OpenAI Makes Scarlett See Red

Sam Altman was on stage at Dreamforce 2023 when he was asked to name his favorite movie about artificial intelligence. It was a deceptively good question. As the cofounder and CEO of OpenAI, Altman was uniquely well-positioned to dictate the path that AI would take in the future, so knowing which fictional version of AI he preferred could shed light on what he expected, hoped, or feared might happen.  

Altman thought about it, then said one word: Her, a 2013 film about a sad man who falls in love with a sentient virtual assistant voiced by Scarlett Johansson. According to Altman himself: “The number of things that I think Her got right, that…this idea that it is going to be this conversational language interface, that was incredibly prophetic, and certainly more than a little bit inspired us…So it’s not just like a prophecy, it’s like an influenced shot or whatever.”

He Likes Her

Little did the crowd at Dreamforce 2023 know, but Altman and OpenAI were already hard at work bringing Her to life. Two weeks later, OpenAI announced that they were giving ChatGPT a voice. Now users would be able to interact with the AI, have conversations, and ask it to do things using natural spoken language – a conversational language interface, if you will. It would have relatively limited functionality and conversational ability at first, but the end goal was clear: Altman and OpenAI wanted to make Her a reality. They just needed two more things to bring Samantha (the AI from Her) to life: a series of massive technological breakthroughs and Scarlett Johansson.

By complete and total coincidence, Altman and OpenAI met with Scarlett Johansson right before and during that very same September. Johansson described the meeting in a statement to the press, saying: “Last September, I received an offer from Sam Altman, who wanted to hire me to voice the current ChatGPT 4.0 system…He told me that he felt that by my voicing the system, I could bridge the gap between tech companies and creatives and help consumers to feel comfortable with the seismic shift concerning humans and A.I. He said he felt that my voice would be comforting to people.”

Unfortunately for Altman, his proposition fell on deaf ears. Johansson said no, and that was the end of it. Altman had everything, but he’d never have Her.

And Then Something Weird Happened

OpenAI announced something big in mid-May 2024: a new AI model called GPT-4o. This new model would combine features from ChatGPT and OpenAI’s other models to be able to see, hear, process, and interact with users like never before. The idea sounded familiar, as did the AI itself.  

“Nine months later, my friends, family, and the general public all noted how much the newest system named ‘Sky’ sounded like me.” Johanssen said in the same statement. “…I was shocked, angered and in disbelief that Mr. Altman would pursue a voice that sounded so eerily similar to mine that my closest friends and news outlets could not tell the difference.”

According to Johanssen, Mr. Altman contacted her agent two days before the demo was released, asking her to reconsider his offer. He did not wait for an answer. As a result, Johansson felt forced to hire legal counsel. Her attorneys wrote two letters to Altman and OpenAI, asking them to detail the exact process they used to create the voice.

It Wasn’t Her

It only took a few days for OpenAI to buckle under the legal and reputational threats from Johanssen’s team and the general public, though they denied any wrongdoing. “The voice of Sky is not Scarlett Johansson’s, and it was never intended to resemble hers,” Altman said in a statement. “We cast the voice actor behind Sky’s voice before any outreach to Ms. Johansson. Out of respect for Ms. Johansson, we have paused using Sky’s voice in our products. We are sorry to Ms. Johansson that we didn’t communicate better.”

OpenAI later posted an update on OpenAI.com breaking down the process by which they selected the actors for ChatGPT’s five voices. They claim to have received over 400 submissions from voice and screen actors wanting to voice ChatGPT beginning in early May 2023, and that they spent the rest of the month narrowing the list down to a handful of prime candidates. The post elaborates further on the selection, preparation, and recording process – presumably to further prove that they totally didn’t want to use or imitate SarJo’s voice – but it doesn’t accomplish what they think it does. Why? Simple: As Johanssen herself pointed out in her statement, Altman posted a tweet on X (formerly Twitter) the same day they announced GPT-4o that simply said “her.”

If Scarlett Sees Red?   

OpenAI taking down the voice should be the end of it, right? Their defense is fairly strong. They can factually state that they used a different actress’s voice; an actress who had been selected for the role months before they met with Johanssen. They can claim that Altman’s tweet was accidental, unrelated, or taken out of context, and that neither the tweet nor his well-documented love of Her have anything to do with his business decisions. That should do it, right?

Here’s the problem: A plaintiff has won this case before.

The Ford Motor Company approached the singer and actress Bette Midler about performing one of her old songs in a commercial for the Mercury Sable. Midler said no. Instead of honoring her wishes, Ford proceeded to hire a voice impersonator to perform in Midler’s place. The ad ran in 1986. Midler filed suit soon after.

Midler argued that, though her image and likeness did not appear in the commercial, her voice was also a distinguishing characteristic and was protected from appropriation. She was not seeking damages for copyright infringement of the song itself, but for the use of her distinctive voice. Midler appealed after an initial loss in district court. A Ninth Circuit court of appeals overturned the loss and ruled that yes, a famous singer’s voice is distinctive to their person, and that it is unlawful to reproduce it or use it without their consent.

We’ll see how things turn out with Johanssen and OpenAI, but the outcome is not a foregone conclusion. OpenAI’s communications with Johanssen and Altman’s tweet imply that Johanssen may have a case.

Related Resources:

Pro Se Litigant Fined 10k for Filing AI-Generated Reply Brief (FindLaw’s Practice of Law)

Celebrated Writers File Copyright Lawsuit Against AI (FindLaw’s Federal Courts)

Silverman Goes for the Gold (FindLaw’s Practice of Law)

The post OpenAI Makes Scarlett See Red appeared first on .