AWS has announced several news related to generative AIamong which are the general availability of Amazon Bedrock and Titan Embeddings. Also the upcoming availability of Llama 2 on Bedrock and new generative AI features for CodeWhisperer and QuickSight.
The Amazon Bedrock managed service offers various high-performance foundational models developed and trained by the most prominent companies in the Artificial Intelligence sector. Among them AI21 Labs, Anthropic, Cohere, Meta, Stability AI and Amazon. Added to this are various features and options for generative AI development, focused on simplifying development while maintaining both privacy and security.
In addition to experimenting with several foundational models for a wide variety of use cases, users can customize the models privately, using their proprietary data. Bedrock also offers, free of charge, an online course so that students and people who want to train in generative AI can acquire the knowledge they need to start using it.
Amazon Bedrock will be the first fully managed generative AI service to offer the Meta Llama 2 large language model, which it will do through a managed API. It will be available in the coming weeks, optimized for rapid response on AWS infrastructure. Customers will be able to develop generative AI applications with the Llama 2 models with 13,000 and 70,000 million parameters, without having to configure or manage an infrastructure.
Also new is the availability of Amazon Titan Embeddings, which makes it easy for customers to get started with Recovery Augmented Generation (RAG) to extend the power of a foundational model with their proprietary data.
Amazon Titan’s foundational models have been built by AWS on large data sets, increasing their power and making them capable of general-purpose operation. The first of these models is the aforementioned Amazon Titan Embeddings. This is a large language model that passes text into numerical representations, known as embeddings. Used to power RAG use cases
For its part, Amazon CodeWhisperer adds the ability to securely customize CodeWhisperer suggestions from its private code base, with the aim of helping developers to be more productive in their task. This new service personalization capability will increase your generative AI-powered coding potential by securely leveraging the internal code base and customer resources to deliver personalized recommendations. Code hints are also more relevant in various tasks.
Amazon QuickSight from now on you will have new generative business intelligence (BI) auditing capabilities. Thus, business analysts will have support to generate and customize visual elements and create graphs with ease, from commands issued in natural language. With this new feature, it is enough to describe the desired result using fragments of questions for QuickSight to generate attractive visual elements that can be attached to a control panel or a report with one click.
