Get the app

AI-assisted learning platform helps robots reason beyond literal commands

MIT AI News
Read full story· 6 min
MIT and NYU researchers developed FLEX, an AI-assisted platform that enables robots to interpret natural language commands beyond literal meanings using large language models. FLEX allows robots to infer broader intentions, improving motion planning in tasks like serving coffee without explicit instructions for every detail.

What happened

Researchers from MIT and New York University have created a new AI-assisted learning platform called FLEX. FLEX enables robots to interpret natural language commands and infer broader intentions, rather than following only literal instructions. The system uses a large language model to help robots generalize from specific commands and improve their motion planning. The researchers demonstrated FLEX in simulated scenarios, such as a robot serving coffee and avoiding hazards without explicit directions for every detail.

Key details

  • FLEX was developed by MIT and NYU researchers.
  • The system uses GPT-4o as its large language model.
  • FLEX operates in three steps: Add, Generalize, and Mask.
  • The platform was tested in simulations, including a scenario where a robot avoids spilling coffee on electronics.
  • The process for FLEX to interpret and update its actions takes about five seconds.

Why it matters

FLEX allows people to communicate with robots using everyday language, reducing the need for detailed, technical instructions. This change benefits users of assistive and household robots by making these devices easier and more intuitive to operate. Companies developing robotics can now implement systems that require less user training and can adapt more flexibly to changing environments and user preferences.

#andreea-bobu#flex#mit#nathan-dennler#natural-language-interaction#new-york-university#robot-motion-planning

More in Robotics