Amazon Web Services (AWS) has unveiled several significant artificial intelligence developments at its re:Invent 2024 conference. The announcements focus on new hardware, software, and services designed to enhance AI capabilities for business customers.
The company introduced the general availability of its Trainium2 chips, which AWS claims are four times faster than their predecessors. These chips will be deployed in new EC2 Trn2 UltraServers, featuring 64 interconnected chips capable of delivering up to 83.2 petaflops of computing power. AWS also announced plans for Trainium3 chips, expected to launch in late 2025.
AWS is expanding its Amazon Connect service with new AI features, including customer segmentation tools and enhanced integration with Amazon Lex and Amazon Q. The update allows companies to process customer service interactions more efficiently and create automated marketing campaigns based on customer behavior patterns.
To address AI accuracy concerns, AWS launched Automated Reasoning checks, a new service within Bedrock that validates AI model responses by cross-referencing them with customer-supplied information. The company also introduced Model Distillation, allowing customers to transfer capabilities from larger to smaller AI models for cost-efficient operation.
The cloud provider announced multi-agent collaboration capabilities in Bedrock, enabling customers to assign AI agents to specific subtasks within larger projects. This includes the ability to designate supervisor agents that can coordinate multiple AI tasks and manage workflow orchestration.
AWS revealed that Amazon Connect now processes over 10 million contact center interactions daily for tens of thousands of organizations. The company also disclosed a partnership with Anthropic to build a massive computing cluster using hundreds of thousands of Trainium2 chips, which AWS states will be one of the world’s largest AI compute clusters.
Sources: Silicon Angle, TechCrunch, Wall Street Journal, TechCrunch, VentureBeat