New AI features unveiled at AWS ‘re:Invent’ conference

Amazon Web Services is tackling AI challenges head-on with new tools that improve model accuracy and lower costs, unveiled at the 2024 re:Invent conference.

Amazon focuses on faster speeds and lower costs in its new AI offerings.

At its re:Invent 2024 conference, Amazon Web Services (AWS) announced groundbreaking tools aimed at addressing common issues in generative AI, including hallucinations. The new Automated Reasoning Checks service verifies the accuracy of AI-generated content by cross-referencing customer-provided information. This tool is integrated into AWS’s Bedrock platform and allows users to refine models with “ground truths” for better reliability, though it shares similarities with earlier offerings from Microsoft and Google.

AWS also introduced Model Distillation, a feature for Bedrock that enables users to transfer capabilities from larger AI models to smaller, more cost-efficient ones. While this helps lower costs, it comes with limitations, such as requiring models from the same family and a slight dip in accuracy. Additionally, Bedrock now offers a multi-agent collaboration feature, which lets customers assign AI agents specific subtasks, improving efficiency in larger projects.

These innovations reflect AWS’s commitment to staying ahead in the competitive AI space. The new features aim to address industry-wide concerns about AI’s reliability and cost while expanding Bedrock’s capabilities for customers like PwC, whose VP of AI and data, Swami Sivasubramanian, highlighted significant growth in the platform’s user base over the past year.