Yesterday marked the opening session of AWS re:Invent 2023 in Las Vegas – the annual conference covering all things AWS. If you’ve been reading my AWS News Roundup or listening to the LogiCast AWS News Podcast you’ll know that new product and service announcements have been ramping up over recent weeks in the ‘pre:Invent’ build up, but the big announcements are always saved for the main event, and AWS CEO Adam Selipsky never fails to deliver in his keynote.
So, what were the big announcements made by Adam Selipsky in his 2023 keynote?
1. Amazon S3 Express One Zone
Amazon S3 Express One Zone object storage was announced into General Availability – a new S3 storage class purpose built to offer the highest performance and lowest latency for customer’s most frequently accessed data. By leveraging this new storage tier, customers will be able to do away with custom caching solutions. S3 Express One Zone uses purpose built hardware and software to speed up data processing. Customers can choose the Availability Zone where they wish to deploy S3 Express One Zone, enabling the storage to be co-located alongside the compute that is being used to process the data to minimize latency. This new storage tier supports millions of requests per minute with single-digit millisecond latency, and has 50% lower access costs vs S3 Standard.
2. AWS Graviton4
Graviton4 is the most powerful and most energy efficient CPU that AWS has built to date. Graviton4 CPUs are 30% faster on average than Graviton3 CPUs, 40% faster for database applications and 45% faster for Java applications.
3. R8g Instances for EC2
R8g Instances were launched into Preview. R8g Instances are the first EC2 instances powered by the new Graviton4 CPUs. R8g is part of the memory optimised family designed for processing large datasets in memory. R8g instances provide the best price/performance and energy efficiency for memory intensive workloads.
4. AWS Trainium2
AWS Trainium2 was announced as ‘Coming Soon’. AWS Trainium2 is AWS’ second generation chip designed specifically for high performance training of Generative AI Large Foundation Models with billions to trillions of parameters. The first Trainium 2 based instances should be available in early 2024. Trainium 2 is 4 x faster than Trainium and can deliver 65 exaflops of on-demand super computing performance.
5. Amazon Bedrock Fine Tuning
A new Amazon Bedrock customisation capability – ‘Fine Tuning‘ was launched and is Generally Available for the Cohere Command Lite, Meta Llama 2 and Amazon Titan Text Lite & Express foundation models, and will soon be available for Anthropic Claude. Fine Tuning enables Amazon Bedrock customers to customise foundation models with their own data in just a few clicks.
6. Retrieval Augmented Generation (RAG) with Knowledge Bases
Retrieval Augmented Generation (RAG) with Knowledge Bases was launched into General Availability. Retrieval Augmented Generation is another technique for Amazon Bedrock customers to tailor foundation models for their businesses using labelled data. Customer data from multiple sources can be retrieved and fed in to the Amazon foundation models.
7. Continued Pre-training for Amazon Titan Text Lite and Express
Continued Pre-training for Amazon Titan Text Lite and Express was launched into Preview. This is a third technique for Amazon Bedrock customers to customise Bedrock foundation models with their own unlabelled data.
8. Agents for Amazon Bedrock
Agents for Amazon Bedrock are now Generally available. Agents for Amazon Bedrock enable users to execute multi-step tasks across company systems and data sources such as answering customer questions about product availability, taking sales orders and more.
9. Guardrails for Amazon Bedrock
Guardrails for Amazon Bedrock were announced into Preview. Guardrails for Amazon Bedrock enable users to safeguard their generative AI applications with responsible AI policies such as ensuring responses with no hate speech. Users can easily configure harmful content filtering, apply Guardrails to any foundation model or agent, and will soon be able to redact Personal Identifiable Information in foundation model responses.
10. Amazon Q
Amazon Q was launched into Preview. Amazon Q is a generative AI powered assistant for work that is tailored to customer’s businesses. Q lets you answer questions quickly with natural language interactions, informed by your own business data. Q has been designed to meet stringent enterprise security requirements from day 1, with an assurance that private customer data will never be used to train the underlying Large Language Models. Amazon Q can help developers build faster – Amazon Q has been trained on 17 years worth of AWS knowledge, and is now available in the AWS console and Slack – to help build out AWS architectures.
11. Amazon Q Architect
Amazon Q Architect can help users to build out a web app on AWS by suggesting AWS services which could be used. Users can interact and refine suggestions by asking natural language questions.
12. Amazon Q Code Transformation
Amazon Q Code Transformation enables developers to transform code in a fraction of the time that it currently takes. Users can complete Java language upgrades in a faction of the time, enhance code security posture and performance, and will soon be able to accelerate the migration from Windows to Linux. Amazon used Amazon Q Code Transformation to successfully upgrade 1000 Java applications from Java 8 to Java 17 in 2 days!
14. Amazon Q in Amazon Quicksight
Amazon Q is available in Amazon Quicksight, the Amazon business intelligence tool. Users can ask natural language questions of the business intelligence data, and then make natural language requests to refine the results.
14. Amazon Q in Amazon Connect
Amazon Q in Amazon Connect provides a generative AI contact center agent assistant which can generate recommended agent responses during real-time customer conversations – no typing requiered. Agents can use natural language search to find responses and links to more information, enabling faster agent onboarding with less training.
15. Zero ETL Integrations with Amazon Redshift
3 new Zero ETL integrations with Amazon Redshift were announced into preview – Amazon Aurora PostgreSQL, Amazon RDS for MySQL and Amazon DynamoDB.
16. Amazon DynamoDB zero-ETL integration with Amazon OpenSearch Service
Amazon DynamoDB zero-ETL integration with Amazon OpenSearch Service was launched in to General Availability. This new integration makes it easy to search DynamoDB data with Amazon OpenSearch Service.
17. Amazon DataZone AI Recommendations
Amazon DataZone AI recommendations was launched into Preview. Amazon DataZone AI Recommendaitons automatically add business context to customer data catalogs in a few clicks. Amazon DataZone AI Recommendations uses generative AI to create business descriptions for customer data, provides recommendations on how to use data for faster insights, and makes data immediately easier to understand and discover.
18. Amazon Project Kuiper
Amazon Project Kuiper is building a constellation of thousands of low earth orbit satellites which will close the digital divide by bring affordable high speed internet to areas where it is currently unavailable. The first 2 prototype satellites have been launched in the last month and all is so far working well.
So, there you have it – if you missed the keynote, I hope my summary has helped you to catch up with all of the key annoucements.