# Visualizing AI Hallucinations: Semantic Entropy as Art
> Explore Semantic Stutter: an artistic project visualizing LLM uncertainty and hallucinations through semantic entropy analysis and aesthetic blur effects.

Tags: ai-hallucinations, semantic-entropy, data-visualization, ai-art, large-language-models, machine-learning, generative-ai
## Visualizing the hidden anxiety of LLMs
- Introduction to using Semantic Entropy to reveal machine hesitation.

## The Illusion of Confidence
- Modern AI interfaces mask uncertainty with polished facades.
- 'Semantic Stutter' aims to break this opacity.

## Technical Foundation: Semantic Entropy
- Based on research by Kuhn et al. (2023).
- Measures uncertainty by sampling multiple answers for one prompt.
- High clustering into different meanings indicates High Entropy (Confusion).

## The Hallucination Detection Mechanism
- Data Point: 95% consistency score for factual questions (Low Entropy).
- Data Point: 45% consistency for ambiguous questions (Med Entropy).
- Data Point: 12% consistency for fabricated topics (High Entropy).

## System Architecture
- Input: Voice/Microphone.
- Processing: Llama 3 generates 5 parallel responses.
- Analysis: Semantic clustering calculation.
- Output: Entropy score determines visual blur intensity.

## Visualizing Uncertainty
- Factual answers remain stable and legible.
- Hallucinations trigger vibrations, overlaps, and text blurring.

## Case Studies
- **Case A (Low Entropy):** Boiling point of water results in sharp, static text.
- **Case B (High Entropy):** Prompts about Napoleon meeting aliens result in trembling, layered ghosting artifacts.

## Artistic Statement
- AI errors are treated as artistic material rather than bugs.
- The goal is to visually display internal machine 'anxiety' and challenge the illusion of objective truth.

## Conclusion
- Semantic Stutter transforms statistical probability into a visceral visual experience.
---
This presentation was created with [Bobr AI](https://bobr.ai) — an AI presentation generator.