Amazon Launches AI Agent Lab, Adept Co-founder to Lead

Amazon has unveiled plans to establish a new research and development facility in San Francisco, named Amazon AGI SF Lab, with the mission of developing foundational capabilities for AI agents. This lab will be led by David Luan, co-founder of the AI startup Adept.

The Amazon AGI SF Lab aims to create agents capable of acting within both digital and physical environments, managing intricate workflows through the use of computers, web browsers, and code interpreters. The lab's initiatives will build upon Amazon's broader AGI team, as referenced in a joint publication by David Luan and Pieter Abbeel. Pieter Abbeel, a specialist in robotics research, joined Amazon through an "authorization and employment" agreement with Covariant. An Amazon spokesperson told TechCrunch that Abbeel will collaborate closely with Luan and the AGI SF Lab moving forward.

Luan and Abbeel stated in their article, "Our initial research will concentrate on several key areas that will enable AI agents to undertake actions in the real world, learn from human feedback, self-correct, and infer our objectives."

The core team of the lab will consist of Adept’s personnel, with Amazon also planning to recruit dozens of additional researchers from fields such as quantitative finance, physics, and mathematics.

In June, Adept agreed to license its technology to Amazon, resulting in part of Adept's team, including Luan, joining the e-commerce giant. Adept is focused on developing AI agents that can perform software tasks, while Luan will continue to work under Rohit Prasad, the former head of Alexa and now leading an AGI team focused on large language models.

Amazon’s partnership with Adept mirrors a similar agreement made by Microsoft in May with AI company Inflection. Both deals have undergone regulatory scrutiny, as policymakers both domestically and internationally assess whether tech giants are hindering competition among AI rivals.

Founded two years ago, Adept aims to create AI models that can execute operations across any software tool using natural language, envisioning an "AI teammate" adept at utilizing various software tools and APIs.

Currently, numerous companies share this vision. According to Emergen Research, the "agent-based" AI sector could be valued at $31 billion by the end of this year. A survey by Capgemini revealed that 82% of organizations plan to integrate AI agents within the next three years to enhance efficiency.

In addition to startups like Orby, Emergence, and Rabbit, major AI players including OpenAI are developing agent products capable of autonomously completing tasks. Earlier this year, Anthropic released its agent technology, and Google is reportedly working on AI agents that can handle purchasing tasks such as booking flights and hotels.

Although Amazon has already entered the agent space, significant strategic moves are yet to be made. In July, the company announced conversational agents for its Bedrock AI development platform, and last week introduced agent functionalities on the Amazon Q Business Assistant platform for enterprise clients and developers. Amazon CEO Andy Jassy has also suggested that future iterations of Alexa will feature enhanced agent capabilities, enabling not only response to queries but also the execution of actions.