Amazon AWS, the cloud computing giant, has often been seen as trailing behind its competitors, Microsoft Azure and Google Cloud, especially in the rapidly evolving field of generative AI. However, at its annual AWS Re:Invent conference this week, Amazon aims to present a bold vision for generative AI that can assist businesses in developing innovative and unique applications using various models and data sources.
Swami Sivasubramanian, Amazon AWS’s vice president of data and AI who manages all AWS database, analytics, machine learning (ML), and generative AI services, shared some insights during an interview with VentureBeat. He previewed the content of his keynote on Wednesday morning and AWS CEO Adam Selipsky’s keynote on Tuesday.
The main focus of generative AI, according to Sivasubramanian, is giving businesses the flexibility to work with different models from multiple providers rather than being constrained to a single vendor or platform. However, he noted that having access to various models might not be sufficient for a competitive advantage as these models could become standard over time. Instead, the real edge comes from proprietary data and how businesses integrate this data with generative AI models to create distinctive applications.
To realize this vision, Amazon is concentrating on two main points at Re:Invent: offering a wide range of generative AI models accessible via its Bedrock service, and providing better, seamless data management tools for customers to build and deploy their own generative AI applications. Sivasubramanian emphasized the “inherent symbiotic relationship” between data and generative AI, highlighting how generative AI can both benefit from and enhance data systems.
Some key highlights for Re:Invent include:
Bedrock Apps Development: AWS’s Bedrock, launched in April, is a fully managed service allowing customers to use foundational generative AI models available through an API. Sivasubramanian noted that developing applications on Bedrock has become even more user-friendly, showcasing customer stories where applications were built in under a minute. Companies like Booking.com, Intuit, LexisNexis, and Bridgewater Associates are already leveraging Bedrock for impactful applications.
Expanded Model Choices: Through Bedrock, Amazon offers enterprise customers access to various models, including its Titan model and other third-party models like AI21’s Jurassic, Anthropic’s Claude, Meta’s Llama 2, and Stable Diffusion. Expect announcements about further investments in model diversity, including details about Amazon’s partnership with Anthropic, supported by a significant investment in September.
Vector Database Enhancements: Generative AI’s potential is substantial in vector databases, which facilitate semantic searches across unstructured data like images, text, and videos. In July, Amazon introduced Vector Engine for its OpenSearch Serverless in preview mode, showing significant traction. Sivasubramanian hinted at its impending general availability and the extension of vector search capabilities to other databases.
Generative AI Applications: AWS plans to unveil new generative AI applications and updates to existing ones, such as Amazon QuickSight and Amazon HealthScribe, designed for ease of use even for those without generative AI or coding expertise.
Zero ETL Vision: Integrating data from diverse sources without a complex ETL process remains a challenge. AWS has been investing in its zero ETL vision, opening new ways for businesses to store and query their vector data alongside other business data in their databases, exemplified by the recent vector search support addition to Amazon’s Aurora MySQL.
Secure AI Customization: AWS enables customers to customize generative AI models securely within their own virtual private cloud (VPC), preserving data security and privacy. This feature is a significant advantage over other cloud providers.
Generative AI Chip Innovations: AWS continues to develop its own silicon solutions for generative AI, including the Nitro hypervisor and the Graviton chip family for cloud computing, and the Trainium and Inferentia chips specialized for generative AI training and inference. Updates on these technologies’ performance and adoption will be provided.
As AWS continues to innovate and expand its offerings in generative AI, businesses can look forward to more flexible, secure, and efficient solutions to harness the full potential of this technology.