Amazon's AGI Lab Reveals Its First Work: Advanced AI Agents
1 min read
Summary
Amazon has unveiled its new AI model, Amazon Nova Act, which it says outperforms those from OpenAI and Anthropic on a variety of benchmarks.
Nova Act is a more advanced version of the company’s Nova model, with additional training to improve its decision-making around actions.
To enhance its Nova model’s capabilities, Amazon is using reinforcement learning, a technique that has successfully enabled other AI models to better simulate reasoning.
Amazon’s new model aims to make AI agents more reliable and less prone to error than those built by its competitors, which often rely on large language models and numerous human-designed rules.
The company has also announced a software development kit (SDK) to make it easier for computer engineers to use Nova Act to build software agents.
This will allow engineers to give their agent specific instructions, such as “don’t accept the insurance upsell,” when booking a rental car.
