New York Times Sues OpenAI, Apple Watch Ban's Implications, Snapchat Rising On AI Success

0:00OpenAI sues the New York Times over copyright infringement. Are Apple's problems with the Apple Watch long-term or a short blip? Amazon brings ads to Amazon Prime Video. Snapchat Plus keeps adding subscribers thanks to AI features. And what is tan mama? All that and more coming up right after this.

0:22You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. we're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to The World on your local public radio station and wherever you find your podcasts.

0:49Welcome to Big Technology Podcast Friday Edition, where we break down the news in our traditional cool-headed and nuanced format. We're here to round out the year, the last show of 2023, three, setting up what's going to be a fascinating and massive 2024. We're here as always with Ranjan Roy of Margins. Ranjan, happy holidays and a happy new year to you. Great to see you, man. How you doing? I'm in historic Lexington, Massachusetts, where I grew up at my parents' house, the house I grew up in, and thought it would be a slow news week, but I'm actually very excited about today's episode. Some stories I think that are going to be very important to set the tone for next year.

1:32Where are you right now, Alex? Absolutely. So I'm going to apologize in advance if I have some sound quality issues. I'm staying at a hostel on top of a lava bed on the big island of Hawaii. It's actually a pretty cool spot, just like in the middle of a lava flow totally off grid, but they do have some Wi-Fi, so we're going to give this a shot. it's not quite suburban Massachusetts, but I guess that'll have to do for you. I think if you average us together, we're like somewhere in Montana. Okay, so let's talk about this really fascinating lawsuit that the New York Times is bringing against OpenAI.

2:09So the New York Times is suing OpenAI over copyright infringement for training on many of its news stories and reviews, including that from the Wirecutter. Now, I just want to start by sharing the context of this, and that this is bigger than just one company, right? It's not just the New York Times that's suing OpenAI. There are plenty of other corporations, news corporations, that are talking with OpenAI about licensing deals. And the sense that I get, and it's being somewhat, you know, not even winked at, but clearly communicated in the coverage of this, is that there will be a slew of lawsuits against OpenAI by these news publishers if they're not compensated in a way that they find fair.

2:46And this is from a New York Times story. The other companies that are talking to OpenAI about licensing deals are News Corp, the IAC, which owns Daily Beast, the magazine publisher Dot Dash Meredith, and the News Media Alliance, which represents more than 2 ,200 news organizations in North America, has also been talking to OpenAI about a deal. And so, Ranjan, I'm kind of curious whether you think that this is something that has the chance of working, whether these publishers are going to get their money, how it might differ from the big Google fight. What can you tell us about this? I think this is massive.

3:22As you said, this is much bigger than one company, than just the New York Times. But I think it's important that it is the New York Times. And Jason Calacanis had an amazing post on this. And it is rare that you will hear me saying that. But he lays out very clearly how in the 2000s, the publishing industry essentially got rolled over by Google while the music industry was strong enough and cohesive enough to take on Napster. The New York Times of today is not that organization it was in the 2000s. It's a gigantic company. It has a lot of technology underlying everything it's doing. It's a different company right now.

4:00the way they've approached it, I think it's clear that this is going to be a much different effort. And it's important because, again, all these large language models have been trained on common crawl, at least GPT-3, GPT-4 on common crawl, which not only indexed and used every New York Times article, it even gave additional weighting to the NewYorkTimes.com domain because it is considered a much more valid domain, that the information is more authoritative. So underlying all the information used to train GPT-3 and 4 was the notion that the New York Times content is more valuable than the average content on the internet.

4:41So I think it's right there that there is value to this. The New York Times, clearly there were discussions around licensing already that must have fallen through because OpenAI did. There are some other publications I know, publishers that they did already sign deals with. But I think this is more than just a cash grab. I think this is really important in setting the tone of how large language model driven companies like OpenAI operate in the future, because the models are not valuable without the content. And the content up until now has been free. yeah they have signed deals with uh the associated press by the way hearing you praise jason calicanis we're just going to go around the horn with all the l and podcast members it's almost new year's it's the season of giving uh from ron john but i did think the jason calicanis post was good and one of the things that really stood out to me was the fact that basically chat gpt was taking wire cutter reviews and spitting them out verbatim and the wire cutter comes with the times like which is a paywall publication.

5:43So do you think that this lawsuit, you know, on its face has some merit? And is there a chance the Times wins? Yeah, so I think there's two important parts to this. One, the wire cutter example is so perfect, because the query that the New York Times used was, you know, what is the top recommended X product from the wire cutter? It provides you that exact answer, which is what any of us go to the wire cutter. I've long-loved the wire cutter, that's what we go to it for. It not only provides that information in a condensed format, it strips out any affiliate link, taking out any value to the New York Times for collecting all that information.

6:23So there's such a clear business impact on the New York Times that ChatGPT and OpenAI are impacting on them. So I think that one is very clear cut. I think the most important part of this lawsuit, though, is going to really be public perception around this. And this is one of the fun things for me to watch, because I think this is going to divide and chop up the tech and tech reporting landscape in ways that are just weird. Again, I'm agreeing with Jason Calacanis today. I think Elon Musk is going to come out strongly in favor of the New York Times, which are words I never would have thought I would say, because I think Twitter's repository of information is of equal value, if not even more than the New York Times.

7:09I think Reddit, even Quora, if you've been on Quora in the last few years, all these places have these incredibly valuable sources of information that should be licensed. They should be able to create their own LLMs and create their own chatbot experiences. So I think this is going to be so interesting in terms of who ends up on what side and how things get portrayed. But I think to the average person, it's just so easy and clear. you can get into the underlying technology of a large language model and how it works. But when you reproduce near verbatim, you know, content and journalism that the New York Times invested millions of dollars and thousands of man hours into, it's pretty obvious that something is wrong there.

7:56It's interesting that you mentioned Quora among your groups of sites that are going to need to charge to get paid to need to charge to get trained on because adam d 'angelo who's the ceo of quora is on the open ai board oh yeah going back and forth being like what do i do about this okay but here so you talked about the uh divide between like tech reporting and or reporting in general the news industry and open ai and this is where it really gets fascinating because jason's post ends with um that the new york times is going to take an open source language model train it on their data and create a chat gpt competitor that much is clear and that is like there's been this meme wouldn't you say within the tech industry that the reporters are competing with tech and that facebook wants ad dollars and the publishers want ad dollars and they're showing that that's not brand safe and that we are brand safe and therefore they have an incentive to knock down the tech uh platforms and and even more than that show that we should be trusted and they shouldn't and so if they're actually building specific competitors to tech products, not just tech products, but one of the most successful tech products in modern memory.

9:06How do you think that changes this fight? Because what I wrote down is that the tech versus tech reporting fight is about to go to 10 here. Well, I would like to recall who predicted, I believe it was two weeks ago when we were discussing the hiring of Zach Seward at the New York Times to research general LLM products that the New York Times could be preparing their own chatbot. We talked about this two weeks ago. And that's why that's what I've been thinking. And they should, I 100 % think they should. And they're already a tech company, when so much of their revenue comes from even gaming right now in other products that live outside of the journalism, They've already become a technology company.

9:51So this would just be another product that is even more suited to a journalism company or a publisher than cooking and games, right? Like this is like the idea of a chatbot queering your body of information. Any company should own that if they have a huge corpus of information. So I think, yes, they should build it and they will build something like that. And I think that makes sense. And they would be able to make it much more reliable relative to their information. But I think the idea, I've never bought the idea that because one part of the New York Times will be building a tech product that competes with another tech company, that that will affect their journalists, you know, view on technology and make it more negative.

10:36This is one thing that's always bugged me because if you read the New York Times tech coverage, there is a lot of fawning coverage over plenty of tech companies that I think is irresponsible and just overly optimistic around it. It's only the few examples that are cherry picked that are negative on very specific companies that everyone runs to. So I think – I don't think it's going to impact that at all. Well, merit or not, it's going to be something that's going to be used as a cudgel by the critics of tech reporting to sort of amp this war up. A direct competing product with ChatGPT. Yeah, but that's why it's going to be so fun next year, especially for us to watch sitting on the sidelines with popcorn.

11:16We are going to have a good time next year. Well, no, because that's again, but Elon Musk cannot make that argument and then claim that Twitter is its own valuable source of information and you should pay for Grok, right? Because then he is a competitor to ChatGPT and OpenAI as well. So I think it's just one of the, and then the citizen journalism on Twitter or whatever is going to be unbiased. I think that's what makes this just so incredibly interesting because it divides what we've called the tech, whatever, technology community. It divides it in ways that it has not been in the last couple of years when it was very cleanly media versus tech.

11:56So up until this point, we've basically gone under the assumption that the New York Times is going to have a good claim against ChatGPT because it can produce wire cutter reviews effectively one for one and other types of stories. Let's get more nuanced here, right? Let's do our show mission and go a little bit deeper. So I'll bring up some counterpoints and you tell me what you think. So first of all, there's actually the amount of training that the New York Times has been responsible for for OpenAI. So this is from Benedict Evans. He goes, looking at an estimate that the New York Times is approximately 0.0083 % of common crawl, which is the crawling that some of these trainers use to build their models.

12:41So he says it's 0.0083 % of common crawl, and reminded again of the disconnect between the social and political value of professional journalism and the amount it actually makes up of internet consumption, traffic, and revenue. And journalists' recurrent refusal to accept that news is important to society is not the same as Google's ad business needs news. Your response? That's exactly the point. 0.0083 % of common crawl, yet the third most heavily weighted domain in common crawl. I think that's exactly the point the New York Times is making, that it is a smaller percentage of overall content, yet it's more heavily weighted because it's important.

13:26And that weighting, when you think about how large language models or companies like OpenAI have to think about licensing, is clearly going to play a part because if you remove that information and that source from the model, then it becomes of less value. If you have to rewrite the entire model because the top few domains that are from publishers all demand that their content be completely removed, it completely changes the way you operate your business and the underlying technology. So I think that it is a reminder, yeah, that the overall amount of content is small, but the importance is big.

14:02And here's the thing I think that looking at the actual percentage misses, it's about the urgency, right? Because we all know that ChatGPT lost users across June, July, August, right? It's not like this ever ascending product. It needs something to keep people coming back. Now, one of those things is assistance with schoolwork, which is why it's seen a resurgence in the fall, but has to be broader and more interested to people. And what's the thing that's the most vital thing that's really good for engagement? It's news. Hey, ChatGPT, what's going on in the world today? How should I think about this news event?

14:37And the news can present that value. And so even though it is a small percentage of the actual training on ChatGPT, it does seem like the news publishers are going to be able to make this point to say, hey, listen, it's not just about the volume, it's about the vitality, the vitality of your product. And we provide a large part of that vitality. What's happening today? Yeah, no, I completely agree from the consumer standpoint, this is big. What I do think is interesting about this is, and we've covered this a lot, the end monetary value or business value is not going to be from this consumer side.

15:14It's going to be from customized models that businesses can actually create their own tools, which right now, a lot of people, I built on GPT 3.5 and 4.0 plenty, like it's very usable, but I think this stuff ends up becoming a lot more customized. And that's where I think either this should force OpenAI to start really building up that enterprise side of their business and customizability and just thinking more from an enterprise standpoint. Or in a good way, it also opens up the game to the open source community, to other smaller players as well who are focused more on the business use cases.

15:52So I think this is also a reminder that New York Times versus OpenAI, these are the names that represent these tools to the general public, but it's not to me where the end value is going to be on this. And in a way, maybe this is good to force OpenAI to start thinking about enterprise. So that was kind of the easy bit of nuance to bat away. This is where it gets a little bit more difficult. So Jan LeCun, chief AI scientist at Meta, posts a very interesting question. He says, genuine question about image generation. If someone uses generative AI tool, a generative AI tool to produce an image that is substantially similar to a copyrighted piece, like a drawing, a painting, a movie screenshot, who should be liable for the copyright infringement?

16:39Should it be the company producing the tool, the human creator of the piece, the person or entity posting or publishing the piece, the communication platform through which the piece is distributed? And he adds some bit of nuance here. He says, when a person distributes a sufficiently similar copy of an art piece, It's a violation of copyright, regardless of the tools and processes used to produce it. The liable person is the person distributing the piece, not the artist and the makers of the tool. Sufficiently detailed prompts will produce images that are substantially similar to images from the training set.

17:15It is not at all surprising that a prompt like the Batman movie rooftop scene screenshot 4K will produce an image very similar to an actual screenshot from the movie. Okay, that brings up a very interesting point here, and he posted it right in the aftermath of this New York Times thing, which makes me believe he was subtweeting it a bit. Which is the point, is it the tool that's the problem, or is it the actual distribution? If it's the act of distribution, is the fact that you're now reading a Wirecoder review on ChatGPT not actually a violation of copyright? Is it the point where you actually take that review, copy it, and paste it, and distribute it on your own?

17:55I mean, that's a very interesting way of looking at a very interesting angle, and it's a wrinkle I expect OpenAI to try to exploit. Well, I think the nuance is in where the monetization occurs. In the example Jan LeCun gives is if I'm creating an image or even if I'm using OpenAI to create chat GPT to create text, if I am then monetizing that, then it could be more clear that the liability is on me. But remember, the monetization and the examples of the wire cutter is OpenAI is monetizing that user paying 20 bucks a month on ChatGPT Plus and to get that information. So then clearly the business impact, you know, falls on OpenAI versus the case.

18:39One thing that I think is really interesting from Jan's example is I think the prompts are going to become more important. Like if an image ends up very similar and, you know, a brand is using it, being liable to show the prompt and did you actually include the Batman movie? I think this is going to be really interesting and where copyright law goes on all of this. But at the moment, I think where is the money going? And I think those people will be held liable for the copyright infringement. There's another side to this that's also really important to discuss, which is the reputation issue that OpenAI is now opened up to, where people already didn't trust that the material that they're typing into ChatGPT or the material that their company is using as they interact with chatbots, they didn't trust that that was not going to be used to train.

19:28and as more and more companies raise their hand and say, hey, they trained on my stuff without my permission and I'm suing, it's going to create potentially an environment of mistrust, you know, regardless of whether or not they're training on specific use cases. And that's something that this blogger, Simon Wilson, spoke about saying that AI is facing a crisis of trust. This is an article that you picked out about how Dropbox has this new feature, NUI, that companies can communicate with their data and that people are going to be say, and I'll just read directly, people will say, OpenAI is training from every piece of data they see, even when they say they aren't.

20:04It's the new Facebook's showing you ads based off of overhearing everything you say on your phone's microphone, right? So it is this big reputation issue that AI is going to run into, and it's really unclear how it's going to resolve. Yeah, so the example he used, and this is a really good piece from Simon Wilson, is that we all still kind of believe this literally came up over Christmas vacation for me, that Facebook is listening to you on your phone. And it's a UI feature, or maybe, you know, like problem bug that ads will show up based on things you just said. But I strongly believe that Facebook is not listening to your phone.

20:45I've read a ton about this. And again, they're just so good at triangulating you and where you are and what your other interests are. They can figure out who you are around and what their interests are, but they're not actually listening. But as much as they say that they are not showing your ads, no one believes them still to this day. And as much as when Dropbox launches their Dash feature that's going to plug into your Slack and Gmail and everything else, are we really going to believe they're not training on the data? Same with OpenAI. Now, with their enterprise tiers, they will say over and over, they're not training on your data, and it will be in the terms of service.

21:24But do we really believe that? And that's a good point, that that's actually another major impact from this New York Times lawsuit, that it could, going back to the enterprise topic, it could affect how businesses trust them. We're here on Big Technology Podcast, Friday edition. I'm joined here with Ranjan Roy of Margins. We're going to go to break, and after the break, we're going to talk about Apple. We're going to talk about Snapchat Plus and a few other stories. Before we do, I just want to say thank you all for sticking with us and listening to us all through the year. We've seen the audience numbers rise and it's been amazing speaking with just a growing and engaged community of listeners.

21:59So thank you for that. Also, if you're in the holiday spirit and feel like giving us a five-star review on Apple Podcasts or Spotify, we would definitely appreciate that. And yeah, why don't we do more as we come back from the break right after this. Did you know your credit card points and miles can lose value to inflation? Credit card companies often reduce the redemption value of your points and miles. Now, imagine a credit card with rewards that can grow in value. With the Gemini credit card, you can earn Bitcoin or one of over 50 other cryptos instantly with no annual fee. Every swipe at the store or gas pump earns you instant rewards deposited straight to your account.

22:37Plus, sign up now for a$200 Bitcoin bonus to kickstart your rewards. Visit Gemini.com slash card today. Check out the link in the description for more information on rates. Again, if you're looking to invest in Bitcoin but don't know where to start, the Gemini credit card makes it easy. The Gemini credit card is issued by WebBank. In order to qualify for the$200 crypto intro bonus, you must spend$3 ,000 in your first 90 days. Some exclusions apply to instant rewards in which rewards are deposited when the transaction posts. This content is not investment advice and trading crypto involves risk.

23:13The Gemini credit card cannot be used to make gambling related purchases.

23:20You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. We're now with you hosting the world together.

23:48And we're back here on Big Technology Podcast, Friday edition, breaking down all the big news that's happened. And Ranjan, at the top, you really mentioned that we are in a sort of traditionally quiet week of the year, but we have so much news coming out. And one of the biggest stories is the fact that Apple had to temporarily pull, and maybe for a longer term, pull Apple Watches because it's been accused of taking some of the blood oxygen reading technology from a company called Massimo and doing it without permission. and this has been such a big issue that it's now effectively been unable to sell these for a very important part of the holiday season.

24:32I'm curious how you read the case right now, whether you think Apple is in trouble long-term, not only in terms of its ability to sell these current models, but it's broader pitch that Apple Watch is really a health device. And where do you see it going from here? Yeah, for me, this was such an interesting story around the intersection of power and technology. And it's one of those topics we usually come up with it around antitrust, but even again, OpenAIA versus New York Times. This is one of those weird moments where in real life, I went to the Apple store on like December 20th or a few days before Christmas.

25:09And you could see the Apple watches were still on display, but you could not buy them. And to see, you know, like the legal nuance of technology battles that you read about in the news, but you actually go to the store and the impact is there. And I think this is a big thing, you know, around 2023 was a major transition in the way we look at power and technology and companies actually being held accountable. And when you get into the details, you know, basically Mossimo, a number of their engineers early on went over to Apple and then built their blood oxygen level monitoring technology in a very similar way.

25:48And the fact that, you know, a court actually demanded that Apple stop selling during the peak sales season is telling that, you know, we are in a different time right now. And I think it's also a good reminder that me included, I get frustrated with Apple, but I still generally think of them as the good company and Tim Cook, you know, a friendly face versus many other tech CEOs. I think this was a reminder. And if this story keeps going, that Apple can play pretty, pretty heavy handed as well. Absolutely. I mean, they've done this, they've effectively ruined the advertising businesses of companies like Twitter, which may have actually had an impact in the company's, you know, fire sale, not fire sale, but being forced to sell.