Can Large Language Models Play Text Games Well? Current State-of-the-Art and Open Questions
This work addresses the problem of evaluating AI capabilities in interactive environments for researchers in AI, ML, and NLP, but it is incremental as it builds on existing text game benchmarks.
The paper investigates the capacity of large language models like ChatGPT to play text games, finding that while they perform competitively compared to existing systems, they exhibit low intelligence by failing to construct world models, leverage existing knowledge, or infer goals during gameplay.
Large language models (LLMs) such as ChatGPT and GPT-4 have recently demonstrated their remarkable abilities of communicating with human users. In this technical report, we take an initiative to investigate their capacities of playing text games, in which a player has to understand the environment and respond to situations by having dialogues with the game world. Our experiments show that ChatGPT performs competitively compared to all the existing systems but still exhibits a low level of intelligence. Precisely, ChatGPT can not construct the world model by playing the game or even reading the game manual; it may fail to leverage the world knowledge that it already has; it cannot infer the goal of each step as the game progresses. Our results open up new research questions at the intersection of artificial intelligence, machine learning, and natural language processing.