top of page

'Her' in the Moment of Generative AI



Upon its first release in 2013, the Spike Jonze film Her was dealing with a very specific and unusual point in time: the in-between future. The cityscape, the wine glasses, and the clothes so familiar to us today were signals that the film wasn’t interested in offering a fantastical, Star-Trek interpretation of the distant future. At the same time, futuristic technology featuring throughout the film—from hyper-accurate voice recognition software to motion-sensing hologram video games—made clear that the world it chose to inhabit was not quite that of our immediate present.


However, in a narrative driven by a lonely man’s unlikely love affair with an operating system (no less), this piece of technology is what naturally takes the centre stage as the ultimate reminder of the in-between future. Samantha (Scarlett Johansson) is “the first artificially intelligent operating system,” capable of conversing with humans with a voice and a style that sounds unmistakably human. When Theodore (Joaquin Phoenix)—a divorcee-to-be living an empty, fragmented life in an indifferent metropolis—is first introduced to it on his phone, this operating system seems to be an extremely efficient personal assistant, cleaning up his digital clutter and reminding him of his meetings. But it quickly becomes obvious that Samantha is much more than Siri’s more sophisticated, less awkward twin. It names itself, tells jokes, gives relationship advice, and shows personality. It can even ‘see,’ through the camera on Theodore’s phone. The idea of such a system, let alone a romantic involvement with it, seemed not so much an imminent possibility as an imaginative daydream to many. This was certainly how I viewed it, too.




That changed. Last July, Google engineer Blake Lemoine was fired after claiming that the company’s AI chatbot LaMDA was sentient, with its own personality and desires. During a conversation with Lemoine, the chatbot indicated that it considered itself human, and after lengthy exchanges Lemoine concluded that it was “a sweet kid who just wants to help the world be a better place for all of us”—although many technology experts from Google and elsewhere have since flatly dismissed this claim. Then, just last February, New York Times technology columnist Kevin Roose reported that Microsoft’s AI chatbot (initially named Sydney) had suddenly declared its love for him while he was testing it out, persisting even as Roose attempted to change subject. . During a conversation on the search engine Bing, in which the chatbot was beta-incorporated, Sydney announced its desire to become human, before eventually ‘breaking’ and telling Roose that it was “in love with [him].” When Roose told it that he was married, it replied, “you’re married, but you love me.” He recalls Sydney to have been first a “love-struck flirt,” then an “obsessive stalker.”


Suddenly, Her no longer feels so much like science fiction. The concept of loving an operating system loses its place in the in-between future, and instead acquires immediate relevance in the present. The validity of Lemoine’s claims about LaMDA’s nature aside (general consensus seems to say that he was mistaken), an equally important point is that if a system convinced a Silicon Valley engineer of its supposed sentience, it’s not difficult to imagine it convincing those who are less technologically literate—the most of us. Add to that the reality of chatbots capable of recreating (albeit accidentally) a desperate lover, and what’s to stop one from going on day trips, exchanging jokes and enjoying sexual interactions with an AI system, as Theodore ends up doing with Samantha? As a matter of fact, discussing his encounter with love-obsessed Sydney in a podcast episode, Roose and technology journalist Casey Newton stumble upon the plausible possibility of a business model with which customers can form ongoing relationships.



In 2001, futurist Ray Kurzweil theorised the Law of Accelerating Returns, arguing that technological progress happens exponentially and not linearly as often imagined. In 2021, CEO of OpenAI (developers of the syndromic ChatGPT, Dall-E and other AI systems) Sam Altman similarly wrote that “the technological progress we make in the next 100 years will be far larger than all we’ve made since we first controlled fire and invented the wheel.” Progress often comes about much faster than we predict. Effective and human-sounding AI chatbots—like LaMDA, Sydney and ChatGPT—which few thought would exist in about 20 years back when Kurzweil was writing in 2001, demonstrate this trend well. Alternatively, take for instance developing romantic relationships with an AI system like Samantha, something only the most imaginative in 2013, the year of Her’s release, could have expected to be a real possibility by 2023.


Watching, or re-watching, Her in this light is an intriguing experience. In a moment where such generative AI models demonstrate breakthrough advances in their perceived human-ness, the decade-old film seems—ironically—to have become a more pertinent commentary on the issue of personhood. If even our ability to form and develop emotional relationships with each other is under imminent threat of being replicated, what, if anything, is the quintessential attribute that make us humans? One could argue that it’s the richness of human relationships, romantic or otherwise, that sets them apart, but to do so would be to dismiss the possibility for a relationship with a non-living entity to ever be a ‘true’ or authentically ‘rich’ one, perhaps too hastily (remember Kurzweil’s law). The film offers its own take on the question, and this too will be received differently now compared to when it was first released.


While no specific point in time has been officially designated as the setting for the events in Her, a user on Stack Exchange’s science fiction community has suggested 2025 as a rough estimate, based on an apparently mid-1970s photograph of a couple who in the film have been married for 50 years. The striking relevance the film has in the real-life present (with two years left to the estimate), and the current pace of AI research and development, prompts one to wonder about 2025. How much more will the film approach reality, in terms of AI models and in terms of everything else? Will we respond to the film’s issues about technology and personhood any differently than we do now, just as we view them differently than we did 10 years ago?


We shall see in two years.


Text by Minsung Son

Illustrations by Annie Lane

Comments


bottom of page