Why is some AI repulsive to us?
AKA: The Uncanny Valley
Since I mentioned it last week, I realized that the Uncanny Valley is a topic that deserves it’s own post. Obviously, I feel I have more to contribute than the hundreds of research papers and probable books already on the subject. You’re welcome.
For those of you unfamiliar with the term, it refers to the phenomenon that occurs when an artificial entity gets just close enough to looking real, but fails. It applies to both animation and robotics as both deal with replicating real things.
The best way to see it illustrated is through computer animated films in the last 20 years.
In case you never knew about it or never cared, The Final Fantasy movie was one of the most realistically animated movies when it came out, and it was terrible. It featured life-like human characters, and while the poor narrative didn’t help, it was also visually unappealing. Since then, most animation companies stopped focusing animating lifelike human characters. The new focus was either non-human characters or characterized humans. Think of the best animated movies since then; Wall-E, The Incredibles, and How to Train Your Dragon. Everyone except Square Enix, who keep cranking out Final Fantasy movies.
Uncanniness occurs because of all the little things that make it obvious the animation or AI is not human. Little facial ticks, the way it moves, and lips not syncing up to the sounds made. It goes from eerily similar to outright creepy very quickly. For a disturbingly thorough example, check out this post from 3 years ago. Also, note that a lot of these are from Japan.
There are a few theories for the reason this exists, but the easiest one is that we are exposed to humans all the time. We have evolved to read subtle ticks as emotional queues and we look at faces all the time. When we see a humanoid robot or animated character that doesn’t fit our model, it’s creepy. Part of it is an evolutionary mechanism, training us to avoid those that seem out of place or ill. Right now it’s one of our main hurdles in getting to human-robot interaction.
The same doesn’t happen when we see animated characters that are non-realistically human. These are safely on the upward slope before the peak and fall into the (Uncanny) Valley. Also, non-human characters aren’t creepy; think Wall-E and the dinosaurs from Jurassic Park. They look fantastic and engaging, but aren’t human at all. Probably because we don’t see dinosaurs and adorable robots every day. I’m sure some robot in the future will watch Wall-E and be creeped out, saying that something about the servos don’t move right.
So, the response to this, unless your Japanese, is to make things cuter and less human for public use. Research will still continue on making humanoid robots (and animation) more human-like, but it will be a while before they hit the mass-market. It’s one reason why screens have replaced faces in a lot of consumer focused robots. Simplifying expressions to just emojis keeps the robot well away from the Uncanny Valley. I’m not even going to get into sex robots. Like all things sex-related, if you can imagine it, it exists.
In short, robots are going to keep being creepy long before they get acceptable. If you want some more nightmare-fuel and proof of that, here’s one last look.
This story is part of The Codex, a collective of independent thought. Subscribe to our newsletter to get a weekly digest of our best stories and be sure to like and follow us on Facebook and Twitter.
