Amazon sets up AI agent-focused R&D lab in San Francisco
Amazon has announced the opening of a new research and development facility, the Amazon AGI SF Lab, in San Francisco. The lab will be led by David Luan, co-founder of AI start-up Adept. The main goal of this project is to create "foundational" capabilities for AI agents that can work online and offline, and handle the complex workflows using computers, web browsers, and code interpreters.
Lab to collaborate with broader AGI team
The Amazon AGI SF Lab will be working alongside Amazon's broader artificial general intelligence (AGI) team. This was disclosed in a joint statement from Luan and Pieter Abbeel, a robotics research lead who joined Amazon through the company's "license and hire" deal with Covariant. An Amazon spokesperson said to TechCrunch that Abbeel shall be working "closely" with Luan and the AGI SF Lab going forward.
Amazon AGI SF lab's initial focus and team composition
The lab's first focus will be on critical research areas that will allow AI agents to take real-world actions, learn from human feedback, self-correct, and infer our goals. This was revealed by Luan and Abbeel in their joint statement. Initially, the lab will be staffed by Adept employees, but it plans to hire more researchers in fields like quantitative finance, physics, and math.
Amazon's acquisition of Adept and future plans
Back in June, Adept had agreed to license its technology to Amazon, and Luan and parts of Adept's team joined the e-commerce giant. Luan will continue to work under Rohit Prasad, the former chief of Alexa who's heading an AGI team specializing in large language models. Amazon's move is similar to Microsoft's deal with AI company Inflection in May, both of which have attracted regulatory attention over concerns about tech giants potentially stifling their AI competitors.
Adept's vision and the growing AI market
Adept was founded with the goal of building an AI model capable of interacting with any software tool in natural language. The company's vision was to build an "AI teammate," trained to use a range of different software tools and APIs. The market for such "agentic" AI could be worth $31 billion by year-end, according to Emergen Research, with 82% of companies planning to integrate AI agents within three years, a Capgemini poll found.