RWKV: Small Team Aims to Be Android of AI Era with Big Model

Source:

Source:
The robotics research team at Google DeepMind recently released a robotics project called RT-2. This project took 7 months to develop and uses a large model for training. RT-2 has capabilities such as symbol understanding, reasoning, and human recognition, and can think and complete tasks based on human instructions. By combining the large model with the robot's operational capabilities, RT-2 can accomplish tasks that involve logical leaps, such as from 'extinct animals' to 'plastic dinosaurs'. The results of this project performed well in various sub - category tests, with performance up to three times that of the previous generation of robot models. This research result demonstrates the potential of large models in robotics research and is expected to drive the development of robots in the future.
Against the backdrop of increasing global competition in AI, the EU has officially announced an unprecedented artificial intelligence investment plan. Today at the AI Action Summit in Paris, European Commission President Ursula von der Leyen formally unveiled a grand initiative called 'Invest AI', showcasing the EU's strong determination in global AI competition. This investment plan, with a total scale of €200 billion, notably includes a dedicated €20 billion European Fund for the construction of AI gigafactories. This decision marks the EU's official entry into large-scale AI development.
Recently, Zhangyue Technology announced on its interactive platform that it is actively promoting the application of artificial intelligence large models in the vertical field of digital reading. With continuous technological advancements, Zhangyue Technology is leveraging its advantages in content copyright, creator ecosystem, and a large user base to deeply integrate leading domestic artificial intelligence large models with the company's business scenarios. Zhangyue Technology aims to enhance business efficiency and user experience by introducing and applying multiple large models, such as DeepSeek, Doubao, and other related technologies. The open-source characteristics of large models and
The ByteDance Doubao large model team announced today the successful development of a new sparse model architecture called UltraMem. This architecture effectively addresses the high memory access issues during the inference of MoE (Mixture of Experts) models, improving inference speed by 2 to 6 times compared to MoE, and reducing inference costs by up to 83%. This groundbreaking advancement opens a new path for efficient inference of large models. The UltraMem architecture successfully resolves the memory bottleneck during inference of MoE architectures while maintaining model performance. Experimental results show that the parameters and activation conditions are the same.
According to media reports, Baidu is set to launch its next generation AI model Ernie 5.0 this year. Sources state that the Ernie 5.0, referred to as a 'foundation model', will see significant enhancements in multimodal capabilities, though specific functionalities were not detailed. This news comes as Apple shifts its potential clients towards Alibaba, leading to widespread speculation that Baidu is seeking to stabilize its stock price and market position amid changing circumstances.