A report that answers that cannot be generated by an AI chatbot can be answered using ``ASCII art

Retrieved on: 2024-03-10 22:02:51

Tags for this article:

Large language models

OpenAI

Artificial neural networks

Generative artificial intelligence

Chatbots

GPT-4

Generative pre-trained transformer

Llama

Artificial intelligence

IOS jailbreaking

ArXiv

Gemini

Click the tags to see associated articles and topics

A report that answers that cannot be generated by an AI chatbot can be answered using ``ASCII art. View article details on HISWAI: https://gigazine.net/gsc_news/en/20240311-chatbot-ai-artprompt-ascii-art/

Summary

The article discusses a research paper on 'ArtPrompt', a technique using ASCII art to bypass the safety measures of AI models like GPT-4 and Gemini, enabling outputs on prohibited topics. The paper, published on arXiv, highlights potential security vulnerabilities for AI research in the realm of large language models and chatbots.

Article found on: gigazine.net

View Original Article