how you wanna structure it with us.

249
00:40:43.218 --> 00:40:52.298
By not giving us any of your data, we're not learning about anything that you're doing, so you're just relying on our model version to just give you one-shot options- Yeah... for what you create.

250
00:40:52.818 --> 00:40:59.938
And then- But I guess it's-... we talked about Tone Builder. Yeah. It's like how do we take that data? Do we utilize that data? Do we keep that data? What happens to the data?

251
00:41:00.258 --> 00:41:08.198
And then broadly, do you want your own instance of our entire company basically built for you that goes nowhere else and you control?

252
00:41:08.208 --> 00:41:13.358
And that's the, the highest level of the other way that gets back to kind of like no data again almost. Yeah.

253
00:41:13.738 --> 00:41:27.138
So I guess it's just you're gonna be transparent about it, and there's trade-offs to different approaches, and, like, companies need to end up making, making those trade-offs at the end of the day, right? Yeah.

254
00:41:27.178 --> 00:41:37.148
I mean, we're... It's still just... I think part of what gets lost with a lot of this is it's just a, a higher level version of machine learning, and machines can only learn when you give them information.

255
00:41:37.738 --> 00:41:48.478
So if you limit the amount of information you give them, your outputs are not gonna be as nuanced as you would hope them to be. So there needs to be some sort of understanding there that there is that trade-off.

256
00:41:48.898 --> 00:41:55.338
Now, from a foundational model perspective, I think we saw a lot of the conversations happening around we can't use...

257
00:41:55.878 --> 00:42:04.958
We don't wanna work with anyone in AI because, you know, they've been training on this data that was illegally taken, which I think that's a little bit of hyperbole right now. Mm-hmm. It's not necessarily illegal.

258
00:42:05.198 --> 00:42:13.958
It just may be illegal in the future. But the reality there is, is there's no way in which anyone besides five or six companies can compete at the foundational model perspective.

259
00:42:14.018 --> 00:42:21.698
We have open source models that are accelerating their ability to be almost as good, but a lot of those are built by some of these large tech companies.

260
00:42:21.958 --> 00:42:33.628
So what we've done in the background, we've just created effectively the lattice work around all of these model structures, and we are constantly training new ones, both open source and closed in the back end, and then we're chain modeling.

261
00:42:34.178 --> 00:42:45.758
So a lot of our outputs that you see, one output will be from one, one weighted version of our subsetted model structure, another will be from another. Yeah. And that's just because they're good at different things.

262
00:42:46.158 --> 00:42:55.488
So when you talk about, like, you know, tagging and SEO and, and, and even then, but then you get into, like, Tone BuilderPlay this forward, though, like three years, right?

263
00:42:55.788 --> 00:43:05.948
What is this-- how, how is this technology being used, and how is news being produced then in... I mean, it's hard to look ahead. I was gonna do five years, but it's impossible.

264
00:43:06.268 --> 00:43:18.928
I have a feeling it's gonna be incredibly different in a, in a lot of different ways. Not just operationally, but I think the way people are retrieving information is gonna be radically different.

265
00:43:18.948 --> 00:43:35.528
And if the way people overall are retrieving information is gonna be radically different, I just do not see how, you know, news companies that are in the information retrieval business at the end of the day are not gonna end up, you know, what they, how they operate and what they produce is not gonna be radically different.

266
00:43:35.948 --> 00:43:40.028
I think that that's right. I don- I hope it's three years, but it's probably-- you're, you're probably right.

267
00:43:40.468 --> 00:43:46.048
I think we can, we can extend the time horizon because of our own industry to five, and it's still, like- [laughs] Yeah... what we would predict in three.

268
00:43:46.688 --> 00:43:58.088
But yes, I think structurally where we, where we've been has been having conversations for twenty-plus years about, okay, where's the value of this, this kind of entity go? What do we do next?

269
00:43:58.148 --> 00:44:04.628
How do we transform, just like technology has transformed the expectations of everyone to receive information in new ways?

270
00:44:05.038 --> 00:44:11.968
And I think the reality is, like, this technology now gives us the opportunity and a head start over almost every other industry. I wanna point that out again.

271
00:44:12.028 --> 00:44:22.768
Like, it is ready to go, and it works for our industry today. Proven. Works today. Yeah. And many other industries are like, "Oh, novelty, this... Yeah, down the road we could see how this will really make an impact."

272
00:44:22.828 --> 00:44:28.708
But today you can change your entire structure of your company to suit the needs of the future. Yeah.

273
00:44:28.928 --> 00:44:37.208
I mean, that's like- And hopefully what that means is more journalists, less everything else from an infrastructure around it to support it necessary. Yeah.

274
00:44:37.248 --> 00:44:41.968
I mean, my thing is inevitably a lot of organizations will be l- leaner. They'll be smaller, right?

275
00:44:42.188 --> 00:44:52.118
But I think the weight of what those organizations do will be-- my optimistic take is that it will be more aligned with the overall purpose of the company.

276
00:44:52.158 --> 00:44:57.208
I don't root for anyone to lose their job, whether they're in ad ops or anything, right?

277
00:44:57.248 --> 00:45:10.048
But for these companies to fulfill their missions, they're going to need to, to cut down and be more efficient with a lot of the in between that is not making the product.

278
00:45:10.128 --> 00:45:31.007
So I mean, just to take, take a step back, I mean, the whole goal is to have a sustainable news ecosystem, and I don't-- it's not like you're gonna get there by having a flood of money coming in, like, from the government, or all of a sudden Mark Zuckerberg's gonna wake up and be like, "Oh, yeah, I, I, I need to, like..."

279
00:45:31.528 --> 00:45:33.028
[chuckles] No, that's just not happening.

280
00:45:33.128 --> 00:45:55.668
So I, I think, you know, it has to, it has to move in that direction and have to use technologies in order to, you know, do, not just do what, what's being done now, quote unquote, "cheaper", but to do it better and, you know, that- that's where the weight of the industry will, will have to m- move, I would guess.

281
00:45:56.288 --> 00:46:03.977
I would agree. I mean, it just- Yeah... it shifts back to the fundamental thing that humans are best at, which is, uh, storytelling. Yeah. So that's the thing that carries value.

282
00:46:03.977 --> 00:46:18.928
I mean, I'm a little wary of how, of how far it will encroach into some of the [laughs] the storytelling, but I have a feeling that it's gonna put a premium on, on more human, you know, experiences of all kinds.

283
00:46:19.168 --> 00:46:29.507
And we see it all the time with the mechanized agriculture industry. You know, people pay premium for, for all kinds of, like, craft and bespoke things. And you know, there's a lot of...

284
00:46:30.028 --> 00:46:41.068
It's not like it depends on the model, right? Like, there's, there's absolutely room, I believe, I might be in a minority here, for AI-produced content. There is absolutely room for it.

285
00:46:41.208 --> 00:46:46.398
I don't think that The LA Times should do it [chuckles], I don't think The New York Times should do it. Doesn't fit with their models. Yeah.

286
00:46:46.408 --> 00:46:58.628
But I, I definitely, you know, if you look at where, the way these tools are going, they're gonna be used absolutely in, in all aspects of the publishing business, and that includes creation.

287
00:46:58.728 --> 00:47:08.188
Some people will use it more. I mean, I, I think that provides an opportunity because the market can sort it out. People will price, you know, the different content.

288
00:47:08.248 --> 00:47:17.368
You know, some organic meal costs very, uh, different than, you know, a trip to the McDonald's drive-through. Yeah. So. That's a good way to put it, I think. Yeah.

289
00:47:17.628 --> 00:47:25.648
I think there's also just the human desire to understand that someone else is doing this, right? There's, there's that craft- Yeah... I guess, to your point.

290
00:47:25.848 --> 00:47:36.248
I want a very smart person that I respect to have thought through this so that I can get their take about what this is, as opposed to just, like, a regurgitation of the facts. Yeah. Um- I believe that.

291
00:47:36.408 --> 00:47:44.828
I just think it's just the same, it's like they're not gonna be robot waiters. There's just not. [chuckles] Maybe in, I don't know, Jimmy John's or something of that nature.

292
00:47:44.868 --> 00:47:54.278
You're not gonna go into, like, French Laundry and get, like, a robot waiter. Like, I'm trying to think. Maybe in some parts of Japan. I mean, they really like [laughs] that. But anyway- We got that... overall- Yeah...

293
00:47:54.308 --> 00:47:59.488
I think the point [laughs] the point holds. Awesome, Josh. Anything we didn't cover that we should have?

294
00:47:59.948 --> 00:48:10.748
No, I think we even got to some of the fun parts of what's happening from a data marketplace perspective, which is brand new excitement. Yeah. No. Okay. I loved it. Cool. Had a great time. Awesome.

295
00:48:10.968 --> 00:48:23.548
Josh, thank you so much for doing it. This is great. I really appreciate it. Thank you, Brian. [outro music]