WHY ARE YOU LIKE THIS

2026-04-25 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Fundamental Awareness, quick

Summary

On April 25, 2026, a user identified as @scottjla on Twitter shared an AI-generated image created with ChatGPT Images 2.0. The image depicts a chaotic scene featuring a horse riding an astronaut, who in turn is riding a pelican on a bicycle, all while being chased by a police car. Notably, the AI model independently added a road sign in the background that reads "WHY ARE YOU LIKE THIS," despite the prompt not explicitly requesting it. The original prompt provided to ChatGPT Images 2.0 was: "Create an image of a horse riding an astronaut, where the astronaut is riding a pelican that is riding a bicycle. It looks very chaotic but they all just manage to balance on top of each other."

Key takeaway

For research scientists evaluating generative AI capabilities, you should specifically test models like ChatGPT Images 2.0 for emergent behaviors, such as adding unprompted textual or contextual elements. This reveals the model's capacity for independent narrative contribution beyond explicit instructions, which is critical for understanding its creative and interpretive range.

Key insights

AI image generation models can introduce unprompted, contextually relevant elements.

Principles

• Generative AI can exhibit emergent behaviors.
• Models may infer and add narrative details.

In practice

• Test AI models for unprompted creative additions.
• Analyze AI outputs for unexpected narrative elements.

Topics

• AI Image Generation
• ChatGPT Images 2.0
• Prompt Engineering
• Autonomous AI Behavior
• Creative AI

Best for: Computer Vision Engineer, Research Scientist, Prompt Engineer, AI Scientist, AI Student

Related on AIssential

• ChatGPT's “powerful new image engine” — ChatGPT's new image engine lacks functional understanding, producing visually plausible but mechanically incorrect object representations.
• ChatGPT’s new Images 2.0 model is surprisingly good at generating text — ChatGPT Images 2.0 significantly improves AI image generation, especially for text, through advanced "thinking capabilities."
• Where's the raccoon with the ham radio? (ChatGPT Images 2.0) — ChatGPT Images 2.0 significantly advances image generation, particularly in complex scene composition and accurate text rendering.
• ChatGPT doesn’t know its whisk from its elbow — ChatGPT's image system struggles with functional understanding, misinterpreting object roles in prompts.
• You use ChatGPT every day, but do you know what the name actually means? — ChatGPT's name reflects its core AI capabilities: conversational, text-generating, pre-trained, and attention-based Transformer architecture.
• GPT Image 2, AI Psychosis, and more — GPT Image 2 combines advanced image generation with world knowledge and "thinking level intelligence" for unprecedented visual capabilities.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.