#USA #San Francisco #Moonvalley #Reka #Physical AI

Strategic Alliance for Advanced AI

In a groundbreaking move within the artificial intelligence realm, Reka, a pioneering AI research lab known for its innovative approaches to foundational intelligence, has joined forces with Moonvalley. This collaboration aims to fast-track the development of models and infrastructure dedicated to the burgeoning field of Physical AI. With the integration of skilled researchers and engineers transitioning from notable tech giants like DeepMind, Meta, and Google, both companies are poised to redefine the possibilities of AI applications in real-world scenarios.

Reka's Vision and Mission

Founded with the ambition to apply AI technologies beyond traditional computing realms, Reka focuses on creating intelligent systems that can understand and interact within the physical world. Reka’s leadership, under the guidance of CEO Dani Yogatama, emphasizes the importance of this collaboration. “Physical AI is fundamentally a research problem,” he stated, underscoring the need for advanced capabilities in reasoning, simulation, and real-time decision-making.

The Expertise Coming Onboard

The partnership brings onboard prominent figures like Dr. Mateusz Malinowski and Dr. Mikołaj Bińkowski, who are highly regarded for their contributions to AI at Google DeepMind. Their accumulated knowledge and experience, especially in models that facilitated video generation, will be crucial in elevating Reka's research leadership. Dr. Malinowski described the transformative potential of the new team when he remarked, “We’re building models that understand how the physical world works.” This sentiment reflects the core mission of the collaboration: to create AI systems equipped to reason about consequences and act safely and responsibly in physical domains.

Introduction of World Language Action Model (WLAM)

One of the primary initiatives resulting from this collaboration is the World Language Action Model (WLAM). This innovative omni model is designed to process egocentric and various physical world data, allowing AI systems to perform realistic simulations for optimal planning. The WLAM’s capability to produce high-fidelity multimodal outputs and execute robotic tasks with precision marks a significant advancement in AI.

Four Pillars of Operation

Reka operates through four distinct but interconnected pillars:

• - Reka Labs: Focuses on the research and development of foundational AI intelligence.
• - Reka Infer: Provides inference infrastructure and an API for multimodal AI capabilities.
• - Reka Vision: A video processing platform that supports tasks like tagging and searching vast quantities of video data.
• - Reka Claru: Develops training data specifically for Physical AI applications, including robotics trajectories and expert human judgment.

With this multi-faceted approach, Reka aims to harness the dynamic capabilities of AI, enabling it to address complex challenges across industries.

Call for Collaboration and Future Prospects

As part of their innovative journey, Reka is actively seeking to expand its team across research, engineering, and product development. Organizations interested in exploring early access to Reka’s platforms are encouraged to connect with the team. The potential of this partnership is immense, with the prospect of setting new benchmarks for AI technologies that can genuinely interact and improve the physical world.

About Reka

Reka is an AI research lab situated in San Francisco, established by a team of leading researchers from notable tech entities such as DeepMind, Meta, and Google. As Reka continues to push the boundaries of AI, the collaboration with Moonvalley marks a pivotal step towards revolutionizing how AI interacts in the real world.