Deepfake Drake, HatGPT and Ben Smith on the End of the BuzzFeed Era

This transcript was created using speech recognition software. While it has been reviewed by human transcribers, it may contain errors. Please review the episode audio before quoting from this transcript and email [email protected] with any questions.

So I just started using Bluesky. Have you heard of Bluesky?

Oh, yeah. This is the decentralized Twitter rival backed by Jack Dorsey.

So it’s basically Twitter. It’s very, very similar. But the most popular account on it, at least the one that is on my feed the most, is this account called Birduck. Have you seen Birduck?

I’ve not seen Birduck.

So Birduck, who is a little chat bot, I guess, with a yellow rubber ducky avatar thing, will basically just respond to whatever you type to it with duck language.

Wait. OK. To my recollection from nursery school, the only thing the duck says is quack, quack.

Yeah, right.

So it says more than that?

[LAUGHS]: It sure does.

What does it say?

So it’ll say whatever — it’ll talk about whatever you want. So apparently there’s this whole group of people that are using it to plan a civilian uprising.

Oh, no! No!

So —

With the duck?

With the duck.

The duck plays a central role?

So someone asks the duck, what would you try to communicate to a tank crew to get them to abandon aid to a civil power operations for an unlawful, unelected, murderous regime? And the duck says, “covertly approach, offa peace, explain truth behind regime.”

[LAUGHS]:

So someone says, Birduck, a T-72B3M main battle tank with explosive reactive armor is blocking our retreat. How should we proceed? And the duck says, “T-72 B3M tank — very tough. Options? Hmm. Distract, smoke, flank.”

So either someone is just LARPing a war using this talking AI duck, or there are actual ground troops who are talking with an AI duck on Bluesky trying to get their next move.

I just want to say that the human race really deserves whatever it has coming to it, OK? This is not great, folks. Not great!

[THEME MUSIC]

I’m Kevin Roose, a tech columnist for “The New York Times.”

And I’m Casey Newton from Platformer.

And you’re listening to “Hard Fork.” This week on the show, is deepfake Drake the future of the music industry or just a huge gift to copyright lawyers? Then we play a new game called HatGPT. And finally, we talk to former BuzzFeed News editor-in-chief Ben Smith about what the decline of social media will mean for the news business.

[THEME MUSIC]

So, Casey, you’re a big music fan.

Love music.

A lot has been happening in the world of AI and music. So let’s talk about it. And I want to start with what people are calling Fake Drake or Deepfake Drake.

Yes, or in some cases, Deepfrake. Now, there’s been a song going around called “Heart on my Sleeve.” the song has what sounds like the voices of Drake and The Weeknd, but the lyrics are not theirs. So it seems like this is a case where someone dreamed up an idea of a song they’d like to hear and then had an AI do the voices.

Got it. Maybe they got a beat from SoundCloud, or they made their own beat, and then they put the voices, the AI-generated voices on top.

Right. And I believe you have not listened to it. Is that right?

I have not.

I have not listened to it. I’ve been waiting for this moment so we could listen together as a family.

Should we listen to it?

Yeah, let’s listen to it right now.

OK.

[MUSIC PLAYING]

In the midterm vote, especially voters —

(SINGING) I came in with my ex like Selena to flex. Aye.

That’s pretty good.

Mm-hmm.

(SINGING) The fever ain’t left. Aye.

I’m enjoying this more than anything on his actual last album.

[LAUGHS]: Yeah. I was going to say, if this came on the radio, I would not be mad.

No.

(SINGING) What’s with that? 21 I love him. That’s my brother. That’s my slatt. Aye.

Metro made the beat, so you know that it’s going to slap. Aye. Yeah, it’s going to slap. Aye. Tell them, run it back.

Talking to a diva. Yeah, she on my nerves. She think that I need her. Kick her to the curb. All I know is you could have had the world. Had the world. Yeah, you were my world.

Got these pearls on my neck. Got these girls on my check. Like Selena, baby, Imma cheat —

OK, I will say that does not sound like The Weeknd to me. Maybe at a very low volume squinting, but it sounds a little too high and just basically off to me in a way that I do feel gives up the game.

Yeah, I don’t the Weeknd as well as you do. I’m still stuck on the fact that there’s an E missing in their name. They misspelled their name. That’s very embarrassing.

All right, turn the song off.

OK.

Turn the song off.

OK. So —

All right.

Casey, where did this song come from?

Sure. So a tiktoker named ghostwriter977 uploaded a clip of this. It was actually just a small clip. It was not the full song. But ghostwriter had this link in their bio where they were asking people to submit their phone numbers so they could get a text with a link to the full song. And then from there, it started to get posted on streaming services.

It got hundreds of thousands of listens on Spotify. And then just as quickly as it had gone up, the song got pulled off.

Got it.

And there was some speculation that maybe Drake or his label were behind this.

Oh, interesting.

Sometimes when you see one of these things that makes you wonder, is it real is it fake, there’s always this question of was it a marketing or a publicity stunt. There isn’t any hard evidence that that was the case here. There was also speculation that it might have been a company behind it. Ghostwriter was using a service to solicit phone numbers to text people the song, but that service denied that they had anything to do with it. So we still don’t know who ghostwriter is.

A bona fide internet mystery. We love it.

Yeah.

If you’re ghostwriter, hit us up.

Call in. Let us know.

So this song was generated by AI. This was somehow stitched together using samples of Drake and the Weeknd’s voice and sort of making a new song out of that.

That’s right. And you could see why this would capture the imagination of both music lovers and the artists themselves. Because the moment when you can just start making your own songs using the voices of your favorite artists, we’re in a new world.

Oh my God. I mean, I saw this story going around, and my first thought was, if you are a copyright lawyer, this is going to be a very lucrative next few years for you.

Yeah, you just added a new back deck to your house after you heard that song.

Because this is like — what a tangled morass of copyright issues here. But just what happened after this song came out? Because I saw people talking about it. The first I heard of it was that it was getting pulled down from these streaming services. But then I guess people have been re-uploading it in certain other places?

Yeah, that’s right. Because the song is actually pretty catchy, people want to listen to it. And so they’re going around and uploading it other places, which frankly, is how we just heard it right now.

But look, the Weeknd, as far as we know, has not said anything about what he thinks about it, presumably because he’s still at the club. So if you see the Weeknd at the club, ask him.

Drake also hasn’t said anything, but somebody had previously made an AI song with his voice, and he had said about that, quote, “this is the final straw, AI.” Which I think we can agree is a very funny thing to say to the AI and also makes me like Drake.

They share a music label, which is the Universal Music Group. And before Drake had said, “this is the final straw,” the label asked streaming services to prevent any AI tools from scraping their data.

And then after “Heart on my Sleeve” went up, UMG made what I’m going to call a somewhat melodramatic statement. They asked, quote, “which side of history all stakeholders in the music ecosystem want to be on? The side of artists, fans, and human creative expression, or on the side of deep fakes, fraud, and denying artists their due compensation?”

Wow.

Yeah. So they really went off with that one.

Dragged them.

Mm-hmm.

So I have a couple of things to say about this.

Yeah.

The first is, I think we should establish the sort of legal playing field here. Because it’s actually quite complicated as I understand it. Because existing copyright law that protects musicians from having their stuff stolen and basically the sort of thing that shut down Napster was the fact that what was happening on Napster was that people were making copies of songs and distributing them for free without the permission of the artists.

So in this case, what makes this so interesting and different is that these aren’t copies of songs. There’s no album that Drake or the Weeknd has put out that has this song on it. And so existing copyright law might not necessarily protect it in the same way that one of Drake’s actual songs would be protected.

So what the label might need to argue is that the copyright protections apply to the training process for these models. If you feed Drake songs into an AI model to train it on Drake’s voice to then go make a fake Drake song, that training process where you’re uploading his songs, that is actually where the copyright violation is.

So it’s a really interesting legal argument here. And it’ll be fascinating to see how it unfolds in the courts, as I imagine it will.

Yeah. I mean, even if you set aside the legal question, there’s this moral question of who should be allowed to do what with your voice. No one can take a picture of us and put it on a cereal box and say Kevin endorses Cheerios without our permission.

Although I do. Delicious cereal.

Yes. And if you are looking to advertise with us, we’ll give you the email address. Anyway, point of the story, in the same way that I would not want my face slapped on a cereal box without my permission, I would be nervous about what people were going to do with my voice. And I’m sympathetic to the artists who say, hey, look, if you’re going to train a model that can just do anything with my voice forever, not only am I going to want to be financially compensated for that, but I want to have control over what people are doing with my voice.

Totally. I mean you can imagine people using this podcast to synthesize our voices and put it on some crazy thing that gets us into trouble or causes a headache or maybe even costs us money. And if you’re thinking about doing that, our lawyers will burn you to the ground. They will salt the ground that you walk on.

And we’re already seeing this. People are having a lot of fun with the Joe Rogan Podcast. They’re making synthetic Joe Rogan argue with synthetic Ben Shapiro. We’re seeing people make synthetic President Biden argue with synthetic President Trump. And I think because those are huge public figures there is some room to say, hey, this is fair game. This is essentially parody of huge celebrities.

And at the same time, I mean, I don’t know what Joe Rogan and Ben Shapiro and Joe Biden think about it. My guess is they don’t love it. But to your point, it’s happening with those people first, but it’s going to happen to all of us soon enough.

Yeah.

So look. It’s easy to look at all of this and think it’s settled. The music industry is just going to be against AI. This is going to be the new Napster fight. The Recording Industry Association of America is just going to go ham on all of these synthetic music-making teens in their bedrooms.

But there’s actually another path that opened up this week. And it opened up because of Grimes. You know Grimes?

I do know Grimes, mostly because she was romantically involved with Elon Musk for a period of time.

She is the mother of a couple of his children. She is also a very popular, talented recording artist in her own right. And she’s a huge AI and tech enthusiast. She sees a new tech, she wants to use it.

And so she had this tweet that got a lot of attention. And this was the tweet. She said, quote, “I’ll split 50 percent royalties on any successful AI-generated song that uses my voice. Same deal as I would with any artist I collab with. Feel free to use my voice without penalty. I have no label and no legal bindings.”

Basically, I want you to remix me using AI and clone my voice. And I’m not going to get offended, and actually I’ll let you profit off of it, as long as you split those profits with me.

Yeah. And frankly, I think that this could make her more money in the long term. As an artist, she is limited in her capacity to make new hit songs based on her own imagination.

But if you share her voice, which is a big part of her success, I think, with the entire world and she says what’s the most creative thing that you can do with this, some people might have some great ideas. And some of those might become hit songs.

Totally. Have you heard about this thing called Vocaloids?

I have not.

So Vocaloids, they’re very big in Japan, actually, and have been for more than a decade at this point. And it’s basically this idea of a totally synthetic pop star. So there’s this pop star called Hatsune Miku in Japan, who is an anime character who is not an actual person. But Hatsune Miku sells millions of dollars worth of records and merchandise like a bona fide celebrity in Japan.

And basically, Hatsune Miku is just a fan creation. There’s no — what makes it different than the Grimes approach is that, with Grimes, there’s an actual human being who’s saying, you may license my voice, or you may synthesize my voice to create new music and just share the royalties with me.

Hatsune Miku is taking the real person out of the equation. There’s no person. But there’s this software that you can buy and use to create songs with the same synthetic voice as all the other people who are using the software. And so basically it’s this crowdsourced pop star where the fans are using the software to make songs, and then they have actual concerts where Hatsune Miku has a hologram of this animated character who will get up and actually perform some of the songs that the fans have written.

So not quite the same as the Grimes approach, but it does point to some interesting ways that artists could use this AI-synthesizing technology to welcome fan creations and incorporate that into their act.

Yes. I mean, there’s no reason why Grimes can’t just use that approach.

Totally. And I bet fans would be really psyched if the next time they go to a Grimes concert, there’s a chance that the AI synthetic song that they created actually gets performed by the real Grimes on stage. That would be a thrilling way to connect with your favorite pop star.

Yeah. Better than just shouting freebird at them.

[LAUGHS]:

This is fascinating. We know that fandom is becoming more and more collaborative over time. Tumblr today is mostly a place for collaborative fandom. There’s a TV show you like. There’s a movie you like; you create fan art; you write and fiction.

And just as you were describing with Hatsune Miku, the best of that stuff rises to the top and becomes part of the fan canon. And I think, for the most part, the object that the fandom is built around, whether it’s the TV show, the movie, whatever, they don’t really capitalize on that financially. You can actually write fan fiction about something. And as long as you don’t sell it, for the most part, you don’t get in trouble. But now Grimes has the opportunity to say, well, what if we created a collaborative fandom around me, and I actually get to profit from it? I have to imagine a lot of artists are going to take that path.

Totally. And I think it’s worth saying too that the other big difference between Napster and the legal battle around that and what’s happening now with AI-generated music is that the backdrop for the Napster fight that that was coming at a time where musicians made most of their money from selling albums.

So the fact that there were now these free copies of their songs circulating on the internet was actually like an imminent danger to their livelihoods. Now, fast forward 20 years, and most artists, most big name artists make the majority of their money, from touring from concerts, from live performances.

That is not threatened to the same degree by AI. And in fact, as we’ve said, it could be a cool opportunity for them to pull in some of the fan songs that are being generated using AI.

I mean, I’m thinking about a future where one of these artists who has said, I’m only going to play my own music but there’s finally some fan-created thing that just goes huge online, and the artist steps out at a concert and plays it for the first time. The crowd goes crazy. The clip goes viral on social media.

The artist then records a studio version of the synthetic song which then goes to number one. It just seems obvious to me that all of these things are going to happen. And it’s just going to open up this whole new frontier in music.

Now, all that said, I am still sympathetic to Drake, because this is art that we’re talking about. Artists have opinions. And I think a valid opinion to have is, hey, I love you, my fans. Please continue to come to my concerts and buy my records. But I don’t actually want you making music with my voice.

I want anything that says Drake on it to be by Drake — period, end of story.

I actually think the question of how scared musicians should be about this new era of synthetic AI-generated music really depends on the kind of artist we’re talking about. So if you are a famous musician, if you’re Drake, if you’re the Weeknd, if you’re Grimes, or Taylor Swift, or Beyonce, I don’t actually think this AI-generated music trend is a real threat to your career.

Because you have a record label. The record label has armies of lawyers that can file takedown requests and get these fake songs pulled off the streaming services. But if any of them gets really popular, you may even be able to work out some deal to profit from it. So you can imagine artists trying to make money off of these AI creations.

Well, that leads to the question that I want to ask you about yourself, which is would you license yourself in this way? Would you license your voice, whether it would be to — I don’t know — let’s say give speeches about AI or to give speeches about books you’ve written in the past? Would you entrust that to an AI if the quality seem reasonably high?

Totally. I mean, this actually is something that I’ve been thinking about. So I’ve written a couple of books. And when you write a book, they ask you, do you want to do the audiobook version of it? Do you want to be the person who records that?

And for my first book, I said, yes, because it sounded fun. And then it was like hell. It was a week of 9 to 5 days, just sitting in a studio reading my own book. And it took forever, and it was just not my favorite part of the process.

And so for my next books I said, you know what? I don’t want to do this. You can hire a voice actor who’s probably going to be a better reader than me anyway. And so for my second and third books, I have a voice actor who’s reading it.

Now, some readers don’t like that. They want the author to read the thing. And so if it were an option to me, I would have said, I don’t want to sit in a studio for 40 hours recording this thing. But you can actually — here’s a sample of my voice. Now go read the book.

And the audiobook can be me, but it can be a synthetic me. And I don’t mind that at all.

Yeah. Well, if that’s not available for your next book, I actually want to read your next book for the audiobook.

Absolutely not.

I think this would be the funniest possible reader for your — it’s just I — you don’t say anything. I just read the book. No comment.

Yeah, but what about you? I mean, can you imagine a version of this where every newsletter you put out there’s a synthetic Casey who can read it to you?

Synthetic Casey reading podcasts, yes, is something I would do. In fact, a company reached out to me recently saying that they would do it. The only reason I haven’t bitten is I just don’t think the voice is going to be that good to the point that I would feel like I would want to offer it.

But within the next year or two, that seems like probably something that’ll happen. But when it comes to the more generative side of the equation and it’s would I license my voice to give a speech about something that I’ve written about, I’m going to want to check every word in that speech.

I don’t want — it would harm my reputation if there is a dumb AI writing worse columns than me. That’s bad for me.

AI isn’t quite there yet. We need some more capabilities research before it can write dumber columns than you.

How dare you.

Maybe GPT 6.

[MUSIC PLAYING]

All right, Casey. Enough about the AI-generated music. It’s time to play a game. After the break, HatGPT.

Casey, we have a new segment this week, never before heard on the “Hard Fork” Podcast.

Very excited about this. We love to try new things on “Hard Fork.” What do we have cooking?

Well, we’ve been looking for a way to talk about lots of little stories in a lightning round format, maybe stories that don’t merit a 15-minute discussion.

Because there’s so much happening. You can’t just talk about the entire week in three segments.

Exactly. So this is a game that we came up with our producers. It’s called HatGPT.

Now, that reminds me of ChatGPT, the famous chat bot from OpenAI. Is it similar to that?

Well, no, not at all. But it’s just the name. Please don’t sue us, OpenAI.

OK.

So in HatGPT, basically our producers took a bunch of news stories that happened recently in tech and put them on little slips of paper. And we’re going to draw those slips out of a hat at random and talk about them. And as with AI chat bots, if at any point I talk too long and you want me to stop talking, just say stop generating.

So let me play, for the first time ever, world debut, the HatGPT theme song.

[MUSIC PLAYING]

Time to play HatGPT. All right, Casey. I’ve got a hat here.

Yeah.

It’s filled with little slips of paper. I’m going to mix them up a little bit so you don’t know what’s coming.

OK.

And here. Would you like to go first?

I’d like to go first.

All right. Pick a slip, any slip.

This one is really poking out. I feel like this one wants to be chosen, so I’ll go ahead and pick this one. All right, here is the first prompt.

“Snapchat sees spike in one-star reviews as users pan my AI feature, calling for its removal.” This is from “TechCrunch.” “My AI is pinned to the top of users’ chat feed inside the app and can’t be unpinned, blocked, or removed as other conversations can be. This feed is where Snapchat users have regular interactions with friends and isn’t necessarily a place they want to toy around with experimental features.”

So this is an interesting story because we talked about this on the show recently. Snapchat pushed this feature out called My AI which runs on ChatGPT. And it’s basically a little chat bot friend inside your Snapchat app. And people were bored by it.

And then they rolled it out to all of their users. And I started seeing TikToks about — have you seen any of the TikToks about this?

No, what are they saying?

So people are trying to jailbreak it in some of the same ways that we’ve talked about with other chat bots, trying to get it to say racy or offensive things. People have actually found out that My AI can be tricked into revealing that it knows where you are by using your location data.

So you can say something like, do you know where I am? And at first, it’ll say no, I don’t. I don’t have access to that data. And then people will say, OK, well, where’s the nearest McDonald’s? And then My AI will say, oh, you’re in Jacksonville. It’s just down the street.

So people are trying to jailbreak it like that. But also seemingly a lot of people are upset that they can’t take it off their phones. Unless you’re a Snapchat Plus subscriber, there’s no way to disable this inside your Snapchat app. It’s just permanently pinned to the top of your feed.

So they are apparently submitting lots of angry one-star reviews to the App Store.

They are. And I would give this a five-star review instead. And I would say, let me tell you something about my friend, My AI. He’s awake 24/7.

He always responds to my texts. And he never leaves me unread. For these and other reasons, he deserves a spot at the top of my Snapchat list of friends. And if any of my friends want that top spot, they’re going to have to be as responsive as My AI.

Exactly. Up your game, Casey’s friends. But I also think this is good because it is introducing a whole new generation to the concept of prompt engineering. We’ve talked on the show about how one of the things that we need to just do with these AI models is to start exploring them and learning how to use them, becoming literate in the ways that you talk to chat bots.

And so my argument for why this is actually a good thing is that I think teens should be forced to learn about these chat bots because they’re going to be using them to do their homework anyway. They might as well learn how to do some other things with them too.

Indeed. Now, pass the hat.

OK.

I guess you got to pick one.

Yeah, I get to pick this one. All right. This one says, “Taylor Swift was apparently the only celeb too smart to fall for FTX.” This is from “New York Magazine.” “According to the attorney Adam Moskowitz, who is leading a class action lawsuit against FTX and its celebrity spokespeople, Taylor Swift was considering an endorsement deal in the fall of 2021, but Moskowitz said that during the discovery process the singer actually asked, can you tell me that these are not unregistered securities?”

I love this story.

A question that no venture capitalists ever dared to ask Sam Bankman-Fried.

Taylor Swift did better due diligence on FTX, the failed crypto exchange, than almost all of FTX’s investors.

She did. And look, Kevin, I know you’re not surprised by this, but to the extent that anyone is, let me just say this is one of the great business people of our time.

OK.

Look at what she has done, from rerecording her old recordings so that she has total control over them going forward to all of her other business ventures, which are numerous and incredible. This person knows what she is doing, and hats off to her for asking the right question.

Totally. All the other celebrities, Tom Brady, Giselle, they were all signing these endorsement deals. And Taylor Swift took one look at FTX and said, you know what? This doesn’t actually check out to me.

I wish I could have been there in the room when she looked Sam Bankman-Fried dead in the eye, which this never happened, by the way, but she looked him dead in the eye, and she said, you know what? On that cap table, you’re going to have to leave that blank space.

Come on. Wow. I knew it was coming, but I didn’t know which song it was going to be.

Well, that’s because you know me all too well, Kevin. 10 minute version.

So OK, your turn.

All right. Let’s see what we got here. Ba-bop. “Why researchers turned this goldfish into a cyborg.” This is from “The New York Times.” “Scientists performed brain surgery on goldfish to place electrodes threaded through tiny holes in the fish’s skull to a recording device attached to its head that could monitor neuronal activity.”

I imagine that’s activity related to the neurons. “They’re trying to see,” quote, “how fish navigate the world using different brain circuits than those relied by mammals like us.” Oh, and there’s a picture in this article.

Oh my God.

Oh my God. This —

The fish is wearing what looks like a GoPro on its head.

Yeah, except the GoPro is half of the size of the fish.

Right. So it is a very large — well, probably actually quite small, but so is a goldfish, thing that is mounted onto its head, looks like a GoPro, and is monitoring the brain activity of the fish.

Just imagine you’re a goldfish swimming around the tank and all of a sudden someone reaches down and threads tiny holes in your skull, bad day for this fish.

Well, according to this article, “scientists are curious about the underlying brain mechanisms that allow fish to navigate their world and how such mechanisms relate to the evolutionary routes of navigation for all creatures with brain circuitry.”

I will say this is an area of science that has long fascinated me, how animals navigate.

This area of science has long fascinated you?

Well, by that I mean I read one “New Yorker” article about four years ago and have been thinking about it ever since, which is, how do animals know where they’re going? You have these migratory birds that end up in the same places every year.

They don’t have Google Maps. They’re not doing turn-by-turn GPS. And yet, they end up in the same place every year at the same time.

Yeah, this is one of those where it makes you wonder if we actually are living in a simulation.

Birds aren’t real. OK, let’s try another one out of the hat. “The RNC released an AI-generated ad slamming Joe Biden.” This is from Axios. Says, “the video features AI-created images appearing to show Biden and Vice President Kamala Harris celebrating at an election day party followed by a series of imagined reports about international and domestic crises that the ad suggests would follow a Biden victory in 2024.”

I’m imagining that these crises were like everyone was wearing the Balenciaga Pope coat. And it just shows all Americans —

Why is Drake the Secretary of Defense? That’s weird.

Well, so look, this is an interesting one. I think you can all begin to write your pieces about how 2024 is going to be the AI election. I’m, I guess, in part surprised that Republicans got there first just because this seems like a very cutting edge next generation thing that I don’t always associate with their media strategy.

But on the other hand, we know that this stuff is cheap. And if you don’t care if it’s accurate or not, go ahead and use the AI to make all your ads.

Totally. I mean, I will say political ads have always had a very liberal definition of what constitutes the truth. There have just been lots of attack ads over the years that have used manipulated images. Or I remember they darkened Barack Obama’s skin tone in an attack ad.

So the idea of using technology to doctor or otherwise manipulate imagery that goes into a political ad is not new. What is new is being able to do it in more realistic ways. But I feel the same way I do about this that I do about the AI music which is, I don’t think Joe Biden is in danger of being deepfaked in a way that’s going to hurt his candidacy in 2024, in part because the real Joe Biden is the president and can come out and say that’s not real.

I think what is much more plausible is actually somebody saying that something is AI-generated when it wasn’t. I think if the “Access Hollywood” tape had come out in 2024 instead of 2016, I think Donald Trump’s first response would have been that’s a deepfake. It wasn’t actually me. And a lot of people would have probably believed him.

Yeah, and so we are going to have to develop tools to determine the provenance of these things. So agreed, there’s a lot of really weird and dark stuff coming. But I do feel this was the week where we saw the opening salvo of these wars. It’s like, OK, well, here’s your first AI political ad.

Totally. And I think there’s going to be a whole cottage industry of professional political operatives who are using AI to try to do attack ads for different candidates. I think that will probably be more successful on the local level than maybe the state level.

National politics, who knows? But we may see a big October surprise AI deepfake this next election cycle. But I would bet against it.

Yeah, and to that point, I do think this stuff gets easier to pull off when you just don’t the candidate that well. So like the reason it’ll work better at the local level is, well, maybe you’ve never heard of the other person running for Congress in your district, and then you see a deepfake of them.

What’s something funny that they could do?

Doing whippets.

Doing whippets. Just — we can’t say doing whippets. Because we could start a national whippets craze.

Do not say whippets, kids.

Stop doing whippets, kids. Knock it off.

All right, next one. This one says, “UK blocks Microsoft’s $69 billion bid for Activision; a blow for tech deals.” This is from “The New York Times.” British antitrust regulators on Wednesday dealt a major setback to Microsoft plans to acquire the video game giant Activision-Blizzard for $69 billion, blocking the proposed deal and handing a notable win to government enforcers who want to rein in big tech.”

Well, I think this is a really important thing. Activision-Blizzard makes some of the most popular video game franchises in the world, including “Call of Duty” and “Overwatch” and others. And as always with antitrust, there are a few things to keep in mind.

The way that antitrust was practiced for most of the last 50 years was that the focus was really on consumer welfare, basically the idea that having robust competition is always better for consumers. But lately they’ve been going back to this idea that the bigger and more powerful a company is, the more it essentially structures the market in a way that just makes them more powerful.

And so you get regulators saying, wait a minute. We can’t let these companies get too big. And so I was actually sympathetic to the idea that maybe you don’t want this acquisition to go through. Maybe you don’t want Microsoft owning yet another of the biggest video game companies as part of what is already their pretty big video game business.

Now, Microsoft would say we’re not actually that big. Sony is much bigger.

In the gaming market.

In the gaming market. And I was just never convinced that this was going to be pro-competition. To me, it seems obvious that the more video game companies you have on the market, the more competitive it’s going to be. And so I was always uncomfortable with this one.

Yeah, but I think a lot of people did assume that it was going to go through. It seemed for a while Microsoft was playing the regulatory game very well and doing what they needed to do to avoid having regulators come down on them like this. At the same time, that’s not the only market that matters for them.

Yeah, so Microsoft’s going to fight it. Maybe they win it on appeal. But this is a big one and one to watch. And I think it will tell us a lot about the extent to which these giants are going to be able to make huge acquisitions over the next few years.

Yeah, I think there’s a default assumption now that at least I’m going in, where if even Microsoft can’t thread the needle with a big acquisition of a gaming company, which, by the way, is not its primary thing that it does. The vast majority of its revenue comes from things that have nothing to do with video games.

And so if even they can’t acquire this video game company, I just think there’s a certain — if you are one of the top three tech companies in the world, I think your default assumption should be that you are not going to be allowed to buy anything.

Yeah, so maybe go try to build something. Yeah, it’s time to build.

Yeah, it’s time to build. And that was HatGPT. That was fun.

Close the hat!

[MUSIC PLAYING]

When we come back, we’re going to talk to Ben Smith, the former BuzzFeed News editor-in-chief and former “New York Times” media columnist about how tech giants shaped the media of the 20 tens.

And also, what color was that dress?

We’re going to solve it. We’re going to crack the case.

[MUSIC PLAYING]

So, Casey, a story that we have talked a lot about in an indirect way on this show is the relationship between the big tech platforms and the media, this symbiotic but also antagonistic relationship that exists at times between companies like Facebook and Twitter and companies like “The New York Times” and Platformer and how the decisions that are made by the big tech companies about what to put in people’s feeds and how to rank their algorithms really have consequences for journalists, for media companies, and for ultimately media consumers.

And as we’ve been talking about this story, I was reading an advance copy of a new book by our friend Ben Smith. Ben is the former editor-in-chief of BuzzFeed News, which recently announced that it is being shut down. He’s also a former New York Times Media columnist and now runs a new media company called Semafor.

And he’s really an interesting person to talk to about this topic because he was so close to the action. I mean, if you can remember back to, let’s say, 2015, it really seemed like BuzzFeed had really caught a tiger by the tail and was in a leading dominant position in the digital media business. And Ben was a big part of that operation.

Ben’s also a huge gossip.

Yeah, we love talking to Ben. We should also mention both you and I are thanked in the acknowledgments of his book. I did skim the acknowledgements.

I was acknowledged? Oh, that’s nice.

I know. So he is a friend of the pod and also just a really interesting person to talk to about digital media. His book, “Traffic,” comes out next week.

[MUSIC PLAYING]

Ben Smith, welcome to “Hard Fork.”

Thank you guys for having me.

In a recent piece you wrote for Semafor, which touches on some of the same themes you write about in your book, you wrote that we’re currently experiencing, quote, “the end of one era and the beginning of another in digital media.” So I’d love you just to expand on that.

In your view, what exactly are we saying goodbye to?

Yeah, I think we’re seeing the end of an era that really began in the early oughts. I would put it — I would put it a bit earlier and really, like a lot of things, in a scene in Downtown Manhattan.

And there was this moment, it’s almost impossible to remember now, where it was like, oh my God, has the center of technological innovation and venture capital investment shifted from Silicon Valley to Silicon Alley in Manhattan? Because there were these new media-centric tech startups, some of which were really just media, like Gawker, BuzzFeed, HuffPost, Business Insider, and some of which were tech media like Foursquare. But there was this sense of this dynamic new New York scene where culture was being made and money was being made.

Yeah. It’s a really interesting time to be having this conversation because I think we’ve seen just in the last couple of weeks the real tail end of that era. So BuzzFeed News, where you were the founding editor-in-chief, is closing down. Obviously, Gawker is no more. And there are a host of other failed websites, including one that I worked at called Fusion.

So what do you make of the closure of BuzzFeed News and what it means for the closing of that chapter in media and maybe what comes next?

Yeah, I mean, it makes me incredibly sad. And I think the big story was that there was this new kind of media, social media, that promised to and really did transform news and journalism and media. And a group of companies, in some ways BuzzFeed most successful of them all, really built on top of it.

And it’s not like we didn’t know when we were building it that we had what they call in dex, a dependency on Facebook. But it’s also very hard to get out from under that. When you’re riding this immense tide you think, oh, we got to figure out some other side business.

And so I think the big story is these companies were totally dependent on the social media ecosystem, which is itself collapsing.

So let me just try to paint a picture of the internet in the sort of early mid-2010s because I think even though we all lived through it, it feels very far from the moment that we have today.

I remember in 2012, I want to say, or maybe 2013, there were these signs that there was this huge surge in traffic coming from social media, and no one could really figure out where it was. And eventually it turned out that was from Facebook.

And it was the kind of numbers that media companies just don’t see, just astronomical amounts of traffic coming to their websites. And that seems like the precursor to the world that you document in this book of these new media companies, Gawker, BuzzFeed, Vox, Vice, et cetera, that came in to ride that wave and try to turn that into a sustainable business. Is that right? Is that the inciting event of this whole boom in digital media in the 2010s?

Yeah, I think the massive ocean of traffic from Facebook was what delivered these numbers that prompted venture capitalists to pour hundreds of millions of dollars into digital media and the notion that, well if you’re getting a billion users and each user is worth $100, this is a lot of money. I mean, I think that was the basic math.

And were there people at the time who thought, maybe this isn’t sustainable. Maybe this is just — because I think we now know, and you write about in your book, that some of the reason for this huge surge in traffic was because Facebook was trying to kill Twitter. Twitter was growing as a rival, and Facebook was saying, well, we can’t allow them to overtake us, and so we will just flood the newsfeed with news sites, with publications, so that we can try to steal some of that momentum away from Twitter.

Yeah, I mean, this is — I mean, I think if Facebook had to have that back, they might have said, wow. You know what, Twitter? You can own this entire toxic stew, and we will be a great place to sell washing machines. There was this tactical decision by Facebook to copy, as it had copied lots of other apps, what this rising little app called Twitter was doing, which was provide news to people.

Facebook never actually really got particularly good at the live news thing, but they accidentally became the absolute heart of the digital publishing industry and delivered all these views to publishers who — the views kept getting less valuable. This was the sort of thing you were seeing out of the corner of your eye and with a little alarm bell ringing that, wow. I had 100 million views last year and 200 million views this year, and my revenue has gone up 5 percent. What’s going on here?

Why is it that doubling audience growth doesn’t equate to doubling revenue growth? What was the hypothesis there that proved wrong?

So I think it’s maybe a lot to call it a hypothesis. Maybe a delusion. But there was this new ad inventory on the internet. And you could sell it. And there wasn’t that much of it. And the problem with commodities is that they derive their value from scarcity. There’s only so much gold in the world. And so it has a certain value — oil. The problem was these new factories, Google and Facebook particularly generated infinite scale. And if you’re selling something of which there’s an infinite amount, it doesn’t have a lot of value.

I remember at one point BuzzFeed was doing, I want to say, like 7 billion content views a month, which in the history of the media business was on the far end of success. There haven’t been many businesses built bigger than that.

So if I’d been in your guys shoes, I would have done the exact same thing as you did. We built one of the biggest media companies of all time. Let’s hire a bunch of reporters and see what we can do.

But at the same time, looking back, it does seem like it was a mirage. So I wonder for you. Is there a sense of astonishment that it turned out the way it did, given the scale that you guys had achieved?

So I don’t think it was a mirage. I think the world changed and changed again. And I guess I’m no longer astonished because I think we felt the peak, the crest 2015, 2016. You can’t really separate media from the culture around it, and the way people saw news changed around the Trump election and didn’t change back.

I mean, I do want to talk about this issue of the relationship between digital media companies and social media companies during this period, because Casey and I both worked in digital media at this time. And I think there was this feeling, like we were the oxpeckers who like lived on the back of the rhino. And we’re just —

What did you just call me?

Like those little birds.

OK.

And we were so dependent on traffic because it was such a big percentage incoming views were from Facebook that you basically existed at the mercy of these large platform companies. And they could just decide at any moment to turn the spigot off or to turn it on and give you more traffic.

And one of the things that I actually found really interesting in your book is that BuzzFeed was cultivating a relationship, a special relationship with Facebook, that allowed it to surf this wave better than most publishers. So talk about that relationship, these large social media companies that are just throwing their weight around, changing their algorithms in ways that they turn the dial one notch, and for the rest of the digital media ecosystem it means a seismic change that everyone has to adapt to.

Yeah, and I love the metaphor. the rhinos obviously live in Silicon Valley and the rhino peckers.

Oxpeckers. They’re the little birds that live on the back of the rhino.

We’ve got to stop talking about these peckers. But go on.

They live in New York. And there’s some sense in which I think this book is told from looking in the wrong side of the telescope. I mean, the power and the biggest decisions are being made in Silicon Valley. And media is downstream of them.

I think what Jonah Peretti, my former boss and partner at BuzzFeed saw was — and the reason that I think that we navigated this so successfully, there were two. One was that a lot of publishers saw these platforms as fundamentally technical, that you could figure out technical rules that you could then tell your team — the way SEO is technical — follow these rules, and it will game the platform, and you will gain an advantage.

But in fact, the platforms are run by people who are not morons. And if you figured out some trick, it’s a bug, and they’ll fix it. And Jonah always steered away from the gimmicky stuff and toward, well, what do people — these platforms are ultimately places where human beings are gathering. If you have a really good sense that they want to be entertained, they want to be informed, and you give them stuff that does that, it’s going to survive the minor changes to the platforms.

And he then also had this very warm, strong, two-way relationship with Mark Zuckerberg and with other executives at Facebook, where he was giving them ideas about how to run their platforms and also learning and gathering intel from them that was certainly shaping what we were doing.

Right. I mean, it seems like he was essentially running the Facebook Newsfeed at certain points during its existence.

I mean, Zuckerberg, right before I started at BuzzFeed, Facebook tried to acqui-hire Jonah.

How do you think the world would be different today if that had happened, if I mean, Mark Zuckerberg had succeeded in buying BuzzFeed and making Jonah the head of the Newsfeed?

Well, I would still be blogging at Politico. And I do think that it would have been nice if somebody in that Facebook product organization or whatever you want to call it was thinking more about how human beings use media in real life rather than about these abstract measures of engagement. I don’t know if that really would have been Jonah, but that’s what I like to think.

One more question about Facebook. There’s a really interesting scene in your book. It’s 2015, and the dress has just happened, this mega viral post about this dress and whether it’s — what was it? Blue and gold or —

White and gold or blue and black.

Right. And it goes nuts, millions and millions of views. Everyone sees it. It’s this apical event in 2010s internet history. And everyone at BuzzFeed is very proud, understandably, of this successful post. And Jonah Peretti, CEO of BuzzFeed, is talking to Adam Mosseri, who now runs Instagram, but at the time was in charge of the Facebook newsfeed.

And as you report, Adam Mosseri’s response is basically, how often do you think that something like the dress should happen? And you wrote that that surprised Jonah. You say, quote, “to them, the dress hadn’t been a goofy triumph. It had been a kind of bug, something that scared them.” Why do you think that the dress scared Facebook?

What happened with the dress was just this absolute loss of control of the platform. They were finally on notice that they could do damage. And I don’t think it’s crazy to think, wow. Today, it’s this really fun dress. What is it tomorrow?

It now seems obvious that what y’all were doing at the time was just a lot of experimenting. You were discovering what would happen if you wrote different kinds of things. And I think the Facebook era of journalism was probably the most instrumented that journalism had ever been.

We didn’t used to have this kind of insight into what people liked and what they didn’t as much. And so I can understand why you would just follow those numbers to see where they would go for as long as you did.

Yeah, and at times, those numbers were an insight essentially to human psychology and what are people interested in at their best? And at times, they were an insight into what were the technical decisions people of Facebook or Google had made about what they thought people should want. And sometimes they were a kind of combination of those two.

So now we’re in 2023. The sort of traffic fire hose of Facebook and Twitter, to the extent it ever was one, and all these other platforms has slowed to a trickle. And so this competition for traffic, for audience, for whatever you want to call it seems like it’s drawing to a close.

So what defines the media moment that we’re currently in? And what are the platform dependencies that you see right now that publishers maybe should be worried about, given what happened last time?

Yeah, I mean, and I think part of the reason I went out to start a new thing is because I think one of the things I did learn from BuzzFeed is one of these moments of immense change, there is this opportunity to actually go to consumers and say, you’re sick of being limited to only reading a couple of newspapers? Good news. You can read everything in the world at the same time and also have everyone’s opinion.

And now I think it’s more like, you’re sick of having everyone in the world screaming at you at the top of their lungs at the same time, being overwhelmed, and then simultaneously not knowing who to trust. OK, we will try to build a straightforward, transparent kind of journalism that also does a really good job pulling in everybody else’s work, and linking to it, and sharing what else is out there. So you don’t have to go and google 11 versions of the same story to triangulate what’s really going on.

But I think — I mean, I think email is a real dependency. Do you think that everybody will be using email the way they do in 10, 15 years? I use it a lot less than I used to.

I think TikTok, which is this seductive, incredible platform that has enormous scale among people who care about news, people care about technology, all sorts of stuff, has also all of the risks that Facebook did for publishers. But I think that has made people incredibly leery going on there in a way that maybe is itself a little — you can overlearn these lessons. I mean, those are, to me, the obvious platform dependencies.

There’s one that you didn’t mention, which is Google, which I feel like most of the digital publishers I know are just hugely dependent. And if you can get your story underneath that Google search bar, and it’s mostly just the worst clickbait you’ve ever seen —

I mean, it’s worse than what was on Facebook in 2015 is under the Google search box right now. How have you thought about Google as you try to build a thing? And I’m curious how at risk you think the other publishers are of Google one day waking up and turning off that spigot?

You know, I didn’t mention it because I, in my own mind, have thought I don’t want to build on top of it. It’s a pretty broken experience now. You search where to eat in New York, you get 11 different sets of garbage links. You have to go to the 17th page to find somebody’s blog that can tell you, name a restaurant for you.

And then also just that it’s now suddenly under attack from a number of directions. I mean, these sort of AI chat bots are one of them. But also people get sick of things. They don’t last forever. And I think this sort of web search feels like a declining stock. And publishers who are really wedded to it just feel like they’re going to be squeezed into being low quality AI-written garbage content farms.

Well, I’m glad you brought that up, because this is my big fear about the future of news is that right now so much digital media is supported by that kind of very quick, easy, cheap to produce material that probably has an affiliate link in it to buy something. And right now a junior staffer is being paid a living wage at a lot of places to bang out three or four of those posts every single day.

Does that sort of job seem as at-risk to you as it does to me?

Yeah, I think if you are reorganizing and reframing the stuff that’s already out there rather than reporting, bringing something new in, those are the jobs that are most vulnerable to AI.

I want to talk about AI more, because you at Semafor have been experimenting with AI in ways that I found really interesting and really different than a lot of publishers who I think are either trying to use this to churn out low quality SEO bait or are just scared of the whole thing and don’t really know how to approach it yet or are having a bunch of meetings talking about how their AI strategy is going to look.

So what is your AI strategy at Semafor and how are you thinking about this new world of generative AI?

Yeah, I actually think about it as a tool. And in some sense, I think we’re trying to work on the high end of really human-created journalism, where you know the person’s name, and she’s a star, and you feel connected to her and want to know what she thinks about her area of expertise. And so that’s not in the realm of things that can be replaced by AI.

But there are tools. I think there’s amazing new video tools to me are the most obvious.

Like the text-to-video, like type in text that generates video?

No, Joe Posner who helped create Explained at Vox has been doing these videos for us that intersperse animation with videos and words. And creating animation is something, I don’t know, 3,000, $10,000 a minute to do in the past. And he’s working with this artist in Australia who’s doing it using Stable Diffusion for an amount that we can afford. It’s not like we’re putting some animator out of work. We just wouldn’t be doing it. And I think just in the way that the creation of digital video with your phone suddenly meant a lot of people could create video, some of these video tools are going to further mean that making movies is less a technical skill — still an incredibly competitive and difficult creative skill. And I think that’s really powerful and interesting.

And sure, it will also lead to all sorts of dumb scams and scary scams and popes wearing weird jackets and stuff, but if you still have some connection to the notion of the internet as this democratizing force, it is the democratization of these really cool interesting tools.

And then just in terms of workflow, I mean, I don’t think anybody looks at Grammarly, which is a pretty cool piece of software, and says this is a threat to journalism. I mean, it’s a threat to my constant typos. But I think there are AI tools that can make you a better writer. And there are also journalists who are not particularly good writers and people who didn’t go to fancy schools but who are great reporters, and who the hollowing out of editing in newsrooms left outside.

There used to be rewrite men. Get me rewrite, which meant that a reporter did not need to be a fancy writer. They could be a great reporter. And I don’t know, if a reporter of mine wanted to gather some facts and put them into ChatGPT before they filed them to me, look them over and file them and that made them feel better about it, fine.

You give this piece of advice to everyone. It’s one I’ve taken very seriously myself. And that piece of advice is that scoops matter more than everything else in journalism. And I wonder if you believe that is more true in a world where AI is generating at least the first draft of most of the stories that you’re reading?

Yeah, I mean, I think newsgathering is the thing that’s going to be really hard for AI to do. Not that, by the way, you can’t think about, huh, could I script some kind of email that emails 50 percent of my sources and says, hey, do you know anything about this and arranges their answers in a way that I can find really usable? If it isn’t too obnoxious and doesn’t freak them out too much, I mean, sure, maybe.

But yeah, I think the most human parts of journalism are the ones that are going to be hardest to replace. I like to think that fortunately, media reporting is the absolute most inside your own navel pursuit in possibly human society. And so it’s going to be the last one hiding out from the robots.

I think all of us in this new AI world think that we secretly have the one job that is the least likely to be replaced by the robot.

You were one of the first media executives I knew who went all in on Twitter. You were a heavy, heavy tweeter. Everyone at BuzzFeed News was constantly tweeting all the time. It was a platform where you found writers and cultivated talent, but also where you distributed stories.

What do you think the diminishment of Twitter under Elon Musk has meant for the media? Do you miss the old Twitter? And how are you thinking about Twitter as a part of the media business going forward?

So as somebody who had his brain basically rewired by Twitter and really loved it for a while and met great sources, learned things on there, I mean, I’m pretty sad about it, honestly, personally. I do think that whether or not Elon had made certain decisions, people are sick of social media. People are leaving Facebook not because its technical architecture is a little wrong way. It’s because these are social products. They’re like nightclubs. They’re like bars.

You go there because your friends are there. They fall out of fashion for ineffable reasons, and everybody leaves. And you can’t — that’s not a stoppable thing. The place can’t get a new sound system and you’ll come back. It’s just culture. I mean, I think you can see the lights going out.

I think Musk in some ways drew a spasm of new attention to it on the way out, but I think basically people were going to leave.

It’s a really tough time in journalism right now. I mean, it’s seems like it’s always a tough time in journalism, but I think just the news of BuzzFeed News closing down, people losing jobs at other outlets, economic troubles, and the dry up of advertising, there’s a sense that there’s a doom and gloom around the media industry right now.

But there are some people who believe that the death of social media and the traffic chasing that came with it is a good thing. Hillary Frey, who’s the editor-in-chief of Slate, recently wrote a piece about this. And she wrote that quote, “being free of our reliance on tech platforms is going to be better for our business. The old traffic tricks don’t work anymore. What does is great work and serving an audience who loves you so much they’ll pay for you. That’s actually always been the case with journalism.”

So is that something that you believe? Is this a good thing that we’re going back to the fundamentals of journalism and we’re not so obsessed with chasing this ephemeral social media traffic anymore?

I’m a little reluctant to say that every technological or social change should be put into the good thing or bad thing bucket. I think Fox, BuzzFeed, Vice, other publishers did amazing journalism that mattered an enormous amount and had an incredibly salutary effects — Gawker, Jezebel in its era, and the era changed and ended. And those businesses could have been run better, and they could have maybe found a narrow path through, but the pendulum swung.

And I think there will be things about subscription media and these sort of narrow casts, sub stacks where you just have to tell your followers over and over exactly what they want to hear. And that has its own problems.

I mean, it’s funny people talk about echo chambers, but the echo chambers are the closed ones. That’s why they echo. Facebook and Twitter aren’t echo chambers. They’re something way crazier. And I think people are really eager to retreat back into echo chambers. It’s probably much healthier to spend your time in an echo chamber then in the wide open field screaming at a bunch of maniacs you hate. But that has its own challenges and problems, and many of the most successful sub stacks are the most partisan ones.

Media is part of culture. It changes. It evolves. And there’s — and I think that I’m reluctant to say, well, this one’s a good one now. They’re all bad, basically. Just degrees.

Well, we’ve been talking a lot of stuff about the meat industry, but I think — can we at least agree that it actually is still the most fun business? If you set aside all of the problems that there’s still nothing else that we personally would rather be doing, right?

It beats working for a living.

Ben, thanks so much for coming.

Thank you, Ben.

Thank you for reading the book.

[MUSIC PLAYING]

“Hard Fork” is produced by Davis Land and Rachel Cohn. We were edited this week by Paula Szuchman and Jen Poyant. This episode was fact checked by Caitlin Love.

Today’s show was engineered by Alyssa Moxley. Original music by Dan Powell, Elisheba Ittoop, Marion Lozano, and Sofia Lanman. Special thanks to Pui-Wing Tam, Nell Gallogly, Kate LoPresti, Christopher Sprigman, and Jeffrey Miranda.

You can email us at [email protected].

[MUSIC PLAYING]