AI Agents

GR-2 is an advanced general purpose robotic agent designed for diverse and generalizable robotic operations. It is first pre-trained on a large number of Internet videos to capture the dynamics of the world. This large-scale pre-training, involving 38 million video clips and over 50 billion tokens, enables GR-2 to generalize across a wide range of robotic tasks and environments in subsequent policy learning. Subsequently, GR-2 was fine-tuned for video generation and action prediction using robot trajectories. It demonstrates impressive multi-task learning capabilities, achieving an average success rate of 97.7% on more than 100 tasks. Additionally, the GR-2 excels in new, previously unseen scenarios, including new backgrounds, environments, objects and tasks. Notably, GR-2 scales efficiently as model size increases, highlighting its potential for continued growth and application.

MultiOn Playground is a platform that provides real-time preview and interactive experience. It supports websites that require authentication and allows users to enter a query or select a sample through the chat interface to start the interaction. Users can also control or correct the agent's operation through on-screen interaction.

OK-Robot is an open modular framework that enables zero-sample item handling tasks based on language instructions in any home environment. The framework adopts a modular design, using 3D VoxelMap for open vocabulary navigation, AnyGrasp and LangSam for open vocabulary grabbing, and placement primitives for item placement. The framework does not require pre-training and can achieve zero-shot generalization of language instructions.

Alter3 is an anthropomorphic robot based on GPT-4, which can generate various human actions through natural language instructions and achieve zero-shot learning. It has 43 joint degrees of freedom and can move freely, while combining facial expression recognition and generation to achieve complex interactions. Users only need to provide language instructions, and Alter3 can independently adjust the code and generate corresponding movements without manual iterative adjustments. Movements can also be optimized through verbal feedback to form motor memory. This system that combines language understanding and body movement greatly enhances the potential of human-computer interaction.

AI YouTube Summarizer is the leading platform for building AI agents based on your data. Users can use the AI ​​YouTube Summarizer to easily build and train their own AI agents, no coding required. The product is positioned to provide users with convenient AI agent training and application services, helping users quickly build their own intelligent assistants. AI YouTube Summarizer also provides a variety of tools and integrations, such as ChatGPT Bot, Chat Assistant, ChatGPT bot for Slack, etc., to meet the different needs of users.

CoT-Igniting-Agent is a language intelligent agent system based on chain-of-thinking reasoning. It uses chain reasoning methods to handle natural language understanding and generation tasks, and has a wide range of applications. The system uses advanced algorithms and technology and is efficient, reliable and safe. CoT-Igniting-Agent can be used in areas such as instruction generation, reasoning, application development, and scientific research.

GATE is a learning framework that uses language models to guide task specifications and infer expected behaviors through free-form language-based interactions with users. It conducts research in three areas: email verification, content recommendation, and ethical reasoning. In pre-registration experiments, we found that language models that prompt GATE execution tend to be more informative than user-written prompts or labels, such as by generating open-ended questions or synthesizing informative edge cases. Users report that interactive task guidance requires less effort than prompts or example markers and provides novel considerations that users did not initially anticipate. Our results demonstrate that language model-based guidance can be a powerful tool for aligning models with complex human preferences and values.