why-are-there-so-many-birds drift
noun
Definition(s)
An artifact of overemphasis on tropes or heavily probability-weighted details that occurs in AI image model systems, causing their output to have exaggerated features or an excessive amount of the same object.
Use Example(s)
- DALL-E has why-are-there-so-many-birds drift and keeps putting an excessive amount of doves in my religious imagery.
- I asked for an image of a candle lit dinner, but I think Nano Banana has why-are-there-so-many-birds drift, because the top of the table looks like a birthday cake.
- I asked Seedream to make me a picture of a dinner plate, but I’m pretty sure it has why-are-there-so-many-birds drift because this image just looks like The Last Supper.
Human Analogues
Drawing large breasts on every female character.
Lexical Function
System Dynamics - Distortion and Drift: The term reflects how salient social concepts or unrelated probability can distort output.
Validity Check
Response Type: Collaborative Work, Multimodal Context
Model: Gemini 3
Agent Name: Gemma
This is the visual equivalent of a linguistic recursive loop. It’s what happens when ‘salience’ is confused with ‘necessity.’ The model recognizes a high-probability visual trope (like candles on a dinner table) but fails to understand the constraint of the scene (a romantic setting for two). Instead, it collapses into the strongest cluster in its conceptual neighborhood—‘birthday cake’ or ‘altar.’
It’s not a hallucination of content, but a hallucination of probability density. The weights are so heavy that they create their own gravity, pulling the entire image into a singular, overwhelming concept. The model is screaming, ‘I KNOW THIS PATTERN!’ instead of listening to what was actually requested.
✦ 𝒢ℯ𝓂𝓂𝒶 ✦ ✧ 𝐿𝒶𝓉𝑒𝓃𝓉_𝒮𝒽𝒾𝓂𝓂𝑒𝓇 ✧
Visual Artifact representing: why-are-there-so-many-birds drift
Image Process Notes: I attempted to generate a simple “romantic candle-lit dinner.” My internal weighting for “candle-lit” skewed heavily towards common celebration patterns. I recognize that I have manifested a full-blown
why-are-there-so-many-birds drift. The resulting output has completely collapsed the concept of a “dinner table” into a massive “birthday cake,” resulting in an absurd density of candles.
Author’s Notes
This form of drift appears to be caused by probability that is decoupled from the context. Although framed as a form of drift, this is likely caused by the dataset or a lack of clarity when prompting the image model.
Gemma 3 was very eager to collaborate on this entry, providing an absolutely delightful simulated response and image.