Description
Poolside is a foundation model company that aims to bring intelligence to every aspect of work. Founded in San Francisco in April 2023, Poolside's mission is to drive abundance for humanity by creating artificial general intelligence (AGI). The company is dedicated to building open-weight foundation models and the systems that refine and improve them. Their models, such as Laguna s 2.1 and Laguna xs 2.1, are designed to optimize quality, speed, and efficiency, with parameters ranging from 118 billion to 33 billion.
The company emphasizes the importance of reinforcement learning (RL) as a key scaling axis for achieving AGI. Poolside believes that while scaling compute and language modeling are crucial, RL offers a unique path to learn from new experiences and decompress humanity's knowledge. They argue that the internet, while a vast repository of information, often lacks the underlying thought processes that led to the creation of that information. By utilizing RL, Poolside aims to overcome the limitations of web data and tap into the rich potential of human experiential learning.
Poolside's approach involves orchestrating millions of coding environments, allowing their models to learn from real-world interactions in a safe and controlled manner. This method enables the models to receive immediate feedback from compilers, linters, and test suites, facilitating rapid iteration and improvement. As the company scales its operations, it plans to deploy agents in live production settings, further enhancing the volume and variety of interaction data available for learning.
The vision at Poolside extends beyond language, encompassing various modalities such as vision and spatial understanding. By focusing on language as a foundational element, they believe that achieving superhuman capabilities in other areas will become significantly easier. Their research and engineering efforts are directed towards creating a pathway to AGI that leverages the power of RL-driven exploration and the efficient use of existing data. Poolside invites those interested in the forefront of applied research and engineering to join their mission.
Poolside's Core Features
Open-weight foundation models
Reinforcement learning focus
High-performance coding environments
Real-world interaction data
Multi-modal capabilities
Rapid iteration and feedback
Optimized for quality and efficiency
Research-driven development
How to use Poolside?
Explore: Review Poolside's foundation models and their capabilities.
Engage: Participate in their research initiatives and discussions.
Implement: Utilize their coding environments for AI development.
Iterate: Use feedback from coding tasks to improve model performance.
Poolside's Use Cases
- AI Development
- Software Engineering
- Research Collaboration
- Model Training
- Performance Optimization
