It took a lot of doing, but I have successfully extracted the essence of uncanny valley. Or, discovered the minimum unit of uncanny valley.

2 years ago, we were researching emotional responses to AI generated faces.

Generate a face. Record emotional response. Tweak generated face on one parameter by a bit. Record response.

Do this 1,600 times.

At first, the results were what you'd expect. We published lots of articles about facial symmetries, about how asymmetry is necessary for attractiveness, etc. You probably read about some of them.

The really interesting research began when we wondered if we could treat the anterior fusiform gyrus adversarially.

We'd hold the face constant, and only allow a 1cm diameter area to change programmatically.

Thousands of college students sat for endless hours, wired up to EEG's and gaze tracking apparatus, staring at a single face with a small multicolored dot between its eyes.

What we'd hoped to find was a bug in how humans perceive faces. Would such a tiny feature be able to statistically significantly alter how faces are perceived?

Our first result was breathtaking. A pattern was found that reliably made faces seem significantly more androgynous, across a statistically significant number of subjects

A 1cm sticker with this pattern, would instantly make you seem androgynous. Breathtaking.

Our next result was even more incredible: a pattern which appears to induce prosopagnosia, or face blindness.

Subjects could identify that they were looking at a face, describe each feature, but could not recognize it whatsoever.

A 1cm dot would make you unrecognizable.

Again: if you were to place a sticker with this pattern on your face, you would be unrecognizable to humans.

Absolutely incredible. We were like dogs finally catching a car, we had no idea what to do with a result this amazing.

We immediately focused our research onto what we called "visually induced agnosias."

Almost immediately afterward, we found a pattern that induces Fregoli Delusion.

A face with this pattern would be identifiable as a face, recognized as a specific person's face, but not recognized as *that person*. Subjects are convinced it is an imposter in disguise.

This was even more incredible.

Basically: until now, all our generated patterns targeted the anterior fusiform gyrus, the part of the brain that deals with facial recognition.

Fregoli's delusion, however, arises from interactions with the emotional processing centers.

Immediately after we found a Fregoli VIA, we stumbled upon a Capgras VIA.

Capgra's delusion, like Fregoli's, involves a disconnect between emotional relational processing and lower-level facial recognition.

A person with a 1cm sticker dot of this pattern would be identifiable, recognizable, but you would be absolutely, inalienably convinced they were someone *else* disguised as that person.

It'd be literally as self-evident as the fact that they have a nose and two eyes.

Now, I was not very clear in distinguishing Fregoli's from Capgras.

The Fregoli VIA we discovered, when placed on multiple faces, caused each face to be identified as a known face wearing a clever disguise.

The Capgras VIA, causes a face to be identified as an unknown imposter.

The first Fregoli VIA we discovered, made subjects universally convinced that each face it was applied to was Robin Williams in disguise.

Tweaking a single pixel was enough to shift that imposter's identity to anyone from Margaret Thatcher to Barack Obama.

The first Capgras VIA we discovered, universally made subjects convinced someone was impersonating the person, but none could identify who this supposed impersonator is.

We all immersed ourselves in the work, just these discoveries alone were enough to ensure decades of research. Each alone was Nobel worthy.

We gradually discovered more Delusional misidentification Syndrome VIA's.

We generated a pattern that induces Christodoulou's Delusion: subjects recognize the face as *their own* but worn by a doppelganger.

A 1cm sticker with this Christodoulou VIA on a face would cause you to recognize it as your own face, regardless of how that face differs from your own by gender, race, age, etc.

You'd be convinced it is your twin.

We discovered a whole family of intermetamorphic VIA's: patterns which cause subjects to perceive a person as transforming, literally, into another person.

Subjects perceive it as the displayed person "dissolving into" a different person.

The first intermetamorphic VIA pattern we discovered, caused subjects to be convinced that displayed persons were being transformed into Danny DeVito.

Thusfar, each VIA had been face-oriented.

When we discovered a VIA that induced reduplicative paramnesia, we were completely taken aback.

It took 2 months to even figure out what it was doing.

Reduplicative paramnesia is the belief that a location has been duplicated.

Normally it is seen in patients with severe bilaterial frontal lobe lesions.

A person with reduplicative paramnesia may be convinced, fully and unshakeably, that their room appears to be the same but is merely a staged location elsewhere.

Normally, reduplicative paramnesia does not extend to the people within a location. Patients are bewildered, for instance, that their doctor does not realize she is actually in a fascimile of the hospital on another continent.

Our VIA version of reduplicative paramnesia, however, extends to everything within the environment, people included.

23 subjects violently attacked our technicians before we even grasped what was actually happening.

After we implemented physical restraints and mild sedatives, subjects reported that they suddenly and entirely believed they'd been teleported, or otherwise instantaneously transported, into a parallel reality.

None reported that they believed they'd been transported into a fascimile of the room they'd been in moments before. None reported actually existing locations.

All, without exception, used the phrases "parallel reality" or "nightmare."

Each reported feelings of intense terror and overwhelming anxiety. None could specify what caused such powerful reactions.

We tried dialling in exactly how their environmental perception was altered.

How did they perceive faces? Inanimate objects? Colors? Angles? Geometric patterns?

Very quickly, we realized this was a whole new phenomenon, far beyond mere facial recognition.

This pattern affected everything, not just facial recognition.

Subjects reported powerful revulsion at their own bodies.

This pattern even had a multimodal effect.

Subjects reported their own voices to be unsettling, tactile sensations elicited powerful revulsion, etc.

Basically: a 1cm dot of this pattern would make you fully and unassailably perceive yourself as having been mentally transported into a parallel reality where literally *everything* is unsettling. From the sensation of your own hair to the sight of a right angle.

This phenomenon became the new focus of research. We'd uncovered a bug in human sensory perception at so fundamental a level that it affected *every sense* and *every perception*.

We developed ketamine and MDMA cocktails to reduce the PTSD that even brief exposure to this VIA caused.

On average, 30 seconds of exposure to this VIA was enough to induce PTSD without such precautions.

We started fine-tuning this VIA, one pixel at a time, ever so slowly exploring its boundaries.

We discovered that we could "dial back" the nightmarishness of the VIA almost, but not quite, indefinitely.

We could dilute it from a debilitatingly fearful experience, into a just-barely perceptible unease.

We had discovered the unit of uncanny valley. The minimum necessary visual stimulus sufficient to induce generalized unease, fear, and anxiety.

It is a multicolored circular dot, 1cm in diameter, capable of transporting you quite literally into a nightmare.

It is a mere 100 pixel difference from another almost identical 1cm multicolored dot which, to all accounts, transports you into hell.

And it is here our research has now been focused since January.

You would not *believe* what we've done in the past 6 months.