New AI platform could help household robots to read between the lines
New AI platform could help household robots to read between the lines
A new AI platform called FLEX, developed by researchers at MIT and New York University, enables household robots to understand implicit meanings in commands rather than just literal instructions. The system uses a large language model (GPT-4o) to infer broader rules from short commands, such as avoiding not just a laptop but also other electronic devices. It operates in three steps: using an LLM to make assumptions, inferring broader rules like treating fragile objects with care, and cleaning up the list based on computer vision. Early tests were conducted in simulation on tasks like serving coffee, moving a candle, and carrying a bowling ball. The team plans to present the system at the Conference on Robot Learning. While promising for personalization in real homes, the system's performance in physical environments remains unverified.
Source: Robotics — Robot learning news discovery · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S6
Reported numbers
cycles
≈5 seconds
View original evidence
a broader rule. For example, after a mug, it infers that fragile objects like it deserve extra care. The third step is to clean up. The robot's computer vision system checks which objects are no longer in the vicinity and deletes them from the list. The cycle takes around five seconds. Early tests only ran in simulation The team tested FLEX in simulated robot arms on three tasks. The robot served coffee, moved a can
Open source S6
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
a broader rule. For example, after a mug, it infers that fragile objects like it deserve extra care. The third step is to clean up. The robot's computer vision system checks which objects are no longer in the vicinity and deletes them from the list. The cycle takes around five seconds. Early tests only ran in simulation The team tested FLEX in simulated robot arms on three tasks. The robot served coffee, moved a can
Open source S6
ell. Household robots learn the reasoning behind a command FLEX comes from researchers at MIT and New York University. They will present the system at the Conference on Robot Learning this month. Led by Andreea Bobu, an assistant professor of aeronautics and astronautics at MIT, the team says FLEX is a flexible feature-editing pipeline for robot motion planning. Here's how the system works. A robot arm carries a cup
Open source S3
New AI platform could help household robots to read between the lines
Open source S1
Source：Robotics — Robot learning news discovery · tech.yahoo.com