Generative artificial intelligence can create text and images, but can it accurately generate system architecture diagrams? This article dives into three use cases for AI-generated diagrams, from generic technology diagrams to whiteboard diagrams for future systems. ChatGPT did impressively well with simple requests, like a basic AWS serverless system. However, whiteboarding a detailed browser-based image processing solution revealed limitations, like difficulty representing different flows. When tasked to diagram a real-life system from source code, both ChatGPT and Claud.ai failed to provide detailed and accurate results. The challenges of automated diagram generation, from the lack of training data to the complexity of code analysis, highlight the current limitations of AI in diagramming.
https://www.ilograph.com/blog/posts/diagrams-ai-can-and-cannot-generate/