Amazon has officially announced the creation of its new research lab, the Amazon AGI SF Lab, located in San Francisco. This dedicated team will focus on developing foundational technologies for creating practical AI agents capable of taking actions in both digital and physical environments. The goal is to empower AI systems that can help users by increasing productivity and improving outcomes across various domains.
The lab’s work will build upon the Amazon AGI team’s recent advancements, including the introduction of Amazon Nova, a new generation of state-of-the-art foundation models. By maintaining a lean and agile team structure, the lab aims to foster rapid innovation while still having the resources of Amazon to drive significant breakthroughs. This approach allows teams to move quickly and focus on high-risk, high-reward research.
The lab is composed of talent from Adept, whose previous work in developing AI agents capable of handling complex workflows using human-like tools (computers, web browsers, and code interpreters) will be instrumental in the lab’s initiatives. Key research areas include integrating large language models (LLMs) with reinforcement learning (RL) for reasoning, planning, and creating agents capable of adapting to physical environments. The lab aims to develop AI that can self-correct, infer human goals, and learn from feedback. 
The Amazon AGI SF Lab is actively seeking a diverse group of passionate individuals from a range of disciplines, including AI, physics, math, and quantitative finance. The lab encourages candidates of all experience levels to apply and contribute fresh perspectives to the field of AI.
