ChatGPT-4o, Scarlett Johansson And Missing The Point Of ‘Her’

One of the strangest tech stories in recent memories unfolded on Monday as OpenAI issued a statement claiming its ChatGPT-4o AI voice “Sky” was not meant to sound like Scarlett Johansson in the movie Her where she plays a conversational AI, but the company was taking it down anyway.

A few hours later, this led to Scarlett Johansson herself releasing a statement to say that OpenAI had indeed tried to hire her, twice, to use her voice for the project, but she turned it down. Then, she was “shocked and angered” the eventual voice sounded like her. As a rebuttal, OpenAI CEO Sam Altman said they’d hired the original actress before contacting Johansson, and that actress was not meant to sound like her. But they are taking it down out of “respect for Ms. Johansson.” Here’s his full statement:

“The voice of Sky is not Scarlett Johansson’s, and it was never intended to resemble hers. 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.”

Sam Altman, also tweeted the word “her,” the day the product launched.

It seems very apparent what happened here, despite Altman’s bumbling explanation. As the most famous media example of a convincing conversational AI, Altman and OpenAI wanted to reproduce that as literally as possible. They may have cast an actress that sounded like Johansson before they contacted her, but by contacting her at all they clearly wanted her voice, and it is impossible to believe Altman nor anyone else noticed the similarities between the actress’s voices before they unveiled “Sky.” That’s simply not credible and his explanation makes no sense.

Of course the second part of this story is that Altman, by holding up Her on a pedestal of something to strive for, has missed the point of that film. In it, a lonely man played by Joaquin Phoenix forms what he believes is a genuine relationship with an AI companion — Samantha, voiced by Johansson.

But ultimately he grows way, way too attached, forsaking other aspects of his life, and devoting himself to a non-person he can never be with. In the end, he’s heartbroken to discover that as an AI, Samantha has formed similar relationships with thousands of users, and he’s not special. She and other AIs, apparently having grown sentient, all disappear into the cloud at the end of the film. Meanwhile, Phoenix is left to start building up relationships with actual humans again.

All of this very reminds me of the whole “metaverse” pursuit, something that has not gone terribly well due to the fact that many points were missed about media like Snow Crash and Ready Player One, that the existence of a reality-escaping metaverse is dystopian, especially when it’s in service of advertising and corporate interests. Corporations like the very people making and monetizing the real-life versions that have failed to catch on.

But “Samantha-like” AIs are probably a more pressing psychological threat than the non-existent metaverse. And the fact that Altman wants to directly emulate the movie as much as possible, right down to the sultry voice of the actress that the main character fell in love with, is disconcerting (even before this, it did not escape many how “flirty” that AI was being during its demos).

This is a colossal own-goal from OpenAI here, but it reflects a larger point about how they are building something to emulate a dystopian sci-fi concept that in its very source material, does more harm than good. This happens in the tech world frequently, but never quite so literally as this.

Follow me on Twitter, Threads, YouTube, and Instagram.

Pick up my sci-fi novels the Herokiller series and The Earthborn Trilogy.