top of page
Alex Iwanoff

A New AI-Era of the Uncanny Valley

The uncanny valley— a term that has haunted roboticists and VFX artists for decades —is experiencing a resurgence due to artificial intelligence video generators. Coined by Japanese roboticist Masahiro Mori in 1970 in an essay, it describes the unsettling sensation we feel when encountering humanoid figures that are almost, but not quite, lifelike.


By Dietmar Höpfl - shockfactor_ai

Minor imperfections in appearance or movement trigger discomfort or even revulsion, tapping into deep psychological responses. Why it happens? We actually don’t really know. Apparently, some even question if this is a scientific concept or not, but research is being conducted in this area.


ROOTS OF THE UNCANNY VALLEY

Mori’s hypothesis stems from human psychology and our innate responses to realism in representation. While exact reasons for this discomfort are debated, researchers speculate it could be due to:


  1. Violation of Expectation: Subtle imperfections in a hyper-realistic face or movement disrupt our expectations, creating a jarring effect.

  2. Evolutionary Psychology: Some theories suggest our unease arise from our ability to distinguish between alive and dead, or healthy and diseased individuals—a potential survival mechanism.

  3. Empathy Gap: We may struggle to emotionally connect with near-perfect imitations that lack a true human essence (this seems to be the case with AI video generators).

“The sense of eeriness is probably a form of instinct that protects us from proximal, rather than distal, sources of danger. Proximal sources of danger are corpses, members of different species, and other entities we can closely approach. Distal sources of danger include windstorms and floods”, Masahiro Mori

AI UNCANNY VALLEY

It’s well-known that CGI has long struggled with the uncanny valley. But, today, we're turning our attention to AI video generators and exploring why they can be so particularly creepy—and whether that's necessarily a bad thing.


First, keep in mind that recreating humans is one of the biggest challenges in visual effects. Our brains are finely tuned to recognize the subtleties of human movement and expression. Even minor inconsistencies stand out and when technology tries very hard to mimic us but doesn't quite succeed, it leaves us with that uncomfortable feeling.


Usually, this has been seen as a problem to overcome—a sign that the technology isn't advanced enough (or that, you know, “CGI is ruining movies”!). Therefore, in VFX, we try to stray away as much from humans as we can, but sometimes… it’s just what the director wants. This gives birth to very unfortunate results, like cat-hybrids or bringing back the dead to life for a movie. A term coined as “digital necromancy”, according to Futurism.


But I diverse. Back with AI. The tool is very good at making us think, at first sight, that it looks human. However, when you look closely, you can see six fingers or slight malformations in the images it creates. But it’s when it tries to generate movement that it becomes really bizarre, going places we’ve never really seen.


This happens because AI lacks true spatial awareness. It doesn't comprehend physical space as we do; instead, it generates content based on patterns learned from vast datasets. This can result in inconsistencies, distorted perspectives or unnatural movements.


To overcome these challenges, developers are working on enhancing AI's spatial awareness and motion generation capabilities. For instance, in robotics, they're training robots in digital simulations (or digital twins) featuring stairs, other robots, metahumans and obstacles, so they learn how to interact with external elements and won't be a danger when put in the real world.


Similarly, in the realm of filmmaking, the aim is to imitate reality, which is why CGI has (mostly) evolved to achieve realistic results. The same is going to happen with AI video generators, with companies looking to cut costs and create cheaper and easier ads to produce.


However, I believe that AI's most compelling work emerges when its uncanny and weird outputs are embraced, especially in the horror genre (but not only!). Many artists on social media are leveraging these eerie qualities to create unsettling videos that we can't help but keep watching. These are the same artists we've included here as examples for our article.


These uncanny visuals produced with AI challenge our perceptions of reality and elicit strong emotions. This shift raises a question: Is the uncanny valley necessarily a bad thing?


While many still strive to eliminate it, embracing the uncanny opens up new possibilities. As an art form, it allows creators to explore places where we would not have been able to go on our own—a realm between reality and artificiality, engaging audiences in ways that are new, fascinating and definitely disturbing.


What do you think about it?


23 views0 comments

Recent Posts

See All

Kommentare


bottom of page