Amazon has announced the creation of the Amazon AGI SF Lab, a dedicated team based in San Francisco that will develop highly capable AI agents.
The lab will be seeded by the team hired from AI startup Adept, and will focus on enabling AI agents to perform real-world actions, learn from human feedback, and infer goals.
Amazon aims to combine large language models with reinforcement learning to solve reasoning and planning, learned world models, and generalizing agents to physical environments.
Amazon CEO Andy Jassy highlighted the cost-effectiveness and multimodal capabilities of the recently announced Nova models, which offer performance parity with industry-leading models.