Where it sits
Affinity rises with realism, plunges just before the end, then recovers. The dip is very close to the finish line.
Tell it apart
| Aspect | Uncanny valley | AI slop |
|---|---|---|
| Judged by | Perception | Process |
| Requires craft | Yes | No |
| Named | 1970 | 2024 |
| Feeling | Dread | Contempt |
Slop is judged by effort and volume; the uncanny valley is a perceptual response to near-miss humanity. A beautifully crafted android can sit deep in the valley. Most slop never gets close enough to fall in.
What it doesn't mean
His running example was a PROSTHETIC HAND, and his central claim is about MOVEMENT: motion changes the shape of the graph by amplifying the peaks and the valleys. A still image is the weakest case for the effect, not the strongest.
There are ongoing challenges — that the effect may be generational, culturally variable, or an artefact of information processing rather than one unified phenomenon. It is an extremely productive hypothesis rather than an established law.
What it really means
The single most useful thing about Mori's 1970 essay is the example he actually used, because it is not the one anyone expects.
Not a robot face. Not animation. A prosthetic hand.
A prosthetic hand can be crude — obviously artificial, clearly a tool — and people are fine with it. Make it realistic: skin tone, fingernails, veins, the right weight. Now, when someone shakes it, they recoil. Mori's point is that the improvement made things worse — and when he asks why we should be equipped with the reaction at all, his own answer is that it is an integral part of our instinct for self-preservation.
Affinity rises with realism. Then, very close to the end, it plunges. Then it recovers on the far side. Hence valley — and it sits almost at the finish line, which is why getting better at this can make things dramatically worse before it makes them better.
The part that gets dropped is his claim about movement. Mori argued that motion amplifies both the peaks and the valley — it deepens the dip. Which is why a still render of a digital face can look perfectly acceptable and the same face in motion is unbearable. A still image is the weakest case for the effect, not the strongest, and almost every modern discussion of the uncanny valley uses still images.
One more thing worth knowing, because it changes how the term feels. 'Uncanny' is a translation artefact. Mori's Japanese phrase means roughly eerie valley. The 1978 English rendering chose uncanny, which dragged in a whole philosophical tradition — Jentsch, Freud, the unheimlich — that Mori was not invoking. The word acquired a psychoanalytic pedigree at the border, in translation, the same way ego did.
And the hypothesis itself remains contested: possibly generational, possibly cultural, possibly not one phenomenon at all. Which is fine. It has been one of the most productive wrong-or-right ideas in the field for more than fifty years.
Where it comes from
Japanese — eerie valley
The English word 'uncanny' imported Freudian and Jentschian baggage that Mori's Japanese never carried — a documented translation artefact from the 1978 rendering.
- 1970
Masahiro Mori publishes the essay in a Japanese journal.
- 1978
The phrase is translated literally into English, creating the accidental Freudian association. An authorised full translation follows decades later.
In real life
- Mori's own example: a lifelike prosthetic hand is fine to look at, but if it moved to shake yours in a dim room, he wrote, you would shriek.
- The digital face looks fine on the film poster; in motion, the eyes don't quite track and the cinema squirms.
- 'The new customer-service avatar smiles about half a second too late. Deep in the uncanny valley.'
Questions
Q01Who coined it?
The roboticist Masahiro Mori, in a 1970 essay in a Japanese magazine.
Q02Is 'uncanny' a good translation?
It is literal but loaded. It pulled in Jentsch's and Freud's philosophical concepts of the uncanny — an association Mori's Japanese phrase, which means roughly 'eerie valley', did not carry.
Q03Does movement matter?
Enormously. Mori argued motion amplifies both the peaks and the valley — which is why a still render can look fine and the same face in motion can be unbearable.
Sources
- Uncanny valley — Wikipedia.
- IEEE Spectrum — 'The Uncanny Valley' (authorised translation).
- uncanny valley (etymology) — Wiktionary.