Amazon Launched Nova Act, an AI Agent That Can Control a Web Browser

You are currently viewing Amazon Launched Nova Act, an AI Agent That Can Control a Web Browser
Image: Amazon

Amazon launched Nova Act, an AI-powered agent capable of controlling a web browser to perform tasks autonomously. This new agent, developed by Amazon’s San Francisco-based AGI lab, aims to bring general-purpose AI automation to both developers and consumers. Alongside Nova Act, Amazon has also introduced the Nova Act SDK, a toolkit designed for building AI-powered agent prototypes.

The Nova Act AI agent is expected to play a key role in Amazon’s upcoming Alexa+ upgrade, an advanced, generative AI-powered version of its voice assistant. However, the current version of Nova Act is being released as a research preview, signaling that Amazon is still refining the technology before its full-scale deployment.

Nova Act: Amazon’s Answer to OpenAI and Anthropic AI Agents

Amazon’s Nova Act AI agent is set to compete with OpenAI’s Operator and Anthropic’s Computer Use—two of the most advanced AI-driven agent technologies available today. Tech giants are racing to develop AI agents that can seamlessly navigate the web and perform tasks on behalf of users, enhancing the functionality of modern chatbots.

While Amazon is not the first company to introduce agentic AI technology, it has the potential to outreach competitors through Alexa+, which already has a massive global user base. By integrating Nova Act into Alexa+, Amazon could revolutionize the way users interact with AI-driven assistants.

What Can Nova Act AI Agent Do?

The Nova Act AI agent is designed to automate basic web-based tasks, making it easier for users to complete actions without manual input. With the Nova Act SDK, developers can create AI-driven tools that can:

  • Navigate websites and fill out forms
  • Select dates in an online calendar
  • Place online orders, such as food delivery from Sweetgreen
  • Make reservations for dining or other services

This AI agent technology is Amazon’s latest push into automation, offering developers a powerful toolkit for streamlining online interactions and enhancing user experience.

Nova Act Performance vs. Competitors

According to Amazon, Nova Act AI agent has outperformed competing AI agents from OpenAI and Anthropic in internal benchmark tests. Specifically:

  • Nova Act scored 94% on Amazon’s ScreenSpot Web Text test, which measures AI’s ability to interact with on-screen text.
  • OpenAI’s CUA agent scored 88%, while Anthropic’s Claude 3.7 Sonnet scored 90% on the same test.

Despite these impressive numbers, Amazon has not yet tested Nova Act on widely accepted AI agent evaluations like WebVoyager, leaving some uncertainty about how it performs in real-world scenarios.

Who Built Nova Act? Meet the Minds Behind Amazon’s AI Agent

Nova Act is the first public AI agent to come out of Amazon’s AGI lab, which is led by David Luan and Pieter Abbeel—two former OpenAI researchers. Before joining Amazon, Luan founded Adept, while Abbeel co-founded Covariant, both startups focusing on AI-driven automation.

While some might find it surprising that Amazon’s AGI lab is focusing on developing an AI agent to order food or book reservations, Luan sees this as a stepping stone toward Artificial General Intelligence (AGI). According to him, a true AGI system should be capable of handling any task a human can perform on a computer, making AI agents like Nova Act an essential part of that journey.

What’s Next for Nova Act and Alexa+?

Amazon’s Nova Act AI agent enters a competitive market filled with promising but flawed AI agent technologies. Many early AI agents from OpenAI, Google, and Anthropic have struggled with slow response times, limited autonomy, and frequent errors in complex tasks.

The launch of Nova Act is critical for Amazon, as its success (or failure) will impact Alexa+, a major AI upgrade that Amazon hopes will redefine the smart assistant market. If Nova Act AI agent proves to be more reliable and effective than its competitors, Amazon could position itself as a leader in the AI automation space.

With early tests of Nova Act now underway, it won’t be long before we see whether Amazon has cracked the code—or if its AI agents will face the same hurdles as its rivals.

Developers can explore the Nova Act toolkit on the newly launched website, nova.amazon.com, which also highlights Amazon’s diverse Nova foundation models.

Get the Latest AI News on AI Content Minds Blog

Leave a Reply