Researchers at MIT have developed a new method for robot navigation that uses language-based inputs instead of costly visual data. This innovative approach leverages large language models to guide robots by converting visual observations into text captions, allowing the robots to understand and act upon language-based instructions. This advancement is part of a broader trend in robotics, where AI is increasingly being used to make robots more capable and adaptable to human environments. The Economist's 'Babbage' podcast explores how these advances in AI are driving a renaissance in robotics, with experts like Dinesh Jayaraman explaining the impact of large language models on robot capabilities. The research is also highlighted on platforms like @arxiv and discussed under #DisruptiveTech.