When Silicon Valley entrepreneurs talk about how we will eventually interact with artificial intelligence, they often name-drop Her, the 2013 movie about a man who fell in love with his AI companion, Samantha.
Voiced by actress Scarlett Johansson, Samantha is a sassy and sensitive digital personality that succeeds in extracting Theodore from his melancholy through a series of heart-to-heart conversations.
But the crushing reality of what Theodore is dealing with is revealed late in the movie when Samantha tells him she is simultaneously interacting with 8316 other people and is in love with 641 of them. Eventually, she disconnects from Theodore to devote her computer power elsewhere.
OpenAI founder Sam Altman was so inspired by Her that he asked Johansson if he could use her voice for Sky, a conversational AI “with 100 unique personalities”. She turned him down, so he came up with a similar-sounding clone, later disabling it and apologising to Johansson who described the whole affair as “so disturbing”.
“This technology is coming like a thousand-foot wave,” she told The New York Times.
Now Meta’s Mark Zuckerberg is taking the concept behind Her and running with it. After spending around US$46 billion to date in an unsuccessful attempt to build the metaverse, an endless digital realm for us to hang out in together, he has pivoted to AI assistants.
One of them, Meta AI, became available here a couple of weeks ago as a standalone app alongside Facebook and Instagram, which both now have AI features built into them.
Meta AI is a “full duplex” assistant, which means that you can talk over the AI and it will change course and respond immediately based on what you are saying. That’s seen as crucial to mimicking the type of natural, human conversations that AI assistants of the last decade have lacked.
But Meta AI has little of Samantha’s charm. It made me cringe when it gave me the weather forecast in Fahrenheit rather than Celsius. All of the usual limitations of conversational AI were there – misunderstandings, irrelevant information served up and a general awkwardness that led to me reverting to typing in questions.
The rapid advances in conversational AI make Her-style relationships a realistic scenario within a few years.
“There are all these things that are better about physical connections when you can have them,” Zuckerberg recently told podcast host Dwarkesh Patel. “But the reality is that people just don’t have the connections, and they feel more alone a lot of the time than they would like.”
He’s right that loss of social connection is a real problem. In New Zealand, loneliness is highest among youth aged 15-24, according to the most recent StatsNZ survey that found 656,000 Kiwis are lonely all, some, or most of the time.
But the core flaw in Zuckerberg’s proposal is the assumption that AI can replicate the psychological benefits of human relationships. Research shows that while AI chatbots might temporarily alleviate feelings of isolation, they cannot provide the authentic empathy, shared experiences or physical presence crucial for meaningful connection.
Nor is Zuckerberg the right guy to serve up AI friends to the lonely. Careless People, the memoir of former Facebook executive and New Zealander Sarah Wynn-Williams, exposed the shocking ambivalence Zuckerberg displayed when shown evidence of the harm his platform was doing to people.
Like Samantha, the social media tycoon has demonstrated his willingness to extract as much as possible from us, mining the data yielded via our Facebook updates and Meta AI prompts to fine-tune the adverts he serves up to users.
Zuckerberg’s focus on AI friendships reduces the complex causes of loneliness to a problem of conversational bandwidth, a framing that serves Meta’s business interests more than the wellbeing of its users. In reality, Zuckerberg’s plan exemplifies the “loneliness economy” – technology that creates isolation then monetises artificial solutions.
