Sharp Tech with Ben Thompson - Generative AI and Copyright, When AI Hits the Music Business, The Social Media that Comes After Twitter

Episode Date: April 17, 2023

Copyright debates underlying text-to-image generation, the stakes for incumbents and upstarts, and what AI could mean for the music business (but not yet the movie business). Plus: A listener’s Twitter reality check yields a discussion of what might come next in a new era of social media.

Hello and welcome back to another episode of Sharp Tech. I'm Andrew Sharp and on the other line, Ben Thompson. Ben, how you doing? Terrible. Yeah, I figured you were probably going to say something along those lines. It's Sunday night. We're recording this in the aftermath of an outright disaster, a Janus, a Deticombo injury in game one of Heat Bucks.

And I, if this were a basketball podcast, I would not make you come on and record. It is a tech podcast. And so my offer to you is to sit here and say, let's provide a distraction from your basketball reality for the next 60 minutes or so. That's very generous of you, very thoughtful. I appreciate your consideration. I appreciate the certain outpouring of support from our listeners at this difficult time in. in my life. And yes, we must, we must push forward. Somehow, some way. The last time it was this dark, 10 days later, Yannis put together one of the most legendary finals performances I can remember I've

ever seen. I did bail on a pod. I was actually supposed to go on the Bill Simmons report after that game a few years ago when Yon got injured against the Hawks. And it was, see, you're very generous. You're being nice to be. Bill gets out of the podcast like says, I was supposed to have Ben Thompson on today, but he bailed. I'm like, come on. You can't. You can't. Yeah, do be dirty like that. Yeah, well, and it was a more sensitive time also. Janus had not won his title, and it seemed within reach. And so the apparent season-ending injury in the middle of that series was one of the more heartbreaking twists, I can remember.

But it all ended well. Hopefully, this will end well also. And on that note, let's shift gears and just escape into the welcome arms of tech news. and a programming note before we get rolling here. We are back to the regular schedule this week. Look at you. We're just a couple of Willis reads right here. Powering through Janice's injury, powering through baby Charles's appearance in the household.

Yeah, we're doing what we can here. You know, you're not going to stop us from podcasting. That's one thing about me and Ben Thompson. And we are back to normal this week. So that's very exciting, a little bit terrifying. I'm going to see how much I can handle here. You can still email questions to email at sharp tech.fm. Help Andrew put a show together.

That's the listener's job. Exactly. If you've got questions for Thursday, email at sharp tech. FM. And as for this Monday's show, I want to start with what I believe is our first ever clip on the podcast. It's a portion of last week's Strateree interview with Adobe's chief strategy officer, Scott Belski, and I've said in the past, I love the weekly Sterecheree interviews. And in this one, you guys were talking about regulation and

copyright enforcement in the era of AI and image generation. So I will cede the floor to Scott for the next two minutes here. Now, I'm not sure if I can extract a truly honest answer from you because you're, you know, you are an executive at Adobe. And so you have a motivation here. I will say, looking at it from the outside, I do, number one, it's not really clear to me why using photos is not fair use. It seems like a pretty textbooked example of transformative work. So, I mean, so there's a bit where Adobe leaning into this and all these are true. It kind of feels like a weird form of regulatory capture in a way where it's in Adobe's

interest to have actually a very narrow view of copyright. And that this is not as my fair use because we, hey, we have a huge stock photo library. and we have the capability of attesting that this is the case. I want to give you a good question that you can sort of talk about here. But what's, is this just your view? My guess would be that, look, that's what the big companies want, so we're going to give it to them. Is that sort of the way to put it?

It was interesting. I was having a debate earlier today, actually, with a friend around, does this go the route of kind of like the Ubers and Airbnbs of the world, where they, on a regional level, we're sort of pushing the laws? and then they just had to kind of fight region by region, and they were able to just sort of get the galvanization of their users to allow them to become, you know, in most areas, legal and accepted. Or is this more like the Napster and Limewire world,

where there was an era where it was just like a free-for-all? And then suddenly, you know, Congress, and I guess like on a federal level, there were just determinations about like what the outcome of this would be. And then it actually like all went away pretty quickly. and then you had iTunes and Spotify and whatever else. Here's a thing, Ben, like I understand today. It's kind of like it seems like it could go either way. But if you just play this out like six plus months from now,

where you can actually make unauthorized sequels to movies, you can say you can watch Succession and be like, you know what? I'm going to make my own script out of prompts with chat GPT5 or whatever, you know, we have by then. So I can't do that. I'm too busy making Balenciaga commercials. Exactly. And then I'm going to.

feed it into a text-to-video model. And I'm basically going to make an unauthorized, you know, episode using the exact actors, their likeness, their voices, and everything. I seriously believe this will all be achievable very soon. And it's just going to be a copyright dumpster fire. Like, what is going to happen? Like, there's no way they're going to be able to commercially monetize that content. There's just no way.

It's like you're using Spider-Man or using the Matrix. You're using whatever. So if you believe that that's an inevitability, which I do, then I think that there will have to be clean models that you can rely on commercially. And yeah, sure, you'll make fun memes and unauthorized things that will creep around the internet. But from a commercial perspective, I don't think it's just like a big company party line. Like I actually believe this is going to be a requirement for this all technology to work commercially. So the question and answer there prompts a million follow-up questions, but as a baseline, can you explain like the technical context and some of the legal tension that underlies the back and forth between you two? Yeah, well, I just want to make an observation, which is I love Scott, but he did not answer my question.

I feel like he deflected because there's two issues at play. And actually, I'm quite eager for your input on this because, we're getting into like some arcane, not arcane, but like difficult legal questions. So this is where you're going to be able to bring the training to bear. But to my mind, there's two distinct issues. I was asking about issue number one, and he deflected by answering issue number two. So from my perspective, issue number one is using copyrighted images in the training set, like when you're actually training these models.

from my perspective, that's a pretty definitive example of fair use. Okay? There's four factors that go into fair use. Number one, what's the transformative factor, the purpose and character of your use, the nature of the copyrighted work, number three, the amount and substantiality of the portion taken, and number four, the effect of the use upon the potential market. Now, for, go ahead, Mr. Lawyer is sharp. I'm sorry, I just opened my mouth and Ben stopped.

I will just say, you sound like the lawyer. I would not have been able to rattle off the four fair use factors. I've written about it before. So I'm reading from a previous daily update. Okay, good. Good. That makes me feel better. I'm just like, oh my God, Ben Thompson's smartest man on the planet.

Best lawyer in the world. Okay. Take me further on the fair used argument with the training corpus. It's clearly transformative, right? These images are going in and you're not getting, it's not a copy machine. That's not how this works, right? to say that this is a violation of copyright is saying like I read a bunch of news articles, then I wrote a daily update that was somehow gweening in which is how the brain works.

I'm pulling in information from all the stuff I've ever read and I'm writing something new and original. Oh, but because you once read the New York Times, you are infringing on their copyright. It's absurd. I just don't think it makes any sort of sense. Now, there is number four, the effect of the use upon the potential market, but the issue he's talking about in. my estimation is not, this doesn't need to be covered by changing the definition of fair use.

You can't recreate Spider-Man because Spider-Man is a copyrighted character, right? Like, like, like the copyright law cuts sort of both ways. And so my question here, and so the context here, let me back up. The broader context is can you use copyrighted material in models, okay? Mid-Journey is doing that. Stable diffusion is doing that. Open AI is doing that. Now, they won't say where they got their training data from, but it's pretty clear they're doing that, right? There's that story where if you use the prop Getty images, it will put like the Getty images like some, a facsimile, facsimile, facsimile, facsimile, facsimile, facsimile, thank you. A facsimile of the Getty image's watermark in there. And everyone took that to mean, oh,

that's clearly copyrighted for you. Well, no, it, it's associated the word Getty with that watermark there, right? Now, there are discussions to be had there for sure. I'm not out and out saying that that is a morally good thing. But as far as like the law is concerned, it seems pretty open and shut to me that it's sort of, it's sort of a black and white question. But Adobe's coming here saying,

oh, we don't use any copyrighted images in our training data. And by the way, this is what Mid Journey and OpenAI would be saying if they did only use public domain images and weren't treading in areas.

That's right. They would definitely tell you. Yeah, yeah. They would be very clear about where their bottle came from if it was all sort of clearly above the board, right? Right. So Adobe's saying we're only using open images and all the images from our stock library, right? So they have a huge stock photo library.

And so they're using that. And they're saying, and so it's in now a lot of people, the initial version of Firefly, they're like, it's not as good as mid journey or X, Y, Z. And, you know, I think that's, that's, that's, it's not. And, but Adobe, it feels like a little bit is sort of almost trying. And this is why I put in the question, it's kind of like, let's play the regulatory, like, with getting a regulatory capture sort of situation. Yeah.

It's in Adobe's interest to have the most restrictive definition of fair use that you can't even use these images in your training model because then here's Adobe with, oh, we have a bunch of images of our own already. And like, I guess if you want to be, work out well, you're going to have to partnership with Getty or whatever might be. The only way you can actually even make the model in the first place is to, that there be no sort of copyright involved. And this isn't a move in the interest of having the best generative AI models.

It's a move in the interest of can we leverage the law to give us a dominant position in the market? Raise the barrier to entry for new players in the marketplace. And my question, just as a baseline before we go any further, we talked about image generation AI and some of these training models back in the fall. but for anyone who missed that podcast or has forgotten some of the context over the last four or five months, as we focused mostly on large language models, when you say they pull in copyrighted images or in Adobe's case, it's all public domain and their stock library and use that for training. Essentially, what we're talking about is they suck up all those images and then associate them with different words. And so when you put in a prompt, it will generate what's ultimately an original image,

but pulling from all these other images across the internet as it sort of scrapes the internet for photos. Is that right? I want to stand up and applaud Andrew Stowe because you want to compliment me on my lawyer skills. And that's a pretty good overview of how it's happening. Yeah, it's exactly right. There are associating these images with sort of words. The way that image generation works is a bit different than large language. models where watch the animals are sort of predicting the next word that's sort of in the sentence.

Whereas this sort of the works image models, they start with noise and they like pull out an image from that noise. And the way they did that was by training it on images and said, you know, basically figure out how to get from pure noise to this text. And that's what I train the model. So, but yes, I mean, I should just stick with your debt. What am I trying to improve on them? There you go.

A great summary. Feel really good about it. So where would regulation leave some of the newer players? Oh, wait, wait. Before we get to regulation, I kind of want to pick on Scott a little bit more. Okay. Because I think there's... What he did was was pretty smart, actually,

because nobody's going to argue with the copyright argument about the Matrix 5 or succession, coming up with your own succession episodes and based on all the same actors. Like, that's a pretty open and shut case if we're talking copyright. Right. And that's already illegal, right? It's like we already have laws against that. I see what he was doing. And frankly, I'm impressed, but expands.

This is always the challenge with interviewing like big company executives. Like they're very good at what they do. And it's like, do I want to really dive into the weeds here? I thought my question was like detailed enough that to the discerting reader, it was pretty clear what was going on. We need to go back and forth on it. But I do think this is the answer. I'm not disputing that there is questions that are raised here.

I do think that what we might need to tighten up and make clear are people having rights to their, like, it's like the NCAA, right? What was a big thing there? Name and likeness, right? That's the big sort of shift in the NCAA where it was ruled all led by the Supreme Court that players have rights to their own image and likeness. And they have the right to monetize that. And the NCAA cannot limit that.

That's completely sort of transformed the nature of college athletics. I mean, I don't know if you want to use that for his excuse for your tar heels were started out the year ranked number one and even make the NCAA tournament. But whatever you want to, you know, you want to go to, that's sort of becoming a precedent. Yeah. And I think that's a precedent that makes a lot of sense in this area. Now, I do grant how you're going to enforce this. What is the line between an actual, you know, image of likeness and what is not is a fuzzy one? There's questions around like, is satire allowed?

Is it not allowed? The rules about libel and slander are very, you know, are hard to enforce. But that's where I think if there is regulatory response, that's where it ought to be. This idea that we're going to take a copyright regime that is already dramatically overreaching, that already dramatically favors large companies, and we're going to fix this from the sort of supply side to me is mistaken. Well, that was my question is, like, how screwed would some of the players be, like, mid-Journey or Open AI and Dolly if the government cracks down on what you're allowed to

scrape from the internet to generate these images? Like, have they made enough progress to this point that they wouldn't necessarily be totally out of the game? Or does it just totally change the paradigm going forward? and you have to be a Google or Adobe or whoever else has these massive photo libraries. I mean, maybe Getty images can get into it too. There's lots of examples of this, right? There's a lot of a whispering, pun intended, that Open AI's whisper model, which is its speech-to-text sort of model,

is basically trained on YouTube transcriptions, right? Because YouTube transcriptions are amazing, right? So there you go. You have all this voice. You have all this sort of like captions of like what they're saying generated by Google. And Open AI is rumor has it. that sort of went into that model, right? And so, yeah, the end result of a real serious crackdown of what can go into models

is absolutely to a dramatically large extent favoring very large companies with very large corpuses of their own content, right? So Google's example, they have rights to all that YouTube content. Believe me, you sign those rights away. They have all that. Facebook has all the stuff on Facebook, right? Elon Musk, oh, by the way, by the way, we talked about that that six-month delay letter. The institute that put that six-month delay together, guess who funds it?

Who? To a large extent. Elon Musk. Guess who is now starting an AI company and buying a bunch of GPUs. Elon Musk, guess who is, you know, guess who is excited about the opportunity to use all the Twitter content there? That letter was so transparently a ploy to stop people in the weed and catch up. And I think that this is the wake up call here. The wake up call is not that copyrighted material is being transformed.

Like if we have any remnant of any sort of freedom in terms of what we could do, to me, again, I want your thoughts, but to me it feels like a pretty open and shut case. But the possible output here of some sort of severe crackdown would be decimating. It would be very, very, very bad. It would basically say no one else can enter this space. One of the great excitements and what's so been invigorating about the last year was the AI's been looming for a long time. Like the idea that this is coming is a big thing. But it's been, it was sort of assumed, and I say this including myself, that AI would only be the big companies.

And it's actually going to deepen the problem of just a few companies basically controlling everything. Right. And what these poys are and what's been so excited about the last year is the turn out that wasn't the case. You could actually, this is why stable diffusion was arguably like even bigger than chat GPT from a sort of big picture perspective. Stable diffusion was such a so exciting because suddenly out of nowhere you had an open source model that was competitive and by and large things people most agree is well surpassed Dali. Maybe not as good as mid journey, but you know, it's not necessarily going to be the big guys. It's going to be someone who can come along. and we're back to actually having competition in the space.

And we don't know what the industry is going to be going forward. And the big companies might be in trouble. And what you have right here is you have people fighting tooth and nail to make sure the big companies get to stay big. This is a important fight. What's your view of the fair use argument? Am I like taking this too literally? No, I think the fair use argument, I lean toward where you are on all of these. this in large part because when you look at the images that these like that a mid journey is

generating, it looks like an original image. It doesn't look like 10 or 15 different photographs piece together. So you're like literally just taking someone else's work and making a collage of someone else's work and calling it original. Like you look at the output and it does all feel original, which matters. And then in a legal context, I just have a hard time believing that these artists who, if you're talking about like a photographer, for instance, maybe their photograph contributes to like 2% of the ultimate AI product.

No, like 0.000000, 0000,000, 0,0,02%, maybe. And then it's completely untraceable if it is. Right. And so in that context, it's pretty difficult to see how we're going to enact this, like, elaborate regime of laws that would be really difficult to enforce. Number one would obviously favor the incumbents like we've discussed. And I'm not sure whose rights are really being protected in that context, because I don't know that the photographer who is contributing 0.000% of the work is really big. That was easier to say than the zeros. Yeah.

Well, and so the other side of the argument, I would add is that if you look at the music business, and we're going to talk about the music business in a little bit here, the way this has worked is if you sample a couple chords of a song and it's a 50-year-old song or whatever, you pay royalties to that artist. And that's how artists get paid. And it's just how the copyright system has worked in entertainment for years. years and years and years. And so I don't know that I buy that that would be applicable to photographers and copyright holders of different images. But that would be the argument I would make if I were

in court, you know, try to make a case on their behalf. Yeah. And that's kind of, I think that's an there's a lawsuit against stable diffusion. I think they make a similar argument. But they sort of say in that lawsuit that it's like a collage, right, which is basically the music sample argument. The problem with an argument is it's not true, right? Right. That's not how it works. Like, just not at all how these image generation, they, they are pulled out of raw noise.

It is not the pasting together of little bits of clips here and there. And, and so, yeah, and the music sample thing, I mean, we won't get into the music sample business. One can question is that actually a good place that we're in where taking, like, would society, would culturally better off if you could, if there was, if that was also fair use. But I will even grant, not you, but the interlocutor that we may have here, that argument and say, it's not even that. That's sort of not what it is. Now, I do think there is a question one of the clear areas of pushback is artists with very distinct styles, right? And then

they go to mid-jurney and they say generate X, Y, Z, and they put the name of the artist, because then it will figure out that it, that's the style to apply to it because like the artist. but this is where I go back to the NCAA idea. That's a name and likeness sort of issue, right? You're literally, like, maybe if I were to institute laws, that's where it would be. Like, if you're putting in the name, that someone has a right to their name. The trigger would be the name and the prompt suddenly puts it in a different category legally and fair use-wise. Is that right?

Right. Well, I just think it's a, like, that's a, it's a distinct question of can you use someone's name and likeness, right? And that name and like this can avoid to imitating their voice. It would apply to this succession example where I'm making my alternative episode of succession. It would apply to I'm blatantly stealing someone's style. Now, again, I grant difficult to police, all these sorts of things. But I think the long-term outcomes and the questions of just how is this even a problem of policing what goes in the models themselves, I'm concerned about. The other point I would push back on is this Napster sort of lime wire bit.

I don't think, I think that that overstates what happened. Now, Napster, LimeWire, like, I was the first person in my floor to just, like, when I started in college, we were like, we got music out, you know, off of FTPs. Like, like, FTP servers. You would find them on sketchy places on the internet. You'd go in and then you would, I mean, it was. No, I believe me. I am familiar with the protocol, but it was always just a bridge too far as far as nerdiness for me. And I wanted to learn how to do it because like the hardcore music pirates were doing that.

But I left that to the experts like you. But then Napster came along and it was like it was transformative because it was so easy to use. Anyone could do it. Right. So what stopped Natser? What stopped piracy? Part of it for sure was like the RAA filing lawsuits, right? You can figure out what the ISP was and then sewing up your door and saying, you know, you're getting served a lawsuit.

There was a deterrent factor, a stick. But that was not, that's not the end of the story. What also happened and what the music industry did, and they were drag kicking and screaming doing this by Daniel Eck, but they did do it, is they embraced the internet. What is the internet? The internet is sort of zero marginal costs as far as getting new content. It was everything is available.

And so they made everything available for $10 a month, right? And the allure of the $10 a month is why go to all the $1,000, trouble, we'll just give it all to you, right? Your, like, your time and also your risk from a lawsuit, but just your time. Like, even on Napster. That's honestly what it was. They were very effective. I mean, the lawsuit threat never felt real to any individual out there who's downloading

like 10 or 15 albums. But what they did was they went to Google. They went to Yahoo. They shut down Kazah, Napster, all these different, like, sort of choke point access points where people were downloading music and finding these like zip files of albums and they made that much, much harder to do. And then on the other hand, you had this really convenient option of iTunes and then eventually Spotify where unless you wanted to work really, really hard, it just made more sense to

spend your money on like actual licensed music rather than, than, you know, navigate the netherworld of the internet. Right. And like what they made it an obvious choice, right? Like for $10 a month, you get access to all the music in the world. You don't have to hunt it down. You don't have to download it. It's just, it's just available, right?

And the best solutions to problems are always ones that lean into that are aligned with that are propelled by the fundamental forces that are causing the problem in the first place, right? The problem for the music industry is that it turned out they weren't selling music. They were selling pieces of plastic. Right? It's like newspapers were not selling the news. They were selling printed paper that showed up on your doorstep. And the internet made that worthless, right?

Because it was even more convenient to get it over the internet. And so the solution was not to try to stop the internet. The solution was to leverage the internet to deliver something better, something that was more compelling. And the music industry is, is, the best example of doing that. Now, what does this mean in the context of these models? Some aspect is to be determined. Like, one of the things that I did talk about with Scott is, you know, the internet is very good at attribution. You can track everything, right? That's people complain about that in the context of ads that everything's being tracked and you want to find out,

did you see an ad and then did you convert and connect it together? Like with it, but it, companies are quite good at it when there's money on the line. And so can we come up with some sort of attribution model where something might work. I don't think we will ever or ought ever get to a place where you get money just because your stuff's in a model. Like that doesn't seem workable to me.

It doesn't seem good in terms of, it doesn't seem to make sense in terms of copyright law. And also it does, I think it would be bad for culture, bad for the internet, bad for competition, all sort of in the long run. There's lots of factors going against it. But if we can get to sort of like names and likeness sort of infrastructure, where

yeah, you could write something in the voice of Ben Thompson, but, you know, the accredited. I mean, YouTube did this as well, right? How YouTube ended up solving the copyrighted music being in video's problem was at first they would just delete the video, right? And they'd be like a strike. Then they would like mute the video.

What they do today is they just like, hey, if you upload a video to YouTube that has copyrighted music in it, guess what? The RA is getting a cut, right? Like that, that's, or, you know, the MPA or whatever it is, right? It just, it just takes care. that. Yeah. Yeah. Because there are certain things that do run a file of copyright and you'll see that it's been removed. So I wonder how they draw that line. No, YouTube's way ahead of everyone else on this line. Like most other platforms, they'll get a copy. They'll get a notice and they'll take it down. Right. And you still see that on Twitter and stuff like that. But YouTube now they just, they just like, look, if videos up and it has copyrighted material, the copyright owner is getting paid. Uh, like, like they're going to collect our ad money and we'll give you guys a cut and everybody can go home happy. Yeah. And that's actually a good thing because now, You have more avenues to make money with your copyrighted material, right?

And so I think we can figure out solutions here. That doesn't mean there's not going to be large upheaval in the artist's world or the design. That's obviously going to happen. But we need to push forward for solutions that allow for continuing innovation that makes sense with the internet and don't make sense with AI and don't try to fight it, right? Let's use AI to find these bits, right? Instead of like we're going to bring down the force of law to try to keep things as they were and make sure that Adobe can win the space. I really like the idea of using AI to police all the copywritten AI content and all the offenders out there. That's obviously the solution.

The solution to the internet making music widely available through piracy is to leverage the internet to make music widely available. at a convenient low monthly cost, right? That was, that was the answer. It's going to be the same thing here. And we're going to see this in lots of places. Like, it's going to be AI versus AI. Like that's, you're going to, like, everyone's like, oh, we're going to get all this AI generated junk. Well, you have AI is policing that, right?

Humans are just going to be hanging out for the next 150 years and then 150 years from now. I'm sure the AI will take over. Lots of time to touch grass while the AI does our work. I feel like Charles will have a good life. And then after that, I don't know, grandkids, you're probably on your own. So technically speaking. I do want to talk about the music, but I also have to ask this. Do you buy the idea that we're six months away from being able to generate an entire episode of succession with, you know, written prompts or something?

I mean, there is an aspect where it's always dodgy to predict any sort of non-total speed and transformation when it comes to AI, given the way the last year is gone. Yeah. There is also a bit where chat GPT was transformative and woke. everyone up, but GPT4 was already done when GPT3, when the chat GPT was released, right? And so it feels faster than it actually was. GPD3 came out in 2020. So there is an aspect where there, there is a maybe a bit of a false sense of speed as far as the underlying models go. That said, I think in many respects, the speed of mid-journey is actually even more impressive. Like, they're on V5 now, and you go back just six months. And it's incredible, the speed of which

that's happened. I've been very surprised at the speed of which video generation has come along. I wrote, you know, it's about a year ago I wrote about what Dali came out. And we had have done a test episode then. And we weren't, the Sharp Tech was not officially launched. But my interpretation of that is like, look, this is going to be transformative, like for metaverse VR sort of things because the wall we've hidden in game generation and virtual worlds is just cost so much to produce content. And that's just a fundamental limitation in what in the more immersive, the more content you need.

And like once computers can generate it, that's going to be completely transformational. And at the time I wrote that, I was obviously thinking about virtual worlds and and, and video, sort of a step on the way to real time generation of content. I didn't expect us to be having video to the extent that we do today. Yeah, well, it's tough. Like, I want to throw up my hands and never doubt anyone when they throw out an AI timeline after the last six to nine months. So it's not to play the skeptic, but I will say that the people in tech, and now I'm in tech. So among all of my people in tech.

With very good AI explanations. Yeah. So the one thing that I kind of push back on is the idea that the text to video is some transformational thing. and we're going to be watching AI-generated movies. Like, maybe eventually we'll get there, but we are nowhere near there now. You look at the videos, even that Valenciaga clip that you had be watched. Like, it's impressive for being AI, but it doesn't, it's not like we're in the uncanny valley. Like, you can tell that this is auto-generated and is kind of a fun novelty, but I'm not sure that it's

actually transformational yet. Maybe six months from now it'll be completely different. But right now, it seems like that. much or do about nothing. I disagree with that timeline. Yeah, as well. Six months is pretty aggressive to basically recreate a very high budget, you know, live actor sort of thing.

Right now we're in the like fun Twitter video phase of this. And then, you know, who knows how long it'll take before we get genuine entertainment. Now, music is the complete opposite end of the spectrum because the music right now is scary accurate and scary realistic. Some of the samples that are floating around, there have been a number of Kanye West samples that people have generated. And Jonathan emailed us and said, over the weekend, a TikTok went viral of some AI generated music that was made in the style of Drake and the weekend. There were discussions around whether or not it was good, but what interested me was the discussion of copyright infringement. When talking about violations,

the best place to start is how the generated music will be used. But even, even in the case of adding the song from the TikTok to the spot to Spotify, what is the move that artists or record companies can make to sue? And I think you're right at the top. Like, there is an existing body of law that makes it unlawful to advertise or sell using another's name, voice, signature, photograph or likeness. Now I'm the one reading the law without that person's prior consent. Andrew, I thought you had it memorized. That was so important. It's too bad. No. But there's an entire body of publicity law and it's very state by state, but it's pretty clearly illegal to do what some of these people are doing. Now, granted, it's unclear whether they're actually selling it overtly. I have actual question for you. Is it illegal to even produce or is it illegal to commercialize? I think it's illegal to commercialize. And it could also be an invasion of privacy to produce and not. commercialize it depending on the legal framework of the state you're in.

Yeah. And that's sort of a similar thing with like that Facebook Lama model, right? It is no one can use that in a commercial product. It's not licensed for that. And also the Stanford sort of Alpaca team, you trained like did the reinforcement training using open AI prompts, which is also against the terms of open AI's license. So no company can touch that with a 10 foot pole. That doesn't stop people from experimenting with it.

it, you know, playing around with it, X, Y, Z. And I don't think that's a terrible place to sort of end up, right? I mean, like, the part of the idea of commercialization is you're attaching the imprimatur of a company or, or an individual sort of to that. That's a necessity for collecting payment, right? Like, there has to be sort of entity to get that payment back anyway. And that should be, that, that should be illegal. And it already is illegal, right? You can't do that. And, you can't do that. And And to me, this is, yeah, I think this is the, we already have the right broad framework in place to deal with this. If someone wants to upload a, you mentioned Drake in the weekend, you can't upload that to Spotify and then make money off it, right? Just like you can't upload copyrighted music to YouTube.

Now, this is a step beyond because it's not copyrighted per se. It's not like an original piece of content that was written by Drake or The Weekend or Kanye West, whoever might be. But I think strengthening or tightening or having a framework for this publicity bit and names and likeness sort of bit, that seems like a pretty clear solution that preserves the space for innovation and also like provides protections and compensation for the people that are getting quote unquote ripped off. Yeah. Well, and my guess would be that as record companies try to enforce this and apparently a number of record companies are already. really unhappy with the state of the landscape as it is record companies unhappy yeah who's at 11 who could guess um well and i think again they will have to go to these choke point platforms in the same way they did with google and you know mega upload and some of the other ones that they

knocked out 10 or 15 years ago that people were using for piracy mega upload yeah yeah throwback um they'll have to go to like a spotify and say look you can't host this music it's illegal because I'm not sure you're ever going to be able to come up with like a framework to prosecute like Ghost Writer 977 who generated this Drake and the weekend fake clip over the weekend. So I am very curious to see how it all evolves. And I look at the entertainment business writ large. And I do think in music in particular, I've talked about this with a mutual friend of ours, Spike Eskin. the internet has already made it much harder for artists to become like real superstars. Like there's a lower barrier to entry, but there's also a lower ceiling on how popular most new artists can get because the market has just been flooded in the Spotify era.

And I wonder whether we're going to see that trend accelerate in the era of AI where like the people like Kanye West, like, Drake, like Taylor Swift, they are going to be able to license their vocals for another 30 years of AI generated music. And then the new original artists are just going to be sort of fighting and clawing for a shred of attention. And we're not going to have like the mega stars. Like maybe Kanye West and Taylor Swift are like the last mega stars in the music business. Because going forward, it's just going to be really difficult to generate any of that. Yeah. I mean, that's like to your and Spike's point, that's, this case now, right?

Yeah. And this is an internet I mean, you already waited out. I mean, you're on a roll today. Like low barrier of entry means it's easier to get started, but that means there's drastically more competition. It's not just music. This is what happened to writing. Right? It happened to writing first. You know, and I can now come on and I

can compete with the New York Times for the time that you're going to spend reading something. And that's tough if you're if you're an incumbent, but it's not clear that that should be our sort of priorities, you know, going forward. You did mention this bit, license your vocals, right? If there's sort of like the official place where it's available, there's going to be some new opportunities here. But big picture, the solution is

to switch to authentication, to instead of trying to stamp out everything that's not real, you ascertain what is real. And that's a knowable space. So Taylor Swift, give me a with if you're a real person here. Well, no, so you think about it in the context of like, like, how do you know that a article is written by me? You go to shratercary.com and like that, that is authenticating that it's, that it's me. Now, is that, can that be spoof? Can it be?

Of course it can. Like, but, but this idea that you need to have authentication that shows that it's real is a, because again, the, the space of generated content. content is infinite. It's zero marginal cost. Again, AI is not quite there to zero marginal cost, but it's obviously on its way. Once we get local models, with local models, it effectively is zero marginal cost. So there's going to be an infinite amount of garbage. You can't police infinite, right? It's a, you would you put zero in the equation, it all falls apart. You can not police, but you can ascertain the world of knowable content of a verifiable content is bound, right?

And so you can know that and authenticate it. And so this is a good thing Adobe is doing. They have an authenticity initiative. Like anything that's generated by their AI tool is marked that it's by that AI tool. And you can mark for real stuff too, right? So you get a song. You should be able to validate. Is this a real song?

Again, this is where we lean into the internet. We lean into AI to help us solve this problem to authenticate and something is real. this is Taylor Swift. This is Kanye West. This is Drake. This is Ben Thompson. You know, just slide myself in with those three right there, right? Yeah.

Rushmore. Well, no, and that's, it's quite clear that that's the next step that the industry writ large is going to have to take because right now you've got already Twitter is like a wash with all these different AI vocals and everything else. This is why the Elon Musk verification batch thing was such a disaster. It wasn't just dumb. It was actually going in the wrong direction as far as what's necessary. I actually think that I was on Twitter about the verification badge thing for years.

The problem with the verification badge is it was too limited. And it was granted at Twitter's whim. They should have had a program for a decade where if you authenticate who you are, then you get a badge. Right. And if you change it. So like input your license, prove that you are who you say you are and suddenly you've got a blue check. And that if you want to be. everyone more reliable.

If you want to be anonymous, that's fine. We're not going to kill anonymity. And nor should we want to. There's a lot of value. And there's people that are positions that can't be out there that, you know, it's like the whole same argument you have for whistleblowers and sources and stuff like that. There's stuff that that the public ought to know that that, like this idea that everything has to be. I disagree with that.

But what you can do is instead of trying to shut all that off, you can actually validate people who choose to be validated. and that can be a sort of a resource. This is an example of like Twitter's just been horribly managed for so long. Twitter should be like the namespace of the internet. Like you already have it, you go on TV and they have their at Twitter handle, right? It's incredible. They have this incredible asset and they could have built something meaningful on that. And they never did.

And then Musk comes in and takes the exact opposite direction where I can sign up and I can say, I'm Nintendo and I can do Super Mario giving you a finger. It's just, it's the whole. issue with Musk is basically doubling down on bad strategies. So it's like hard to pick apart. Like it's not just that what he's doing doesn't make sense. It's like doubling down on what Twitter is doing that hasn't made sense for a decade when there was a clear other way to go. Yes.

And it's the opposite of where the entire internet is going to end up at the end of all this. Hopefully, uh, well, the other, authentication is the way. The other interesting thing just to note to put in here and this will probably annoy a few people is this is exactly, If you go back to listen to what I said before, I've been consistent on this.

This is exactly where blockchains are interesting. The whole idea of a blockchain is it is a spread widely on the internet yet unique, yet uniqueness and scarcity. And so the idea that you could have the ascertainment of content on an entity that is not controlled by any company that no one is in charge of is super compelling. And I think there is a bit where there is a bit of. of a symbiosis between this future of AI just generating stuff like crazy and to have the capability of verification at scale that's not controlled by anyone company. Now, that's going to take a long time to get there. There's a lot of scalability work.

In the meantime, I think there's a lot of room for validation by established companies. Again, this is what Twitter should have done. Twitter should be the place where you attach your name. It links to your profile on Twitter. It's art. This fits in with Twitter as protocol that connects to everything else. But in the long run, you could also see there actually being the value for this more of a feature as opposed to an industry, but real value to have something that is knowable and scarce and ascertainably so. All I ask is that we don't call it blockchain, if that's the direction we're going with authenticating the entire internet.

Blockchain is the FTP of the modern era where my normie alarms go. off and I'm just like, I don't understand this shit. I'll let the nerds figure it out. So we'll have to make it more accessible. Or blockchain is the FTP of the modern era. It's like the most niche t-shirt that we just think of potential. I like it. Oh, boy.

But speaking of Twitter, I want to close as we continue the distraction for you. This is therapy for Ben in the wake of Bucks Heat. Jorge says, I think that perhaps you two are both too close and invested into Twitter to take a step back and recognize that Twitter has really lost its relevance. No one in my circle of friends, family, or colleagues use Twitter at all. At this point, I only hear about Twitter because of the media. My friends and I mostly communicate via text or through Instagram Reels. This is what brings us to a place where substack notes and artifact can dare to launch. There is a large unmet need for people to absorb content and have a conversation around it, which is what I think these platforms.

are attempting to do. Now, Ben, I got Jorge's email. I read this email last week and I was like, man, you know, he's absolutely right. Like, Twitter has lost its relevance. No one in my life relies on Twitter on a daily basis who's not in the media or, you know, one step removed from the media. But at the same time, that has always been true of Twitter. Like, it's never really had its relevance among like normal people.

And the issue with Twitter is it's this like paradoxical thing where no one is on it except the media. So it has this like outsize relevance and has this agenda setting function that I think persist to this day, despite what Jorge is saying there. I think there's absolutely an opportunity for new social media for people to use. But at the same time, Twitter still has its hooks in, at least from what I can see. that's exactly the takeaway, right? I was going to, before you, you jumped in to make the point, I was going to say, Jorge, you just said that they only hear about Twitter because of the media. That's the exact problem, right?

It's like Twitter is the social network for the media. There was an old sort of phrase back in the day that the New York Times publishes on Monday, CBS News broadcasts on Monday night, and local newspapers published on Tuesday morning. And speaking to the New York Times sort of. of traditional agenda setting role in the U.S. sort of media ecosystem. What's happened today is there's a Twitter controversy on Sunday night, then the New York Times publishes on Monday and on down the line. And that's the, that is the case.

It remains the case. If anything, it's got even more of the case as people, to the extent they do leave Twitter. And all that's left is media people online, argue with themselves and like aligning themselves. When you had a geographic diversity and it would take days, if not weeks, if not ever, to find out what an opinion calumness in, you know, Madison, Wisconsin Center, Portland, Oregon, or whatever, that you have so much more diversity. What you have when it comes to Twitter is everything online and this, this massive sort of coordination function where anyone who disagrees is sort of beat down.

There was a bit last week with, I'm just to go back to the NBA where there is a big controversy because the rookie of the year vote was not unanimous for Powell Ben-Bankero. It was because Walker Kessler got a vote. He got it from the Utah jazz sort of beat writer. There's a few problems there. Number one, Kessler was better than Ben-Kero, but that's neither here nor there. Wow. Hot-take at the end.

Number two, the idea that we should be outraged because Powell Bancero would not get a unanimous MVP vote when he's fine. Yeah, he shouldn't be a unanimous MVP. This is a LeBron James sort of story here, right? It would be super clear. But I'll set aside the NBA hot take. for now. It used to be that different people would vote lots of places. Yes, the beatwriters would tend to favor of their own players, but it all sort of evened out in the end. And that was fine. On Twitter, you are vilified and the character of the day because you diverted from the group. And what's the message that comes out of this controversy?

It's every single voter for an NBA award now knows it's just, if you just go with the conventional wisdom, you won't get attacked on Twitter. How is that good? How is that a benefit? Now, Take that toy example for a sport and a rookie of the year vote that doesn't really matter and apply it to literally everything in life where the only people that actually raise a ruckus are the most extreme, are the activists on either side. Everyone in the middle is like, if I raise my hand here, if I object, I'm going to get attacked. And getting attacked on Twitter is hard. It's not fun at all. Like, it's hard to explain if it's not happened to you. but it's it's like it's a brutal experience like it's when i got super attacked and vilified

for the whole title two stuff i was it took me like it should be for years like it's it's it's hard to we are not as humans meant to have tens of thousands of people saying go die and and i'm going to come kill you and or you need to be fired even it doesn't have to be that dramatic but it can still be pretty abusive when yeah it's just completely out of proportion like that's what happened to that jazz writer. It's like the entire internet freaks out. It's like, whoa, whoa, whoa. This is a rookie of the year vote.

It's not the end of the world. Yeah, and it doesn't need to be like tens of thousands of people. I just don't tweet anymore because it's not worth it, right? Like you have people coming in. It's probably hurt my business, but like, who wants to view abuse for fun? It's just not, it's not like, but anyhow, we're now we've, we're, we're a Twitter problematic thing. It's a good question, though, because I wonder how long these dynamics

will persist because I do think Jorge is right that Twitter, number one, was never that relevant among the mainstream. It has lost relevance from whatever it was five to seven years ago, like among my circle of friends and peers. And so I wonder how long everyone is going to remain super conscious of the conversation on Twitter and potential for attacks and criticism and everything else because it's not just the media. I mean, like, C-suite. in corporations also have to be really conscious of all this, and they end up minimizing risk wherever possible to appease Twitter and avoid like crazy controversies over, you know, X, Y, or Z. And so everything sort of becomes more vanilla and safe in that environment.

And I do wonder how long that will last, because like Jorge says, most people aren't on Twitter and don't care about these stupid controversies. So at some point, we have to be a able to break away from this model of making decisions, but I don't know when we'll get there. Well, this is the bit where Jorge is right and I am too invested. There is a part of me that wants Twitter to fail and go out of business because I think it's bad for society. I think this, and I think it's bad because of what it does to the media. I think the media is important. It does shape how people think. I think people have a perception of the U.S. of being far worse than it actually is in reality that is driven by Twitter. And this absolutely applies to both sides.

Like we, you know, the most sort of galaxy brain take is that it's actually good because we fight online and the normal life is mostly fine. So like we use our energy online. Perfectly pleasant. Yeah. Yeah. Yeah. So, no, so I've said that a couple of times. Like, I do think Twitter, I think social media by and large has gotten an overly bad rap.

I think Facebook in particular. I think part of it is, again, the media uses Twitter. They don't use Facebook. and all the angst and problems that everyone can see with social media has been put on Facebook when it's actually been Twitter all along. Like that's like and again, that's a subjective take, but that is a bias that I will admit to. But that said, Jorge makes an excellent point here, which I think I'm going to steal and this is my give you credit, which is what makes substack notes an artifact interesting. By and large, I don't think there's going to be an entity that replaces Twitter, right?

you don't do a one-for-one replacement. But there is, to his point, there's lots and lots of people that have nowhere to go, that have nowhere to post that, like, we want to comment on a news story, right? And if you can comment on a news story, an artifact and you can get upvoted because it was super insightful,

like that's the ABCs of sort of social network, right? You could get status, you can get that feedback by offering something smart and there's an opportunity and you could, there's reputation that accrues over time. There's actually a motivation to go in and do it now and be like one of the best commenters on Artifact, right? And yeah, if you're not on, yeah, sure, most people who are on Twitter will just drop their takes on Twitter,

but there's lots of people that aren't on Twitter. And I think it's a very excellent insight by Jorge that there's a huge addressable market of people that realize Twitter's terrible. And it's going to be really interesting to see how it evolves. You wrote about this last week. You wrote about Substack Notes. You wrote about Artifact and Blue Sky, I believe. and I just wonder over the next five years what the social media landscape is going to look like

because reading your post, I started to think, like we've sort of been in the network television era of social media where you've got Facebook, you've got Twitter, you've got Instagram, you've got like a couple major players, and those have been the only options for people. But as some of these specialized options emerge, maybe there will be fewer people

watching or not watching, I'm sticking with a cable analogy. Maybe there will be fewer people using these apps, but there are going to be maybe more people using social media generally as newer options and newer, more specialized options hit the marketplace because I do think right now what we have isn't meeting most people's needs. So I wonder how tech will address that over the next five years. Yeah, I mean, the thing to remember is humans do, it is kind of interesting that we're having this like social media like reemergence and explosion at the same time. All the other news is about humans talking to machines. Like humans still also like to talk other humans, right?

Right. And yeah, and so the opportunity space is basically infinite of like building companies here and getting ever more specialized. Like what is Reddit? Reddit is just like basically can we have one broad platform that that, that, serves the needs of every niche imaginable and ones you probably didn't want to know existed, right? Like the, and so,

yeah, I think it's exciting. I think, like, it's very interesting to this develop. I think it would be healthier overall for there to be more distinct, uh, you know,

heterogeneous, heterogeneous homogenous, whatever, one of the two sort of, uh, communities that, and, you know,

this was one of those bits. I, you know, I referenced this last week. I wrote social networking 2.0, I think in 2020. And at the time, I was a little head.

to write it because I'm like, I feel like I'm living in the future as far as social media is concerned, where I had multiple networks. Yeah. Yeah. And like this is where things are going. But there's always a question, well, I do live abroad. My business is in the U.S. You know, I have, there's lots of exogenous factors that are driving me to have, like, I have, for

years, I've had a folder. I have a folder in my doc that has all the chat clients that I have in one folder. Now I just get one notification and I can click it and go to it. Because that's the reality of my life, right? And is that because I'm someone living abroad that has to deal with different networks in different countries and all these sorts of things? Or is this the way things are going? And actually, I've been thinking about that article for a long time. One of the reasons I was hesitant to write it is I didn't want to read too much into my experience.

And maybe for regular folks, it's just going to be Facebook and Twitter forever. But it sort of was, it felt sort of inescapable then. I think it's turned out to be the case that, you know, the most mistaken thing Mark Zuckerberg ever said. And this goes back to the authenticity bit is like, oh, you know, that's over. Everyone needs to be their whole self online. And that's why I like the real name policy, all those sort of things. No, we're not, that doesn't apply to the offline world either, right? When I am at my in-laws house versus when I'm at home versus when I'm at cigar night,

I'm basically three different fends. And that is not being dishonest. It's three different aspects of my personality and my persona that is appropriate, right? You don't want to be the same person everywhere. It actually is healthy to have different spaces for different things. That's why we have the touch grass movement, right? There's a bit you need to go out and just hang out, right? And be someone different than you are at work or that you are in other sort of social spaces.

And I do find it exciting and encouraging that we're seeing more and more products cognizant of this reality and sort of serving that. And I think it's a good thing. I think it's healthy. I think having everyone in the same spot and say everyone has to be their whole self all the time, that's not a good place to be. You get in situations. It's part of the Twitter problem. If you have to be your whole self all the time, then attacking any aspect of that self feels like an attack on you, right? And so someone attacks a Janus or if someone attacks like my political view, that's an attack on me.

It's like, no, it's actually not. And also, I don't want to transmit to everyone I've ever met professionally and personally. personally, and I've had several different careers at this point. And so social media that's all centralized is so stressful to me that I don't even really engage anymore because it's just like, let me talk basketball with my basketball group chat. I'm mad at myself. I didn't even mention WhatsApp in the new social media landscape and the emergence of group chats, which we've emphasized over and over again.

There's also Discord, which is having a moment here, a fun little national security. adventure, but I'm very excited to see how all of it evolves. And I think you're right that, like, ultimately it will end up in a healthier place than we've been over the last 10 years or so. Yeah, it's a very low bar. Yeah, it's a very low bar. And I think this goes, we mentioned this in like the pause thing, right? It's like, pause where?

Pausing in 2023 because things are so great. Like, I do think we'll look back at the 2010s as a real sort of low point, but a low point that was sort of inherent in any major transition. Like, again, you can pine for the old days, but there is no going back. What you don't want to do is stop right in the middle of the wrenching transition. You want to get through to something better. And I think the AI is exciting in that regard because it potentially points to something better.

I think there's going to be a lot of peaks and valleys with the AI stuff that sort of going forward. I think social media, though, is an example. V1 kind of sucked, right? And it started out great. And then you realize having everyone online and having everything you ever posted. Like, one of the big mistakes I think all social media is made.

And Elon Musk, you want to make a good change, make this change. Why is every tweet saved? All tweets should be deleted after a week, right? Like, like, no, you think about it, though. You should at least have the option to toggle that. No, it should happen automatically. Save people from themselves. I'm fine with that.

Twitter. The whole idea of Twitter is a real-time social. network. It's what's happening right now. So why do we care about a tweet from three years ago, right? And like, anyhow. No, it'd be a lot more fun. The entire internet should delete every seven days. Maybe not the entire internet. But certainly all of social media is more fun when you don't have to be accountable for your dumb jokes 10 days ago. So I'm with you 100% on that front. And it's a long list of things that Elon should be doing that he isn't doing. I had no idea. Yeah, you still want to take accountability for your bucks are not.

contenders right in pencil take, which now is annoyingly true if Yannis is injured. I'm not going to rub it in. It's a sensitive stretch right here. Everybody say a prayer for Janus's tailbone. Hopefully he comes back healthy and the bucks can get back on the right track. But a real Willis Reed performance from you here fighting through heartache to deliver another great podcast to the Sharp Tech family. And Sharp Tech family who is now larger by one. Yeah, that's right. Good performance by you. Good, good job by you. Okay. Well, on that back patting note, we will come back later in the week. Email at sharptech.com. If you've got questions, we would love to hear from you. Thursday's show is going to be a more traditional mailbag. We'll just sort of bounce around. So hit us with whatever you've got. We'll double back to some questions we missed on the vacation episode that was upstaged for a live human birth. And Ben, on that note, I do sincerely wish you well processing the Janus news.

Don't wish me well. Wish Janus well. That's right. I wish Janus well more than anything. And I look forward to coming back soon. Yonis, you can have my tailbone.