There’s a story that’s been with us for thousands of years, told and retold across space and time. It goes something like this: A man carves a statue of the perfect woman, then falls in love with it. With a bit of divine intervention, this artificial being is brought to life.

Nina Beguš, a researcher at the University of California, Berkeley with a PhD in comparative literature, sees this narrative template—which in Western literature has roots in the ancient Greek myth of Pygmalion—popping up in a surprising place: Siri, ChatGPT, and the other so-called “AI” tools taking the world by storm. As Beguš argues in her new book Artificial Humanities: A Fictional Perspective on Language in AI, the Pygmalion myth is one of the main tropes technology developers fall back on when designing new AI systems. It’s a big reason these systems are so often feminine, submissive, syncophantic, and even sexy.

To Beguš, the power of fiction to shape emerging AI technology should not be dismissed. Rather, it calls for an entirely new field of study, the Artificial Humanities, which brings philosophers, literary scholars, historians, gender studies experts, and other humanists to the table of AI development. Without their participation, Beguš worries that the most obvious cultural memes, from ancient mythology to Hollywood science fiction—killer robots, sexy voice assistants, even the idea that humanity represents the pinnacle of intelligence—will limit our collective imagination of what AI can be. “In order to shed off the Pygmalionesque ballast we all implicitly foster, we need the humanities at the table,” Beguš writes.

The Science of Fiction spoke with Beguš about the new field of Artificial Humanities, the feedback loop between fiction and AI development, and how better stories can help us build more useful, ethical, and interesting machines. 

The conversation has been edited for length and clarity.

Tell me about your academic background and how it came to intersect with AI?

When I was doing my masters in continental Europe, I saw all these breakthroughs happening in technology. It was ImageNet at the time, and Siri. And I was looking at those breakthroughs and reading literature and saying, ‘Wow, the overlap is so uncanny, what’s going on here?’ 

Now it might seem obvious to say “Yes, the influence between the world of fiction and the imaginary, the world of ideas, and the world of practice and technology building is definitely there, and it goes both ways.” But at that time, it wasn’t such a given. Especially since people tend to see fiction as not doing any real work in the actual world.

But this is so much not the case. And I think now with AI, that’s really obvious. It’s clear there’s this unprecedented power of fiction, especially in technological spaces, doing actual work on how we build AI and also how we relate to these products.

And so I was thinking about all these stories, and I framed it around the Pygmalion myth. Which is such a bizarre myth: you’re building something, then you fall in love with it, you start to take care of it and there’s this romantic, erotic aspect. It just seemed so bizarre, but so present.

And I looked into different mythologies and folk stories, and you find it everywhere. You find it in Europe, North Africa, on the Silk Road, in Native American folk stories. It’s such an archetypal myth. So in my book, I pretty much limited myself to the Pygmalion myth.

I think the first time that really struck me that something was going on was Siri. It was a woman—all of the early [voice] assistants were women. It had this subservient nature of a helpful assistant. It was trying to be a product that builds rapport with you.

Something that today, 15 years later, we still see with large language models and a lot of other products. We kind of can’t break out of this persona.