Recently, Amazon announced the establishment of a new research laboratory in San Francisco—the Amazon AGI SF Lab—dedicated to developing artificial intelligence systems capable of interacting with both digital and physical environments simultaneously. The lab's primary objective is to create AI agents that can handle complex workflows by utilizing common tools such as web browsers and code interpreters.

The establishment of this lab is based on Amazon's recent acquisition of the Adept team, and it plans to recruit dozens of researchers from various fields, including physics and quantitative finance. This initiative further expands Amazon's existing AGI team's efforts, particularly in the research of its Nova foundational model, demonstrating Amazon's collaborative strategy in AI development.

The lab is led by Adept's co-founders, David Luan and Pieter Abbeel, who believe that these new AI agents have the potential to overcome current technological limitations. In their announcement, Luan and Abbeel stated that the lab's initial research will focus on several key areas to enable AI agents to perform real-world actions, learn from human feedback, self-correct, and infer human objectives.

In a blog post announcing the lab's establishment, Luan mentioned that they are developing practical AI designed to genuinely serve individuals, enhancing customers' productivity, capabilities, and satisfaction.

The Amazon AGI SF Lab will feature a streamlined team and cutting-edge research agenda, representing Amazon's latest strategic initiative aimed at achieving breakthrough results that could redefine AI capabilities and user interactions.