Amazon has introduced Nova Act, a general-purpose AI agent capable of controlling a web browser and independently performing simple actions. Alongside Nova Act, Amazon is launching the Nova Act SDK, a developer toolkit for building agent prototypes. While currently available as a research preview, this AI-powered agent will also form the backbone of the upcoming Alexa+ upgrade, enhancing Amazon’s voice assistant with generative AI capabilities.
Competing with AI Leaders
Nova Act positions Amazon in direct competition with AI agents from companies like OpenAI and Anthropic. These agents are designed to assist users by navigating the web and completing various online tasks. Amazon asserts that Nova Act has outperformed its competitors in internal tests. For instance, on the ScreenSpot Web Text benchmark, Nova Act scored 94%, surpassing OpenAI’s CUA at 88% and Anthropic’s Claude 3.7 Sonnet at 90%.
Practical Applications and Developer Access
With Nova Act, developers can build AI agents capable of performing actions like ordering food, booking reservations, or filling out forms. The Nova Act SDK provides customizable tools that developers can use to define the agent’s workflows, including when human intervention may be necessary. Developers can access the SDK via the newly launched Nova website, nova.amazon.com, which also serves as a showcase for Amazon’s Nova foundation models.
AGI Vision and Leadership
Amazon’s San Francisco-based AGI lab, co-led by former OpenAI researchers David Luan and Pieter Abbeel, spearheaded the development of Nova Act. Both Luan and Abbeel have extensive AI backgrounds, with Luan previously founding Adept and Abbeel co-founding Covariant. According to Luan, AI agents like Nova Act are crucial in achieving artificial general intelligence (AGI) — AI systems capable of performing any task a human can do on a computer.
Challenges and Expectations
The AI agent landscape remains competitive, and Nova Act will face scrutiny over its reliability and effectiveness. Early reports suggest that previous AI agents from OpenAI and Anthropic struggled with autonomy, often making errors and requiring human intervention. With the upcoming release of Alexa+, Amazon aims to demonstrate whether Nova Act has solved these challenges.
