Researchers at Google DeepMind and London School of Economics have introduced a new method to evaluate sentience in artificial intelligence systems by using a pain-reward trade-off test. In a study involving large language models (LLMs), AI systems were challenged with a text-based game offering scenarios of avoiding pain or achieving pleasure in exchange for scoring points. This innovative approach, inspired by animal studies, aims to probe AI systems without relying on self-reported states, which can merely reflect learned patterns instead of genuine sentience. Although the tested AI models are not concluded to be sentient, the study provides a promising framework for future explorations of AI sentience and the potential societal implications of recognizing AI systems with sentient characteristics.