AI as a User of AI : Towards Responsible Autonomy

Abstract

Recent advancements in Artificial Intelligence (AI), particularly in generative language models and algorithms, have led to significant impacts across diverse domains. AI capabilities to address prompts are growing beyond human capability but we expect AI to perform well also as a prompt engineer. Additionally, AI can serve as a guardian for ethical, security, and other predefined issues related to generated content. We postulate that enforcing dialogues among AI-as-prompt-engineer, AI-as-prompt-responder, and AI-as-Compliance-Guardian can lead to high-quality and responsible solutions. This paper introduces a novel AI collaboration paradigm emphasizing responsible autonomy, with implications for addressing real-world challenges. The paradigm of responsible AI-AI conversation establishes structured interaction patterns, guaranteeing decision-making autonomy. Key implications include enhanced understanding of AI dialogue flow, compliance with rules and regulations, and decision-making scenarios exemplifying responsible autonomy. Real-world applications envision AI systems autonomously addressing complex challenges. We have made preliminary testing of such a paradigm involving instances of ChatGPT autonomously playing various roles in a set of experimental AI-AI conversations and observed evident added value of such a framework.