If you've been snoozing through the Scarlett Johansson-OpenAI drama this week, let me quickly catch you up:
A woman rejected a man's request, twice, and he responded like a child — demonstrating just one of the reasons the world needs to be skeptical of the people and companies profiting from their own breathless AI hype.
Bear with me for the slightly longer version.
Back in September, Scarlett Johansson, who played the hauntingly complex AI assistant in the 2013 Spike Jonze film "Her," got a request from OpenAI's CEO, Sam Altman. He wanted to hire Johansson to voice his company's newest ChatGPT model, "Sky."
She said no.
Several months later, just days before the product launched, he asked her to reconsider.
But before she could say no a second time, OpenAI released the new model with a voice that sounded strikingly similar to Johansson's. The AI's debut was both mind-blowing and cringe-y. Mind-blowing because of the bot's apparent sophistication and human-like qualities, and cringe-worthy given how uncomfortably flirty it sounded, like a caricature of a computer developer's fantasy.
Johansson quickly lawyered up, saying Monday she was "shocked, angered and in disbelief" that Altman would use a voice "so eerily similar" to her own.
The company, which hit pause on the update after the legal threat from Johansson, said the voice "belongs to a different professional actress using her own natural speaking voice" and was not intended to resemble Johansson's.
Altman — who is on the record about his love of the Spike Jonze film, and even tweeted the word "her" on the day of the launch — apologized to Johansson on Monday.
Safety takes a backseat
The rift reflects a broader anxiety among artists, academics and even some AI pioneers over the speed at which tech companies are building and releasing AI tools to the public, with seemingly little regard for intellectual property and safety concerns.
It also points to one of the original sins of AI that its developers have yet to resolve: that all of these products are being greenlit, funded and built by the 0.01% of Silicon Valley — a largely young, white and male cohort, whose natural human biases are getting baked in to the AI in ways that even they don't fully understand, or are aware of.
OpenAI was forced to confront some of those concerns late last week, after two prominent employees left the company.
One, Jan Leike, wrote on X that the company's "safety culture and processes have taken a backseat to shiny products." The other, OpenAI's co-founder and chief scientist Ilya Sutskever, announced he was leaving to work on another project, without disclosing details, though his departure is especially notable given his key role in (briefly, dramatically) firing Altman last year over concerns that Altman was pushing AI technology "too far, too fast."
Altman even acknowledged Leike's point, saying: "He's right we have a lot more to do; we are committed to doing it."
That's the kind of diplomatic candor that's helped make Altman, a 39-year-old billionaire, the face of "responsible AI." He is a thoughtful, charismatic figure, and he really wants us to trust that he's looking out for regular folks as he mints money and races to beat Google and Meta to the AI mountaintop.
I'd be curious to ask him, though, when was the last time he watched "Her" — one of his favorite movies that he called "incredibly prophetic" — all the way to the end.
Because part of what makes the film great is how Johansson's warm and empathetic voice lulls the main character (and the audience, ultimately) into an almost hypnotic state of infatuation. Maybe the ease of human-AI companionship is acceptable, or even preferable, to the messiness of real love, we wonder as we watch Joaquin Phoenix's character walk alone on a crowded beach, the camera of his phone peeking out of his shirt pocket, smiling because he's so happy with his AI lover.
Of course, by the end of the movie, the AI's have decamped to some other plane of existence, leaving the humans, with all their messy human needs, to take care of one another once again.
"Being friends with AI will be so much easier than forging bonds with human beings," wrote Wired editor Brian Barrett in a recent essay about the movie. "That doesn't mean it's better. Sometimes it's much worse."
Comments
Post a Comment