Silicon Valley bigwigs aren’t very imaginative. They know a few things about culture — A16Z’s Ben Horowitz likes rap, Elon Musk likes Spaceballs — but for the most part these middle managers have been heads-down building shareholder value, so they don’t get out to the mall much.
But when one of them gets a bug up their behind about a specific bit of cultural ephemera, watch out. They’re going to steal it.
Sam Altman, CEO of OpenAI, had such a bug up there a few weeks ago when he tweeted the single word, “her.” The Tweet seemed innocuous at the time. It was in reference to the 2013 movie Her featuring Joaquin Phoenix and Scarlett Johansson about an AI assistant that eventually goes supernova. It’s a love story — Phoenix is a nebbishy guy that every Silicon Valley startup is trying to sell to but not hire — and it ends badly for him.
The movie is an admonition against humanizing tech. But Altman, being tone deaf, thought it was instruction manual. The result? He asked Johansson if they could use her voice in the latest version of OpenAI’s text-to-speech generator, she refused, and he used someone who sounded like her. Let’s call it the first Realfake— a copy of a human that is aimed at making money, not titillating.
Johansson released an open letter in reaction to the controversy, first reported by NPR reporter Bobby Allyn.
“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 AI. He said he felt that my voice would be comforting to people,” she wrote.
Whether or not Sky sounds like ScarJo, the primary problem involves the use of cultural artifacts in a technological context. What Altman and his team did was humanize a tech product through the use of cultural signifiers and in doing so they made a comforting shell. But they stole the voice, just as they stole billions of web pages, millions of videos, and hundreds of thousands of books. They consumed the corpus of the web and spit it back out at us in a novel form. The results, as we’ll probably learn over the next few years, are disastrous.
We love the promise of AI, just not the current implementation. As creatives, we must ask hard questions about rights and access. Does Sky really sound like ScarJo? Not really, but the suggestion that they are the same hovers in the air like a stink. AI businesspeople — not researchers — are so desperate to escape the uncanny valley and sell us new products that they are completely ignoring the legal and societal implications of their work.
In the end, Altman stole Johansson’s voice because it was easy and funny and clever, things that Altman and his ilk aren’t in real life. They’ll talk a big game about protecting us but in the end all they want is a robot that will make you fall in love with it and, eventually, them.
Leave a Reply