Amazon Vice President Data Storage Services and Cloud Migration, Mai-Lan Tomsen Bukovec, discusses the evolution of AWS storage, intelligent tiering impact on cloud-storage, preparing for generative AI, and ensuring customers adopt the right LLMs for datasets. They emphasize the shift to prioritizing data quality over volume and how companies leverage custom datasets for successful generative AI applications.
Data quality over volume determines business success in the modern era.
Intelligent tiering and adaptive storage solutions support generative AI applications efficiently.
Deep dives
Evolution of Amazon S3: Transition to Strong Consistency
The discussion covers the evolution of Amazon S3, particularly focusing on the transition from eventual consistency to strong consistency. Originally launched in 2006 as a storage service, S3 had eventual consistency, leading to delays in data retrieval. Due to increased data activity, S3 underwent a significant shift to strong consistency, ensuring immediate data retrieval upon successful addition. This evolution to strong consistency was a meticulous process involving reworking S3's core components without disrupting customer accessibility, showcasing AWS's commitment to enhancing services.
Intelligent Tearing for Data Lakes and Generative AI
The episode delves into the concept of intelligent tearing, a storage class introduced to address fluctuating data access patterns. Customers, particularly in data lakes, benefit from automatic cost-saving mechanisms based on data retrieval frequency. Intelligent tearing received rapid adoption due to its ability to adapt to varying data usage, making it a valuable tool for managing data sets efficiently for generative AI applications like pre-training models and handling intermediate data sets like prompt histories and checkpoints.
Securing Customer Data and Guarding Against Model Bias
The conversation emphasizes the importance of data privacy and security in utilizing third-party models like ChAg GPT. AWS ensures customer data confidentiality by preventing proprietary data from influencing the model and not saving prompt histories within AWS. The focus on transparency, highlighted through model and data cards, aids in assessing model accuracy, bias reduction, and compliance. AWS's dedication to enterprise AI security underpins trust and reliability in leveraging generative AI models.
Enhancing Enterprise AI with Bedrock and Custom Models
The discussion highlights the significance of evaluation criteria and model selection when utilizing open-source models like Meta's LOMA or Anthropic. Factors such as accuracy, bias reduction, latency, and cost form essential evaluation dimensions, influencing the choice of a generative AI model. AWS's Bedrock service offers flexibility in model evaluation, allowing comparisons and integration of custom models based on specific criteria, empowering users to easily navigate and adopt evolving generative AI technologies.
“Every modern business is a data business,” says Amazon Vice President Data Storage Services and Cloud Migration, Mai-Lan Tomsen Bukovec that “will be won or lost on the quality of the data, not the volume.” In this edition of Bloomberg Intelligence’s Tech Disruptors podcast, BI senior technology analyst Anurag Rana talks with Tomsen Bukovec about the evolution of AWS storage, specifically S3, and how features like intelligent tiering have shaped the cloud-storage landscape. They also discuss how the notion around storage is changing to support generative AI and what AWS is doing to make sure its customers are prepared to adopt the right LLMs for their datasets.
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode