Clicky

Robotics

What is the Uncanny Valley?

By March 3, 2017 No Comments

Because human interaction is such a core part of our nature, it makes sense that we want to replicate much of what makes this form of communication so rich. This is why a lot of technology interfaces are moving toward replicating human faces, bodies, and movements, in a bid to make us feel more comfortable with them.

In doing this, however, we’ve discovered a major problem. At a certain level of realism, human-like machines or images can cause a deep sense of revulsion or discomfort in people – something that’s been referred to as the “uncanny valley”.



Falling Into the Valley

Now more than ever, you are likely to run into simulations of human beings. While we don’t have realistic humanoid robots roaming the streets just yet, we do have computer-generated faces in our entertainment, virtual reality, augmented reality, and video games.

Repliee Q2

Who’s Idea Was It?

Back in 1970, professor Masahiro Mori identified this phenomenon; it made it into the English-speaking world around 1978. Mori hypothesized that as a robot (or other human simulation) became ever more realistic, we would find it more and more appealing. However, at some point it would become too creepy and we would feel revolted. But keep making it more realistic and eventually we would feel OK about it again. That dip in the amount of positive feeling people have towards a human simulation is the actual “valley”, because that is what it looks like on a graph.

Why Does it Happen?

The short answer is that we don’t exactly know. One thing that’s for sure is that humans are highly evolved to detect even the slightest abnormality in a human face. It’s one of the reasons it’s so incredibly hard to create a CG face that will fool people.

Take the latest, most cutting-edge facial CG from the film Star Wars, Rogue One as an example. The actor who played the character Moff Tarkin in the 70s films had passed away, but instead of recasting a new actor, Disney created “CG makeup” to reproduce his face. It’s an incredibly sophisticated and complex piece of CG, and for the most part it works. However, it is clear to any viewer at certain points that the face is not real.

To be sure, this sort of CG is getting better and better. One of the key problems is that we can’t always nail down why a particular face falls into the valley in the first place. A replica could be technically flawless and audiences will still see right through it – the power of our neurons proving to be too much to overcome.

Some of the most interesting research related to the uncanny valley was done in 2011. Researchers from the University of California put subjects into an fMRI machine, and monitored their brains while exposing them to uncanny valley stimuli. Based on the brain areas that lit up when subjects looked at uncanny valley subjects, it appears that we get that creeped-out feeling when the way something looks and moves violates our expectations. The set of sensory inputs that tell us something is human needs to match precisely, so if one aspect says “human” and another contradicts that, you get the uncanny valley. That’s not the final word on the causes of the uncanny valley, but it’s one of the most convincing arguments that I’ve seen so far.

Not Everyone is Convinced

While most people agree that the uncanny effect does exist, not everyone agrees that it neatly happens within the “valley”. Instead, they say that it can happen at various point along the continuum of realism. There’s also an idea that younger people, who are exposed more to CG, for example, may be inoculated against the effect because their life experiences have generated a different set of expectations for their brains. This suggests that the uncanny valley is at least partly learned.

Why Does it Matter?

You may be wondering why this relatively obscure hypothesis matters, but when you think about it, the uncanny valley may have profound implications for the future. For one thing, it may mean that people will reject some attempts at bringing realistic human-like systems into everyday society. Unless you can nail down that human look perfectly, you may as well not bother; rather, just stick with designs that are clearly not human and are more “robotic” or cartoonish.

It also matters because if you know about the uncanny valley it arms you with the knowledge to explain why you feel a particular way about a human-like simulation. One of the issues with the uncanny valley is that the people who feel the discomfort can’t nail down why they feel that way. This makes it more likely that they will attribute the feeling to another cause that has nothing to do with it. So we have a long way to go before we can start creating human simulacra that is 100% convincing. Since we ourselves may one day live through digital avatars or as uploaded minds inhabiting robotic platforms, it’s important that we come across as convincing to others. Getting a handle on the uncanny valley is key to that long-term goal.

Humanoid

It’s Uncanny

If you want to get a good first-hand experience of what the uncanny valley feels like, there are a few films I could “recommend” to really get you into that discomfort zone.

One of the first CG movies to attempt photo-realism was Final Fantasy, The Spirits Within. I’m a fan of Final Fantasy, but I have no idea why this film carried the franchise name. In any case, for its time the technology was mind-blowing, but the faces of the characters were creepy and uncanny.

One of the most notorious examples has to be The Polar Express, featuring a godawful digital recreation of Tom Hank’s face. This is probably the one film that brought the idea of the uncanny valley to the mainstream. Surprisingly, the makers of this horror went on to make an even creepier-looking film in the form of Mars Needs Moms, which is one of the biggest bombs in animation history.

The list goes on, but by inspecting just these few movies you’ll become intimately familiar with the revolting zone known as the uncanny valley.