A Compositional Paradigm for Foundation Models: Towards Smarter Robotic Agents
By: Luigi Quarantiello , Elia Piccoli , Jack Bell and more
Potential Business Impact:
AI learns new things without forgetting old ones.
The birth of Foundation Models brought unprecedented results in a wide range of tasks, from language to vision, to robotic control. These models are able to process huge quantities of data, and can extract and develop rich representations, which can be employed across different domains and modalities. However, they still have issues in adapting to dynamic, real-world scenarios without retraining the entire model from scratch. In this work, we propose the application of Continual Learning and Compositionality principles to foster the development of more flexible, efficient and smart AI solutions.
Similar Papers
Leveraging Foundation Models for Enhancing Robot Perception and Action
Robotics
Robots learn to do more things in messy places.
Toward Accurate Long-Horizon Robotic Manipulation: Language-to-Action with Foundation Models via Scene Graphs
Robotics
Robots learn new tasks without special training.
The Future of Continual Learning in the Era of Foundation Models: Three Key Directions
Machine Learning (CS)
Keeps AI smart and updated with new information.