Why LLMs are vulnerable to the ‘butterfly effect’

Researchers found that the smallest of perturbations, such as adding a space at the end of a prompt, can cause an LLM to change its answer.

ai ai newsletter change chatgpt conversational ai data infrastructure newsletter featured data-labelling found generative-ai innovator spotlight newsletter featured jailbreak large language models llm llms ml and deep learning nlp openai prompt prompting prompt learning researchers science space vulnerable