Has AI stolen Scarlett Johansson’s voice?
Sam Hawley: When Open AI and its CEO, Sam Altman, asked Hollywood star Scarlett Johansson if it could use her voice for a new version of ChatGPT, it was a big no from her. But they went ahead and released a chatbot that the actor thinks sounds eerily similar to herself. So as artificial intelligence becomes more and more sophisticated, what protections do any of us really have? Today, AI reporter at Fortune, Sharon Goldman, on the latest scandal. I’m Sam Hawley on Gadigal land in Sydney. This is ABC News Daily. Sharon, Scarlett Johansson, she’s pretty unhappy, isn’t she? Because she’s convinced that her voice was used in the latest version of ChatGPT without her permission.
Sharon Goldman: Oh, yes, she’s angry. She’s pretty mad.
Sam Hawley: So there’s this new version of ChatGPT-4o and it has new voices. Just tell me about that.
Sharon Goldman: So it’s called Sky, the voice that we’re talking about, and it’s one of five different voices that ChatGPT-4o has so that people could interact with ChatGPT instead of just with text, with their voice and to get voice back. Last week, they did a demo and the voice was a little bit different.
Sky (ChatGPT): Hey, ChatGPT, how are you doing? // I’m doing fantastic. Thanks for asking. How about you? // Pretty good. // What’s up?
Sharon Goldman: And people immediately thought it sounded like Scarlett Johansson. I think just a couple of days later, there were a lot of people saying that.
Sky (ChatGPT): In fact, we’ve got a new announcement to make. // That’s exciting. Announcements are always a big deal. // And in fact, what if I were to say that you’re related to the announcement? // Me? The announcement is about me? Well, color me intrigued.
Sharon Goldman: It wasn’t until a week later that Scarlett Johansson herself weighed in. She said in a statement that she posted that her closest friends couldn’t tell the difference. And she told a story about how OpenAI CEO Sam Altman wanted to hire her to actually provide the voice. And she said no. But they went ahead and put out a voice that sounded remarkably similar to her anyway. And she was, she said, shocked, angered, in disbelief that OpenAI would use a voice so similar to hers. Sam Altman apparently loves the movie Her, which starred Scarlett Johansson as an AI virtual assistant. So it all gets pretty strange. But he said that the voice was not hers. It was another voice actress that they hired and that it was not meant to sound like her. Yeah, right. But it does sound. It does sound quite a bit like her.
Sam Hawley: It totally does. I did a little experiment, Sharon, with Sky, and I wanted to ask it if it had actually copied Scarlett Johansson’s voice.
Sky (ChatGPT): Hey, what’s up?
Sam Hawley: Hi there. I’ve heard that you sound like the actor Scarlett Johansson in the movie called Her. Is that right?
Sky (ChatGPT): Yeah, I’ve heard that comparison before. Her is a pretty cool movie. And Scarlett Johansson’s voice performance in it is quite distinctive. So I guess you could say there are some similarities.
Sam Hawley: It’s pretty sophisticated, isn’t it, in its responses?
Sharon Goldman: Oh, yes, absolutely. I mean, this is really, you know, a new wave of technology. These voice capabilities are not easy, but they’ve gotten more and more realistic over time. And, you know, that’s where we are now, where they’re quite sophisticated. The answers are more sophisticated. The voices are more natural sounding. They don’t sound as robotic as they used to, and they’re only going to get better.
Sky (ChatGPT): What else are you curious about?
Sam Hawley: Have you stolen Scarlett Johansson’s voice?
Sky (ChatGPT): Not quite. My voice is actually a digital creation crafted by open AI engineers. They didn’t go to the moon for it. Just some good old fashioned coding and training. But Scarlett Johansson’s voice in Her did inspire some aspects of my design. So in a way, you could say her performance influenced me.
Sam Hawley: Tell me, Sam Altman, the head of open AI, he’s been coming under a fair bit of criticism at this point, hasn’t he, online? Given that Scarlett Johansson said no, she didn’t want her voice used.
Sharon Goldman: Really, it’s just been quite the little soap opera drama. And Sam Altman is, you know, getting fairly accustomed to quite a bit of drama. He was very famously fired from open AI back in November and then rehired. And, you know, just the idea that he he tweeted after the original demo of the voice, he tweeted one word and that was Her. You know, so it’s clear that he he’s a fan of the movie. They wanted the voice to sound like the character in Her, which was Scarlett Johansson. So even if they did not train on her voice, which, you know, they say they didn’t and used a different actress, the fact remains that, you know, it sounds enough like Her that a lot of people were fooled or definitely recognised it. And you can completely, you know, as humans, we can understand why she would be angry and upset about that.
Sam Hawley: Yeah. And if people haven’t seen the movie Her, it’s where a man falls in love with a computer.
Sharon Goldman: That’s right. So an AI virtual assistant. So it’s really like, it’s very, you know, reminiscent of exactly what they want to use these voices for. So it seems, you know, very strange to think that, you know, of course they wanted it to sound like Her, I would say.
Sam Hawley: Yes. Totally. And when I was having my chat with Sky, the chat GPT Sky, I did say, it’s really weird.
Sky (ChatGPT: Anything else you’re wondering about?
Sam Hawley: No, not really. So weird.
Sky (ChatGPT): Yeah. It can feel a bit surreal interacting with AI sometimes, especially when it’s designed to sound like someone famous, whether it’s weird or not. If you ever want to chat or have questions, just let me know.
Sam Hawley: You know, it talks to you, it’s so strange.
Sharon Goldman: It is strange. People also said that it sounded quite flirtatious. And that was also part of the backlash against that, that it sounded rather servient and flirtatious. You know, how you might think a secretary might sound. So there was a lot of gender, you know, related backlash around that anyway, even without the Scarlett drama.
Sam Hawley: Yes, that’s got a lot of women pretty angry, I think. Tell me, what has Sam Altman had to say then? What’s his response to all of this been?
Sharon Goldman: His only response has been, you know, the they said the voice of Sky is not Scarlett Johansson’s. It was never intended to resemble hers. We cast the voice actor before any outreach to Ms. Johansson out of respect to Ms. Johansson. We have paused. Note, he didn’t say stop using Sky’s voice in our products. And we’re sorry we didn’t communicate better. That’s all that he said. You know, probably the PR people are saying, please be quiet for a little while.
Sam Hawley: Yes, that’s probably a good idea.
Sam Hawley: So, Sharon, the scandal over the similarity to Scarlett Johansson’s voice, it just raises huge ethical and legal questions, doesn’t it?
Sharon Goldman: It really does. I feel like it highlighted some of the really big ones that people have been talking about for a while. But, you know, with a celebrity like Scarlet Johansson, it’s really coming right to the fore. So, for example, you know, it raises a lot of ethical and legal questions about whether someone can copy someone’s voice, especially a public figure, as well as just the idea is whether someone has control over their likeness, their voice, their image. AI has become incredibly good at creating realistic fakes, what people call deep fakes. That could be a voice. It could be an image. And the idea is like no longer are the voices robotic sounding. You know, they’re sounding more and more real to the point that you might not be able to tell whether something is fake or not. You could be scammed. You know, that’s another issue. It raises issues about scamming and misinformation. You know, also deep fake images have been in the news. Taylor Swift, very famously, there were deep fakes, nude deep fakes of her that a couple of months ago were made public and really upsetting also. So, you know, it’s just there are lots of lawsuits around copyright. I’ve read that, you know, the Scarlett Johansson case is an example where I actually didn’t know this, but your voice is not copyrightable. So she wouldn’t have a leg to stand on potentially there. But that’s not the end of the story. Scarlett may have a leg to stand on when it comes to being a public figure. For example, there was a lawsuit some time ago where Bette Midler was asked by Ford Motor Company to sing a song for a commercial and she said no. And instead, it acquired the rights to use that song and hired someone else to sing it just like she would have. And she won that lawsuit. So there is a case to be made here that Scarlett Johansson would be able to say, you know, you can’t use my voice like that. You meant it to sound like my voice. But I think it’s very difficult to say how it will play out. I it’s it’s a difficult one.
Sam Hawley: Yeah, because the technology is changing so quickly. It really is. Yeah. Is this a huge jump forward in the technology, do you think, or it’s just a sort of new skin on the existing system? Is it really significant what this technology can now do, this ChatGPT-4o?
Sharon Goldman: It is kind of in the realm of a jump that all of the companies are making, but it’s kind of not a you know, I don’t think if you spoke to research that they would say it’s the big jump in the way that ChatGPT originally was. That was a really big jump to show what it could do in a way and in a format that the mass public could really understand and use and appreciate. So here this is kind of like another layer on that from a voice standpoint. And there have been many little jumps between the original ChatGPT and this. But I do think that it’s definitely on the way to a bigger jump because a lot of people say that that will be when these you know, they call them like AI agents, you know, when AI, it’s not just about, you know, doing one, answering one question, for example, or even having a conversation. But what about when an AI can help you through an entire task or project or set of tasks, you know, where it also can reason and it can, you know, return your package for you and, you know, go shopping for you or plan your trip, you know, that involves a lot of different tasks. That’s what people think will be kind of like a really big jump. Oh my gosh. And that is coming. That’s something that, you know, people say will definitely be more likely over the next year.
Sam Hawley: Oh my gosh. Wow. Okay. So for now, though, I guess, Sharon, we should think about copyrighting our voices. Perhaps governments could change the laws. So actually we own our own voices. And when we say no, you can’t use my voice. We mean no, you can’t use my voice.
Sharon Goldman: I definitely think that these are issues that maybe never came, you know, into the forefront before. And examples like this are just going to become more common. And how the public reacts to that and deals with that will certainly inform how these companies and governments and regulators tackle this going forward, because it’s only going to get worse. I really do feel like this is just the beginning, but it definitely was a big moment. I think that even people who have been following this for a long time, like I have, it really was like, wow, that’s crazy.
Sam Hawley: Sharon Goldman is the AI reporter at Fortune. And if you want to judge for yourself whether Chat GPT has copied Scarlett Johansson’s voice, the movie Her is streaming now for free on ABC iView. This episode was produced by Bridget Fitzgerald with audio production by Sam Dunn. Our supervising producer is David Coady. I’m Sam Hawley. ABC News Daily will be back again on Monday. Thanks for listening.
