Today we bring you a "short" summary of the key announcements and sessions in which the Micropole team took part!
GenAI on every floor
Let's start with AWS CEO Adam Selipsky's Keynote yesterday morning, announcing a host of new hardware and software features aimed at companies wishing to leverage AWS Cloud resources at the lowest cost to develop their AI and GenIA projects.
Selipsky then reminded the audience that a modern AI foundation is made up of 3 layers (Infrastructure / Tools / Apps), before going into the details of the announcements, the first of which concern the Infrastructure layer. For some time now, AWS has been investing in the creation of processors optimized for specific tasks, offering a good cost/performance ratio and a moderate carbon footprint.
On the menu, then, is the arrival of the Graviton Gen4 general-purpose processor (the first version dates back to 2018), between 30 and 40% faster than the G3, available in a new EC2 R8g instance type announced in the process. This type of machine is notably used by SAP to run large workloads on the AWS Cloud. Selipsky also announced the arrival of Trainium 2 chips, optimized for training deep neural networks (Deep Learning) and in particular LLM (Large Languages Models) generating inferences and FM (Foundations Models), and recalled the availability of the Inferencia 2 chip optimized for generating Deep Learning model inferences.
Another announcement designed to significantly improve performance is the arrival of Amazon S3 Express One Zone, a new class of S3 storage offering the best possible performance combined with very low latency, designed to accelerate very data-intensive activities, including Machine / Deep Learning model training or ad-hoc analytical querying of data stored on S3.
On the tools side, Amazon Bedrock (GenAI application creation) has seen its customization functionalities extended: after support for Fine Tuning, now comes RAG (Retrieval Augmented Generation) and Continuous retraining. Bedrock is now integrated into CloudWatch and CloudTrail and can provide recommendations. Also noteworthy is the ability to create policies for Bedrock AI models to meet the challenges of ethical AI...
But the announcement that undoubtedly received the best reception was that ofAmazon Q, a GenAI assistant designed to facilitate everyday tasks at work (more or less the equivalent of Microsoft's Copilot), trained on 17 years of AWS knowledge base. It can index numerous data sources, including Salesforce, Office 365, Box and Onedrive, as well as AWS data sources, and consolidates these elements in a vector database used by the agent to answer questions or provide advice.
provide suggestions. Amazon Q is already integrated with QuickSight (in preiew) and enables users to obtain recommendations for designing reports & Dashbaords, obtain narratives on graphs / KPIs, or create Stories to make QuickSight report presentations more punchy. These features are undoubtedly a major step towards the democratization of self-service BI.
With all these announcements, AWS is stepping up a gear, clearly demonstrating its ambitions in the highly promising GenAI market. But also its determination to democratize its use for all types of profiles, and to improve the operational efficiency of users in their day-to-day tasks.
One step closer to an Enterprise Data Driven
Being data-driven remains the holy grail for many companies. It's not just a buzzword. It's about creating an ecosystem where data is essential and ubiquitous, ensuring ease of access, fluid collaboration and strict compliance.
Imagine a world where self-service analysis is intuitive. Data is easy to find, understand and use.
Introducing AWS Datazone: a tool that redefines the way we catalog, discover and manage data in organizations. Datazone comes in the form of a domain portal, a one-stop shop with APIs that enable your analytics or BI solutions to connect easily, supported by a data catalog and projects that facilitate governance and access control.
Data resources (such as Glue and Redshift tables) are part of a vast, searchable catalog, enriched by a business glossary and technical metadata defining data ownership and meaning.
It's not just about the tools, it's also about the underlying vision. As such, Amazon Datazone is positioned as a key layer for a true Data Mesh approach.
With Datazone, creating, defining and then providing access to data products from one department to another is child's play, enabling direct access to data via tools such as Athena, QuickSight or Sagemaker, for example.
This is a real step forward in the democratization of data.