The Taylor Swift Deepfake Problem, Tech All Time Highs + Layoffs, Quitting Amazon Prime

0:00Deep fake nudes of Taylor Swift are circulating online, leading to call from all sides to ban the practice. So what does it say about the future of generative AI? Let's also talk about the weirdest tech recovery, where stock market all-time highs are being met with mass layoffs, and why quitting Amazon Prime might not be a bad idea. All that and more coming up right after this. Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format. We're back here for another week of news, big week of news, consequential week of news. We're streaming live on LinkedIn once again, so we might take some questions from the audience.

0:39And for those listening on the audio feed, we appreciate you. Ranjan, Roy, Ranjan Roy joins us as always. Ranjan, welcome. We're so familiar, no introduction needed, Alex. So before we get started, I just want to share something very quick. We got such a nice review on Apple Podcasts from someone whose username is Kay Parwani. Kay Parwani writes, this podcast is my tech sanctuary in a world filled with digital noise. No tech babble, no smoke and mirrors, just straightforward, insightful discussions that cut through the BS. We asked for the reviews because people who are evaluating the podcast, they tend to use them as a proxy for how many people listened to the show and whether their executives should be on.

1:23And we've had a couple of really close calls. We had some great guests last year, of course, Brian Chesky from Airbnb, among others. And we want to definitely deliver the top tier executives to you this year and the way that you can help us do that, not for our egos, but for the guest recruitment is to do a five star review on Apple Podcasts or Spotify. Now that I have gotten that out of the way, I just want to make that plea and say thank you to such a nice review. Let's get to the news of the week. And it's been a bit of a troubling story, Ron John, where there's been AI porn of Taylor Swift circulating through the Internet.

1:58basically people using image generators to make naked images of Taylor. And now it's become this huge story. Was it inevitable that we were going to get here with image generation? It was inevitable. And as I read a couple of places, it's not a big deal until it happens to Taylor Swift. And I think this is where the rubber meets the road. This is where generative AI. We've known it's going to have to be addressed at some point. So what better platform than Elon Musk's X slash Twitter? I'm still having trouble calling it X. And who better than Taylor Swift? And I think regulation around generative AI, at some point, there's going to be some kind of platform standards that are needed.

2:43There's going to be some kind of regulation that's needed. Most image generators have protections and censorship against this, but clearly lots of people are able to do this but but yeah we've all been talking about this and we've known it was coming so it's it's here yeah and you could totally you could jailbreak these uh these image generators and even if there are protections there are ways to always like i guess inject prompts and get around some of the protections which is going to be a bigger and bigger issue and you're right the fact that it's taylor swift has now brought everybody's attention and the New York Post as it tends to do, just so artfully put it, it's no big deal until it happens to Taylor Swift.

3:22Now that you have your attention, Houston, we have an AI generated porn problem, an oozing societal store many folks were seemingly happy to ignore until it ran afoul of America's sweet sweetheart. This week, deepfake images of the wholesome singer, digitally altered photos portraying her in a compromising sexual position circulated on X to Swifties discuss. And now, this is important, Swifties are wondering how there are no regulations against someone creating fake porn using the likeness of real people and distributing it for the whole world to behold. So of course, I try to consume a good deal of media on all sides of an issue.

4:02And what I found interesting here was that the interests of the New York Post and folks on the other side of the aisle or the other side of the political spectrum are lining up saying that this is a real problem and needs to be dealt with. Here's Casey Newton and platformers saying, generative AI has many positive creative uses and I still believe in its potential to do good. But looking back over the past year, it's clear that any benefits we have seen today come at a high cost. And unless those in power take action, and soon the number of victims who pay that cost is only going to increase. That is notable to me, that we're seeing this as an issue that is starting to drive consensus.

4:42And of course we've seen consensus for regulating big tech across the aisle and you know it never amounted to anything but when this stuff starts happening taylor swift do you imagine that it's going to be a pretty important legislative issue fairly soon already i mean it is ironic that we've already seen examples of deepfakes around i mean biden and trump for a long time around in other political races So the future of American democracy and global democracy did not create this call to arms in the way that Taylor Swift has in the last week, but it's still good that it's happening. I think this is the year that the technology has become as democratized as possible and effective as possible.

5:27Just anyone who's used like mid-journey version three versus where it is now in version six has seen how quickly the technology has just exploded and just how effective it is. But yeah, I think going into this election year, this is going to be such an issue. We are going to see certainly some very, very viral deepfakes that are circulated, that are confusing. Or like in this case, I don't think, I mean, I'm guessing vast, vast majority of people who see the Taylor Swift deepfakes are not thinking it's actually Taylor Swift. It's more a very clear coordinated harassment campaign. But overall, it's now and it's happening.

6:09But I get one question. You had said big tech regulation. Do you think this is an issue of platforms? Do you think that this is an issue for the tools like MidJourney, Dolly, Stable Diffusion, whoever else? or who gets regulated or who gets controlled or the user? It's a great question because there will be a big movement that will say it's only a problem for the folks that distribute it and the platforms that allow the distribution, which is a great legal argument. But I think ultimately when it comes down to who's going to take heat for it, it is going to be the tools that enable this to be possible.

6:43Like Twitter existed, you know, despite all of its issues before this happened. right this is a new technology that's enabling it to to take place and so therefore the new technology will be the thing that gets the spotlight and yes of course it will play a role in politics but i also think that you know maybe like an image of joe biden a fake image of joe biden or a fake image of donald trump um i think a lot of people can can like you mentioned pick out the fact that that's fake the same way that they picked out the fact that taylor swift the Taylor Swift image was fake. But when it comes to creating these nudes, that's the real problem, fake or not.

7:20And to me, one of the most interesting things about this is, and the post is right, it has opened our eyes to the fact that this has been a problem for many others. There was a post that Casey Newton picked out that was talking about how on 4chan, they're finding women that are testifying in court and then deep faking them without their clothes on. And this is sort of like a way that like they you know sort of revel in the technology and that's a huge problem and i think that like my eyes are open now in particular about how widespread this already is and how and where it's going to be and maybe it's not going to rise to the point where it's going to throw an election but certainly to the point where like everyday people are going to have to deal with the fact that nude images of them are circulating deep faked or not that's a problem no that that's that's a good point.

8:12I think that distinguishing between disinformation and harassment, and when this is clear that this is pure harassment, not disinformation, basically, and the example you made, which is terrifying, is pure harassment or intimidation. Yeah, that's actually, I mean, a really good distinction that I hadn't thought of as much and probably up to this point has not in the center of the discussion and needs to be more so um but yeah this is this is a tough problem and it's and it's also tough because again i'm like the first to say that this technology is incredible and is going to have so many positive and interesting and amazing use cases and and i love talking about it using it so yeah this is going to be an interesting year on this Rajan, this is such a new technology also.

9:04And so it's coming into an environment where there are broader awareness of the downsides of technology. And maybe in an instance like this can alter its trajectory. Can something like the Taylor Swift nudes actually set back progress in generative AI? Or is that something that is still so far fetched given the direction of technology? that's okay one thing i would take issue and i think any time in discussions around regulation of technology is the phrasing set back progress because i think you know is this progress taylor swift nudes deep fake nudes being circulated like i think a lot of the times the language around regulation especially from the technology side of the industry especially the ones benefiting or profiting is that everything is progress and you will be setting it back.

9:59I think it's, I guess in a way, it's a good thing because you can imagine seven to 10 years ago, no one would have even really thought like this might even happen and everyone would just look the other way and just let things go. Whereas I think we're in a environment now, as you said, where at least everyone is a lot more conscious of risk in technology and at least willing to think through how to work with it and work to prevent it. But I don't know. Okay, well, let me attach this specific definition of progress then. Progress in terms of, okay, we can talk about the broad definition of progress.

10:35I'm talking specifically about building better technology in this vein. Will it set that back? Progress in that way. better technology meaning like if mid-journey version 8 has more safeguards implemented to prevent this i gotta ask i gotta get the real answer here though like yes of course but i'm talking about more sophisticated like we have been on an innovation curve with this generative ai movement where we've seen this stuff improve its capabilities so yet i'm not saying we shouldn't build for safety i'm just trying to say like can this be a screeching halt moment or are we still going to see folks continuing to build this technology i don't think this will be a screeching halt moment i think this is going to be it's good that these disc it's not good that this is happening but it's good that it's brought it to the forefront and the discussion because again maybe there is some law around i mean there already are definitely laws around harassment and if you generate it and distribute it not the platform the user that can be you know maybe there's actual legal recourse that can be taken and is like very structured and created in a way that you know the legal system recognizes so i think i think it's it's going to not bring it to a screeching halt and will be built will be built in a better way yeah i think all the companies He's thinking about bringing out the next generation of this stuff.

12:09We're watching this now and being like, are we sure? Are we sure? Because this stuff is moving quickly. And we spoke a little bit last week about how video generation is going to be the next thing. And then like literally two days later, this happened this week. Google released a video generator called Lumiere, which is an AI video generator. This is from Ars Technica. It does a great job of creating videos of cute animals in ridiculous scenarios, such as using roller skates, driving a car, playing a piano. And one of the things that I thought was so interesting about is how it could alter. You like you like you can alter existing video.

12:48So you say, you know, make that person's white shirt black or make that person's, you know, gray hair blue. And it will be able to do that just with AI, which is unbelievable. So we talked about how this video generation stuff is really going to, I mean, if you thought, if you think bad images of like, let's say Taylor Swift are causing a problem, just wait till the videos come. If I may rant about Google for just a moment, they did not actually release Lumiere. And we'll definitely get into this in a segment later. they released a paper and as they often do, you know, saying that this capability will be possible.

13:24I think it's really reflective of how Google has approached product development and generative AI overall, just a small rant that'll come back in this episode. But yeah, I agree. Video is going to be a major area, but video is still far away. Video, if you've used runway ML, if you've used the video generation capabilities that people can actually use at this point, there's still a ways to go. So I think photos are still going to be where the, where were the like kind of behavioral norms, legal norms, all these things are kind of set. It would be interesting to get someone who's like working on training some of these AI generated videos to come on the show.

14:04I'm just kind of curious. They must see all sorts of in the, in the development process, They might see, must see all sorts of like, I don't know, fun house hell that these video generators are creating. No, I've worked on, I've trained a stable diffusion model and like on custom images and stuff. And yeah, it's wild and weird. And especially this was about like a year ago. So it was, I mean, it's progressed very quickly since then. But, you know, that's like the 12 fingers, eight hands, 10 arms, that kind of stuff was very common before. which is why I think we do have time. There is a window here, but I also agree.

14:44I think the point you made, someone working at MidJourney is a product manager, someone working at OpenAI on Dolly. This Taylor Swift situation will make them at least think, which is at least a start. It's a lead start versus I feel there was probably a three to four year window where people probably did not, like almost willfully ignored potential use cases of what they're doing. It is kind of sad that it takes something like this to like make them pay attention. Like you would imagine that the idea of people being AI generated naked after appearing in court would be horrifying enough, but I guess it takes Taylor Swift.

15:28Larger statement on society and culture. and look we're here spending the first 15 minutes of the show talking about it when it was taylor swift so i mean where is she going to be sitting at the afc championship that's all we know in the box that's for sure you got the chiefs or the ravens i'm gonna go ravens i also think if taylor sat in the cold i could you imagine how powerful that would be could you imagine a statement Nothing would bother her. Nothing would bother her. The cold of the people, Taylor. I mean, with bodyguards all around, I understand this is like a case in point that being too close, Jason Kelsey shirt off among the people has a different wherewithal, but yeah.

16:13I mean, he's a 300-pound offensive lineman. What about you? What about you? Ravens or Chiefs? I think my Super Bowl is going to be Chiefs-Lions. Let's see how it goes. Yeah, I'm Lions this year. I'm rooting. strongly for the Lions, Detroit, you know, it's a, it's a feel good story. I mean, the coach is amazing. Bite the kneecap off guy, you know, potentially taking Detroit to the Super Bowl. We're going to get up and on the way up, we're going to bite a kneecap off. All right. And we're going to stand up and then it's going to take two more shots to knock us down. All right. And on the way up, we're going to take your other kneecap.

16:44Anyway, we should, before we get too deep into the sports stuff, let's, let's talk quickly about the business side of generative AI, because there was a very interesting story about Anthropik. Like you and I are always trying to determine like how much of a business this is. Will margins come down, et cetera, et cetera. This is a story from the information Anthropik's gross margin flags, long-term AI profit questions. Okay. It's a bit of a jumbly headline, but basically it means that they're not making as much money off of it as other technology. And this is interesting. So it says new data show that profit margins for startups may end up lower than for existing enterprise software firms.

17:25Let's talk about AI startups, right? After paying the cost of customer support and service to power its AI, Anthropics gross margin, so gross profit as a percentage of revenue, was between 50 and 55 percent in December. And this is according to two people with direct knowledge of the figures. that's far lower than the average gross margin of 77 % for cloud software stocks. So we are looking at the fact that there are so much money involved in running these models that the margins are fairly low. And it's interesting because, or not fairly low, 55 is still pretty good, but not as high as traditional software.

18:02And it does make me wonder a little bit about like, what is the business future of this? I mean, think about all the billions, trillions may be invested into AI. And if you can't make a lot of money off of it, does that eventually come back to roost? I thought this was such a good story and a good point. Again, every listener, remember, average gross margin for cloud software stocks, 77%. Anthropic reported 50 to 55. And 50 to 55 % gross margin, you can see that in a good retailer, someone actually making clothes. So you're getting into a very, very different space than traditional software. And it really got me thinking around how does generative AI scale?

18:48Because we know that generative AI is a more resource intensive process. There's definitely hope that over time, again, as this technology advances and improves, the cost per query will come down the way things work will come down and you know this will all be the right models will be used for the right use cases because again do i need the full power of gpt for or for anthropic the latest model of clod to do simple text queries or you know generate like 10 email subject lines probably not so maybe that's one area where you can start to see some more cost efficiency but But overall, is this going to look different than traditional software?

19:33And I think it raises a really interesting question around that, that is this a new hybrid where the cogs behind technology are no longer just kind of infinite scale software like we've seen in most other industries where it almost looks more like Tesla or an Airbnb or someone else calling themselves a tech company, but actually having a lot of underlying costs. maybe it looks more like that and what does that say about where this is all heading then i mean oh the valuations are yeah it says the valuations are insane actually i think even more so because these are not even being valued like cloud companies these are being valued like covid era the whatever peloton work from home crypto everything so so i think you're in trouble when in peloton covet territory yeah when you're in peloton covet territory and raise the flag you know uh but yeah i think i think it says very clearly that valuations are out of control in this space because again it's it's one thing to say this is a pure traditional software play that's going to dominate traditional software but if you start seeing that the underlying economics are different than traditional software already, then those like 50x future sales or whatever they're trading at right now definitely look a bit rich.

21:00Yeah. And speaking of COVID era valuations and market madness, can I briefly tease Wednesday's guest? Please, please. Okay, I'm going to do it. Vlad Tenev, the CEO of Robinhood, is coming on. So we're going to have such a fun conversation. I can't wait. So folks, stay tuned for Vlad coming on Wednesday. It should be about 6 a.m. Eastern that I drop the episode. Just for listeners' context, I do not know ahead of time these guests. And Alex amazingly weaves into every conversation. Somehow I tee him up perfectly to introduce the guest that's coming. Thank you so much. So all right. Speaking of economy.

21:41Very interesting moment right now. And I really, I think we're going to just have to figure out what the heck is happening here because it's so confusing to me. We're seeing all-time highs in the stock market and not just all-time highs in the stock market, but we're seeing all-time highs in tech, which is powering the stock market. And yet, it looks like 2024 might be a bigger year for tech layoffs than 2023 was. I would understand if so many of these cuts and let's just talk about them quickly we saw that Microsoft is going to cut 1900 people from Activision the Salesforce just announced a layoff we've talked about the ones at Google we've talked about ones continuing at Amazon that we're already outpacing January 2023 which is hard to believe and Quartz says that we have 24 ,000 layoffs just this week in the tech industry.

22:39Why in the world are the tech companies performing so well and cutting so deeply? So there's a really good article in The Atlantic from Andy Lowry around layoffs being contagious or the idea of copycat layoffs that I thought was pretty interesting. Yeah. Can you explain that? Can you dig into that a bit? Yeah. So when a company lays off and their peers are not laying off, it gives the impression that your company is potentially in trouble or something's going wrong. When your competitors are laying off people and it becomes an industry trend, that is exactly the moment that not only if you've been thinking about laying off anyone, you do it.

23:24It almost looks bad if you're not laying off people because at that point, like right now in the market, it started with year of efficiency and everything else last year, layoffs not only have become normalized, it's almost a show of strength by tech CEOs and most of these firms. So I think it's clear that every leader has the incentive to layoff right now and in some ways is, again, incentivized to do it. And I think we can definitely discuss, should they be doing it? Who should be taking responsibility for the decisions being made that were made in the past few years. But it's clear this is going to continue.

24:03I mean, how do you do it? Like, okay, so you said maybe we should. Let's do it. Like, let's talk about it. How can you lay off so many people when the valuations are at an all-time high? Because you hired too many people during the last couple of years. But isn't the market telling you you didn't if they're putting your stock at an all-time high? Well, no, but the market is still just trying. The market just cares about your revenue, sales growth, EBITDA, profit, all these things. But this is what I'm saying. So if the stock market isn't punishing you for your costs, what do you - It could be better.

24:35Well, actually, hold on. But the other major point, and this is one thing that I know, and I will be a bit controversial here, is I don't have the greatest sympathy for large tech firms and employees at many of them because over the last number of years, I heard firsthand endless stories about people very, very generously compensated, barely doing anything at a lot of these firms that they did become bloated. They did become, you know, just like kind of, you know, rest and vest or just places to relax and hang out, do a little, do side projects and get insane compensation packages relative to most people in the world and even most people in tech.

25:21But I think where it gets interesting is at what point does that hiring bloat negatively affect the actual work being done, the product development pipeline? And if you can imagine those additional layers of bureaucracy and complexity probably slow things down. Like Google, going back to what we had said earlier, Google has been a disaster in terms of product rollout, especially in the generative AI space. I mean, BARD is a mess, how that works within their existing products. Meanwhile, Microsoft has just been on fire. Is that because there are too many people? I mean, that's what Mark Zuckerberg and Meta openly said.

26:02I was about to say, you're sounding Elon Musk-ish. I know, but this is something I will agree with Elon Musk to an extent, But I will agree more with Mark Zuckerberg in the sense of, you know, culturally it changed the company. And then again, remember, it was in their year of efficiency that they launched threads in like how many was like three months. There was a great piece. I remember reading about the entire launch process. Like they got to work. They cut things down and really focused and started executing very well. They revitalize their entire ad business, again, while downsizing and becoming more efficient.

26:39So I think there's something to be said where hiring bloat, even if the market is allowing for it, still actually will have negative long-term implications on the company, both culturally, in terms of actual product efficiency, development, all these things. So let's accept your argument. Why didn't this happen in 2023? Why did they wait till 2024? Was it that they were so slow that they couldn't figure out who they wanted to cut for a year? this is like we're gonna see more cuts this year than last no but well one we did see many cuts last year i think that uh quartz had the graphic that cuts this year are outpacing on average for january every month from last year but still are just only slightly higher so there was plenty of layoffs last year and i do think it'll continue but but as you said these companies have been running slow.

27:32You know, these companies, there is a lot of bureaucracy and work within this. And it's not like you're at a early stage startup where hiring and firing are the fastest things that happen. You know, these things do happen slowly. But going back to the contagious copycat idea, now there's all the appetite in the world. Again, should Sundar and other leaders take more responsibility for the mismanagement or the overhiring? I think that's a reasonable question. But is it the right thing to do for the company at this moment? I think for a lot of these companies, it is. Let's talk briefly because it involves like one of the themes that we talk about here.

28:16Let's talk briefly about the media layoffs, right? 115 people cut at the LA Times, I think 8 % out at Business Insider, Sports Illustrated effectively demolished. We know there've been layoffs at Condé Nast. I mean, I could keep talking forever about the media companies that have laid off my old employer, BuzzFeed News, No More Newsroom, although that happened mid last year. I think there's a misnomer here that people are saying that this is because of generative AI. I just don't think it's like people have gone to generative AI to get their news. It seems to me that it's just like broader business mismanagement.

28:51Can we agree on that one? I, well, one definite distinction for anyone listening is that But media layoffs, I have a very different, at least, empathy for versus kind of like big tech layoffs. Because, again, it's not like people's lives were super comfortable beforehand versus a lot of the big tech employees. I think within the media industry, it's been tough for a long time and it remains tough. But I agree. I don't think it has anything to do with generative AI. Maybe, like, if anything, it'll be used as an excuse by management to push through layoffs. But I think this just still goes back to the business model problem that's plagued media for the last 15 years.