Title

AWS re:Invent 2023 - Putting your data to work with generative AI | AIM250-INT

Summary

Mylon Thompson-Bukovec, VP of Technology at AWS, discusses the importance of using high-quality enterprise data to drive generative AI.
Generative AI can create new content based on learned data patterns and is transforming business operations.
The talk covers how to customize generative AI applications with business data and unpacks three critical data initiatives: prompt engineering, retrieval augmented generation (RAG), and fine-tuning or continuous pre-training.
Amazon Bedrock is highlighted as a tool that supports all three customization capabilities.
IDC forecasts a 22% growth in data over the next five years, with 90% being unstructured, emphasizing the need for high-quality data.
Adobe and Pinterest share their experiences with generative AI, focusing on data quality, training, and integration with existing data architectures.
AWS continues to innovate with new releases and capabilities to support generative AI workflows, emphasizing the importance of being your own best auditor for compliance and responsible AI.

The rapid growth of unstructured data presents both challenges and opportunities for generative AI, with a need for high-quality data to improve model accuracy and reliability.
Customizing generative AI models with enterprise data is becoming a common practice, with techniques like prompt engineering, RAG, and fine-tuning being essential for businesses to differentiate their AI systems.
Amazon Bedrock is a pivotal tool for customizing generative AI, offering support for various customization techniques and simplifying the process of integrating enterprise data.
Adobe's approach to generative AI emphasizes the importance of data quality and governance, using their Adobe Stock Marketplace as a foundational dataset and investing in data set management as a product.
Pinterest's use of generative AI for analytics productivity showcases the practical benefits of integrating AI with existing data architectures, resulting in significant productivity gains.
AWS's commitment to responsible AI is evident in their AI service cards, IP coverage for generative AI outputs, and new features like guardrails in Amazon Bedrock, which automate the adoption of responsible AI practices.
The talk underscores the importance of preparing for future regulations and compliance in the generative AI space by being proactive in auditing and responsible AI practices.