One of the factors contributing to fear of AI is the Frankenstein Complex.1 The term, coined by sci-fi writer Isaac Asimov2, originally described fear of the “mechanical man” in the science fiction of old. The complex is named for the young scientist, Dr. Victor Frankenstein, protagonist of Mary Shelley’s 1818 novel Frankenstein, or The Modern Prometheus. He stitches human body parts together to create a monster. There is no AI in the story; he brings the monster to life based on an intelligence that was assumed to be resident in biology. “Frankenstein’s monster” later came to be called simply “Frankenstein.”
Thomas Edison first filmed the story as a silent movie in 1910 but Boris Karloff’s depiction of the monster in the 1932 Universal classic motion picture Frankenstein became, as you will see (below), the iconic one.
Today’s movie monsters are a lot scarier than those of the 1930s. But even today, Karloff’s Frankenstein monster makes your skin crawl. Why? He moves clumsily, seemingly in slow motion. Even on crutches, you could avoid him. A blow from Mike Tyson would probably take him out. A swift-striking bobcat or teased alligator sinking teeth into your hand would be much more dangerous. Yet we get chills just looking at Karloff’s monster. Why?
The Uncanny Valley Hypothesis3, named after a dip in a regression curve, posits a correlation between the degree of an object’s resemblance to a human being and our emotional response to it. We can see it as funny and sometimes creepy.
Our baby brains are prewired to recognize faces so we can discern Mommy from others. Our sensitivity to small changes in facial expression is uncanny. Cartoonists use this fact in their art. In the following figure, just changing the eyebrows and the mouth displays a plethora of different emotions. Each of the three columns of faces has a different mouth shape: flat, smile, and frown. Each of the three rows has different eyebrows: pointed down, pointed up, and flat. And each of the nine combinations of these simple features is strikingly different:
We see faces everywhere we look and we assign personality to what we see. The Uncanny Valley Hypothesis explains why we see the humor in the following images where we immediately interpret patterns as faces. 4
The emotional response is not always one of amusement. A feeling of eeriness results when viewing the faces below. The spoon-sucking demon in your coffee and the evil drain grate make us squirm. The reflection of a demon-possessed stove in the stainless steel kettle is the stuff of a Stephen King novel:
Those who market AI rely on the Uncanny Valley Hypothesis to hype their wares by packaging them in human-like containers. Think of Amazon’s ubiquitous Alexa chatbot’s voice. To make chatbots more appealing, AI is often embedded in a human-appearing body with human social features. Sophia the Robot, shown below, has her own Facebook page. Her speech is augmented by facial expressions using small feature changes akin to those used by cartoonists. Such seductive optics makes the Sophia AI more human-like than her voice alone would do. But the human container in which Sophia lives has little to do with her chatbot AI. If you want to brave the ads, there are many interesting videos of Sophia on YouTube.
My personal Frankenstein Complex reaction would diminish if Sophia wore a wig. I suspect her marketers want me to feel a pinch of the Complex and thus purposely left off the wig.
Contrary to what we might expect from the hype, the AI optics representing humans today is not a big improvement over the robotic animations in The Hall of Presidents that Disney World has offered starting in 1971. Although they have been updated and improved, the US Presidential robots are hardly chatbots. They perform with fixed prerecorded audio and preprogrammed gestures. But visually, they are similar to Sophia the robot.
The obsession with AI packaging over AI content is evident in the news media excitement about a Buddhist robot who delivers messages to the faithful. Unlike Sophia, the robot features little AI. The messages appear prerecorded and are not the product of AI. It is technologically closer to Disney’sthe US presidents when the exhibit opened in 1971.
The Frankenstein Complex and the Uncanny Valley Hypothesis contribute to fears of AI. Anything not human that looks like a human or displays almost human characteristics is scary. Fear can grow from incomplete knowledge and our predisposition to associating the almost-human with the bizarre. There are other factors that contribute to the fear of AI but the Frankenstein Complex looks to be a major one.
I suspect AI representations will improve to the point of being visually indistinguishable from humans if they are not closely examined. Currently, though, seamless representation of the human form is not well developed. The marketers of AI take advantage wonderfully of their immature technology to pique our emotional interest, relying on the Uncanny Valley Hypothesis. Today, when the competition for audence attention has never been more intense, making things look “almost human” is a successful technique.
And I finally have a diagnosis for some part of my emotional reaction to AI. I understand AI. But my skin still crawls while watching Sophia the Robot and other AI robots embellished with seductive human-appearing optics. I suffer from a case of the Frankenstein Complex.
1 Rushing, Janice Hocker, and Thomas S. Frentz. “The Frankenstein myth in contemporary cinema.” Critical Studies in Media Communication 6, no. 1 (1989): 61-80; Lehman-Wilzig, Sam N. “Frankenstein unbound: towards a legal definition of artificial intelligence.” Futures 13, no. 6 (1981): 442-457.
2 McCauley, Lee. “The Frankenstein complex and Asimov’s three laws.” University of Memphis (2007).
3 Mathur, Maya B., and David B. Reichling. “Navigating a social world with robot partners: A quantitative cartography of the Uncanny Valley.” Cognition 146 (2016): 22-32. (open access)
4 Face images courtesy of TheChortle.com
Also by Robert J. Marks: Random thoughts on recent AI headlines