Elon Musk has released the Grok-1 model, a 314 billion parameter language model that features a unique tone aligning with his free speech approach, now available open source under the Apache 2.0 license. Other recent developments in the field of machine learning include Nvidia introducing new GPUs and a humanoid robot foundation model, OpenAI leaks revealing undisclosed model names, and the European lawmakers passing the AI Act to regulate AI technologies.
In the latest news related to machine learning (ML), Elon Musk has openly released the Grok-1 model, a 314 billion parameter language model. The Grok-1 model has a unique tone, described as more quippy and sarcastic compared to other models, aligning with Elon Musk’s free speech approach. The model is fully open source under the Apache 2.0 license, available for exploration and use.
Nvidia recently announced new GPUs at their GTC conference, featuring chips with improved performance and new features like FP4 tensor cores. Additionally, Nvidia introduced Groot, a foundation model for humanoid robots, highlighting their focus on advancing technologies in the field of robotics. They also announced support for the Robot Operating System (ROS), a widely used standard in robotics.
OpenAI leaks suggest that individuals are exploring undiscovered model names accessible via the OpenAI API. This has led to the discovery of model names such as Jan Street and superhuman, indicating a level of secrecy and varying purposes for these models. The leak raises questions about the usage and nature of these models within the AI landscape.
In other news, a major AI regulation act has been passed by European lawmakers, marking a significant step in AI legislation. The AI Act aims to regulate AI technologies and their impact on society, becoming a global standard for AI governance. The act is expected to come into force after final checks and endorsements, demonstrating Europe’s commitment to integrating AI regulations into legislation.
Additional updates include advancements in AI research such as Apple’s release of multimodal models and Google’s efforts to combat spam content in search results. Coher has introduced Coher Embed V3 focusing on embedding models, supporting INT8 and binary embeddings to reduce memory requirements. These developments highlight ongoing progress and innovation in the field of machine learning and AI technologies.