Amazon has introduced a new general-purpose artificial intelligence agent called Nova Act. This agent can control a web browser and independently perform certain simple tasks. Alongside the new AI model, Amazon has also released the Nova Act SDK, a toolkit that allows developers to create their own agent prototypes with Nova Act.
Nova Act is a technology developed by Amazon's newly opened AGI lab in San Francisco and will be a key feature in the upcoming Alexa+ version of Amazon's popular voice assistant. Alexa+ is an AI-enhanced version of the voice assistant, but this version is still in research preview and not yet fully polished.
Through the Nova Act SDK, developers will be able to use new tools to automate simple tasks, such as navigating web pages, filling out forms, and selecting dates on a calendar. Amazon claims that Nova Act outperforms its competitors, OpenAI and Anthropic, in internal tests. For example, in the ScreenSpot Web Text test, Nova Act scored 94%, while OpenAI's agent scored 88%, and Anthropic's Claude 3.7 Sonnet scored 90%.
Nova Act is the first public product to emerge from Amazon's AGI lab, which is co-led by David Luan and Pieter Abbeel, both former OpenAI researchers. Luan and Abbeel emphasize that creating AI agents is a step towards building superintelligent systems, and Nova Act is designed to automate small, simple tasks.
While Amazon enters the AI agent market with this new technology, it acknowledges that there is still much ahead, especially with the Alexa+ project. The first tests of Nova Act show the potential for this technology to create even greater possibilities in the future.
