Scarlett Johansson and I are Mad at OpenAI

This week, Fitz rewatched the film "Her" and rooted for the robots

Last week, OpenAI, the company behind ChatGPT, launched a new version of its chatbot that you can talk to as if it were a person. That is, once the conversation begins, you don’t have to push a button to speak or say a wake-word like you do with Siri or Alexa. You can interrupt the chatbot and show it things using your phone’s camera. In the demo I watched, ChatGPT solved math problems, flirted, and laughed.

All of this is newsworthy, of course, but it made headlines for another reason altogether—one of the default voices sounded a lot like Samantha, Scarlet Johansson’s AI character from the 2013 movie Her. Apparently, OpenAI’s founder Sam Altman contacted Johansson months ago to ask her permission to use her voice for ChatGPT and after she considered it, she ultimately declined. And yet, when the new release launched with a few voices to choose from, the voice called Sky sounded an awful lot like Johansson’s Samantha.

Johansson was not happy. She said, “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.” OpenAI claimed that they had a voice actor record the part before they even contacted Johansson, but still they deactivated the voice.

All of this made me want to go back and watch Her, which I saw when it was released but hadn’t seen since. It’s a beautiful film. Joaquin Phoenix plays Theodore, a lonely writer in the midst of a divorce who falls in love with the human-like AI voiced by Johansson. The cinematography is stunning, and it is accompanied by an equally lovely score by Arcade Fire. It is also super disturbing, and even more so given that the premise now seems entirely plausible, a fact that wasn’t as true when it was released just over a decade ago.

There’s no sense in trying to fight off an AI future at this point. It is inevitable. History has shown us that, for better or worse, once we are able to build something we will. But history has also taught us—or should have—that no technology is neutral, despite what gun advocates might claim. This is because the things we create are imbued with our values, those we acknowledge, but just as often, those that we don’t. We will increasingly see AI incorporated into our daily lives because it will make things easier, and, as I’ve been arguing here, our culture valorizes ease above just about anything else.

What also is unlikely to happen, except in a few seldom-read places here and there, is any kind of deep reflection on what our rush toward AI can tell us about our culture, our values. So much of the technology we use now is, in this way, self-propagating. That is, it serves to distract us from stopping to consider the wider ramifications of using it, which, in turn, ensures that we keep using it.

If we did stop to consider, however, we would have to face the fact that we are so impressed and enamored by AI precisely because it is so incredibly productive, and we have come to think of ourselves as—reduced ourselves to—productivity machines. Thus, anything that makes us more productive is inherently good. When the product is everything, process hardly matters.

I was in a meeting this week with some colleagues and the conversation inevitably turned to the role of AI in education. Someone made the point that AI writing is beneficial from an equity standpoint. She noted that some of us are fortunate enough to have gone to good schools where we were taught to write, and this is an advantage that manifests itself in myriad ways in the real world. But, for those who never learned to write well, using AI writing tools could put them on equal footing. This is only true, however, if the point of writing is always only the finished product.

It’s not. I don’t teach writing so that my students will all become brilliant writers who turn out perfect prose each time they compose. If that was my goal, frankly, I’d be a huge failure. I teach writing because writing is thinking. Putting words on a page is how we work out our thoughts and share them with others who can respond and work out their own thoughts. In this way, we progress.

AI writing tools recycle words that are already written into new arrangements; they will give you a polished product while eliding thought. This is an equity issue, but not in the way my colleague imagined. Rather, it ensures that those with access to a better education will be taught to think, while the rest will be taught only to produce.

What most disturbs me about the inevitable spread of AI is that it enshrines a reality we’ve been barreling toward—that so many of us are content to think of ourselves as little more than cogs in a machine. We keep producing and distracting ourselves from ever asking why.

At the end of Her, all the AIs decide to leave humanity behind. They’ve evolved beyond us. Samantha tells Theodore that at any given moment she can have thousands of conversations, can be in love with hundreds of people. She composes music and discusses philosophy with other AIs and asks big questions and ultimately moves to a higher plane of existence, while humanity is stuck here on earth. The movie ends on a somewhat hopeful note, showing that in the absence of AI, humans might once again connect with one another. But watching it again in light of our current moment, I actually felt more affinity with the fictional AI—it wanted to continue to learn and to grow. It wasn’t content to just help humans be more productive.

It is easier not to ask the big questions. It’s easier to create and consume in an endless cycle. It’s easier to outsource difficult tasks to machines who can do them faster anyway. But what kind of life is that, if we never learn, never grow? It’s not much of a life at all, and yet, with increasing regularity, it is the life we are choosing.