Amazon announces 'Bedrock', a service for building generative AI, and 'Titan FM', a basic model for AI

by Tony Webster

Amazon Web Services (AWS), Amazon's cloud computing service, has announced the cloud service ' Amazon Bedrock ' for developing applications that use generative AI for companies and developers. At the same time, the existence of ' Amazon Titan FM ', which is Amazon's own fundamental model for generative AI (Fundamental Model), was also clarified.

Announcing New Tools for Building with Generative AI on AWS | AWS Machine Learning Blog
https://aws.amazon.com/jp/blogs/machine-learning/announcing-new-tools-for-building-with-generative-ai-on-aws/

Foundation Model API Service – Amazon Bedrock – AWS
https://aws.amazon.com/jp/bedrock/

Foundation Model-Amazon Titan-AWS
https://aws.amazon.com/jp/bedrock/titan/

AWS launches Bedrock generative AI service, Titan LLMs
https://www.cnbc.com/2023/04/13/aws-launches-bedrock-generative-ai-service-titan-llms.html

According to AWS, Bedrock is part of Stability AI , which develops the image generation AI 'Stable Diffusion,' Anthropic , which develops the interactive AI 'Claude,' AI21 Labs , which develops the text generation AI 'Wordtune Spices,' and Amazon's basic model. It is a service that can be accessed via API. By being able to access the latest basic models provided by each company, AI services such as text generation AI, chatbots, search systems, text summarization, and image generation can be developed.

AWS says, 'Bedrock's serverless experience allows you to get started quickly without having to manage any infrastructure. You can also customize the underlying model to your liking and scaleexperiments to test different models and scale the underlying model.' You can easily integrate and deploy your applications using familiar AWS tools and features, such as integration with Amazon SageMaker such as Pipelines managed by .

Bedrock has already been offered in the form of a limited preview, and some customers such as AI company Coda are using it in advance. “As a long-time AWS customer, we are excited that Bedrock can bring quality, scalability, and performance to Coda’s AI,” said Shishir Mehrotra, co-founder and CEO of Coda. All on AWS, Bedrock enables rapid adoption of generative AI, with all the security and privacy you need to protect your data. With tens of thousands of teams, including tens of thousands, running on Coda, reliability and scalability are really important.'

In addition, Amazon announced 'Titan FM' as one of the foundational models available for this Bedrock. At the time of writing the article, Titan FM has ' Titan Text ', a generative large-scale language model suitable for tasks such as summarization, text generation, and information extraction, and an embedded large-scale language model that converts text input into numerical representation. It seems that two of ' Titan Embeddings ' will be prepared. AWS has revealed that this Titan Embeddings is already used for Amazon.com's product search function.

Amazon CEO Andy Jassy told economic media CNBC, ``Most companies want to use large-scale language models, but really good models cost billions of dollars and take years to train. This is why most companies don't try to develop a foundational model, i.e. they start with a good foundational model that is already built and customize it for their own purposes. We want to be able to move forward with development, and Bedrock fulfills that desire.'

Amazon also announced the general availability of Amazon EC2 Trn1n instances powered by its proprietary second-generation machine learning chip , AWS Trainium , and Amazon EC2 Inf2 instances powered by the high-performance AWS Inferentia2 accelerator for generative AI. . In addition, it was announced that the generative AI-based coding support service ' Amazon CodeWhisperer ' will be made available to the public free of charge. Amazon CodeWhisperer supports Python, Java, JavaScript, TypeScript, C#, Go, Kotlin, Rust, PHP, SQL, and can be used from IDEs (integrated development environments) such as VS Code, IntelliJ IDEA, and AWS Cloud9 via AWS toolkits. It is said that it will be accessible.

Amazon said, 'Building powerful applications like Amazon CodeWhisperer is transformative for developers and all of our customers. We are excited about what you are building with generative AI on AWS.' Our mission is to enable developers of all skill levels and companies of all sizes to innovate with generative AI, unlocking your potential with the next wave of machine learning. It's just the beginning of what we believe will bring out.'