To achieve this, there are several key aspects to consider:

To achieve this, there are several key aspects to consider:

1. Enhanced Capabilities of Large Models:
   The application of large models in robotics has been revolutionizing the field. These models have the ability to process and understand complex information across multiple modalities, from language to language-visual models. However, the true power lies in their ability to incorporate state estimation information, further enriching the modalities within the large model. By encoding different modalities of information into the same vector space, regardless of the input modality, the large model can generate implicit mathematical descriptions for cross-modal tasks. This advancement opens up new possibilities for robots to interact and understand the world around them.

2. Microsoft Research: ChatGPT for Robotics:
   One notable example of the application of large models in robotics is Microsoft Research's ChatGPT for Robotics. This project leverages the power of ChatGPT, a language model developed by Microsoft, to enable robots to engage in natural language conversations. By using large models, robots can understand and respond to human commands and queries, making human-robot interactions more intuitive and efficient. This development showcases the potential of large models in enhancing the capabilities of robots and bridging the gap between humans and machines.

3. PaLM-E: Enriching Large Models with Visual Information:
   Google has made significant progress in enriching large models with visual information through their project called PaLM-E. Initially, PaLM-E focused on mapping semantic information to a single image, enabling image classification. However, Google expanded upon this by incorporating object instance-level segmentation, allowing the large model to extract detailed information about objects within the image. This additional modality of object state information is then encoded within the large model, further enhancing its understanding of visual data. By expanding the capabilities of large models to include visual information, robots can better perceive and interpret their surroundings, leading to improved decision-making and task execution.

When examining the commonalities between these two developments, we can observe the shared objective of enriching large models with new modalities. Both Microsoft Research and Google have recognized the potential of incorporating additional information, whether it be state estimation or visual data, to enhance the capabilities of large models in robotics.

In conclusion, the application of large models in robotics is transforming the field by enabling robots to process and understand information across multiple modalities. Through advancements like Microsoft Research's ChatGPT for Robotics and Google's PaLM-E project, we are witnessing the integration of language, visual, and state estimation modalities within large models. This integration opens up new possibilities for robots to interact with humans, understand their environment, and make informed decisions. As the field of robotics continues to evolve, it is crucial to explore and harness the power of large models to unlock the full potential of artificial general intelligence and beyond.

Actionable Advice:

1. Embrace large models: Incorporating large models in robotics can significantly enhance the capabilities of robots. Explore ways to leverage these models to improve tasks such as natural language understanding and computer vision.

2. Cross-modal integration: Investigate methods to encode different modalities of information into a shared vector space within large models. This integration allows for a more holistic understanding of data and enables robots to handle cross-modal tasks effectively.

3. Continual learning: As large models evolve and new modalities are incorporated, it is essential to ensure continual learning and adaptation. Develop mechanisms to update and fine-tune large models to keep up with advancements in the field and maximize their potential in robotics applications.