Amazon Bedrock, a tool for using generated AI on AWS, is officially released, Claude who can speak Japanese and Stable Diffusion for image generation can be used



``Amazon Bedrock,'' a tool that allows you to use generated AI on AWS

, which was announced in April 2023, was officially released on September 28, 2023, and has started offering to general users.

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


Amazon Bedrock Is Now Generally Available – Build and Scale Generative AI Applications with Foundation Models | AWS News Blog
https://aws.amazon.com/jp/blogs/aws/amazon-bedrock-is-now-generally-available-build-and-scale-generative-ai-applications-with-foundation-models/


Amazon Bedrock is a tool for leveraging serverless generative AI that requires no infrastructure management. Since the service is provided on AWS, it is possible to monitor usage and API status using Amazon CloudWatch and AWS CloudTrail .

To access Amazon Bedrock, enter 'bedrock' in the search field on the console screen and click 'Amazon Bedrock' displayed in the service field.



It is not available in the Tokyo region, so you will need to switch to one of the following regions: Oregon, Singapore, Northern Virginia, or Ohio.



The menu in Amazon Bedrock looks like this.



In 'Examples' you can check examples of settings tailored to your use case.



In 'Playground' you can actually try using AI models for chat, text, and images.



You can also access Amazon Bedrock via the API. For example, the code to display the available underlying models using

Boto3, an SDK for Python, is as follows.
[code]import boto3
import json

bedrock = boto3.client(
service_name='bedrock',
region_name='us-east-1'
)

bedrock.list_foundation_models()[/code]



The basic models listed at the time of article creation are as follows.

・Claude v1.3
・Claude v2
・Claude Instant v1.2
・Command v14.6
・Jurassic-2 Mid v1
・Jurassic-2 Ultra v1
・Stable Diffusion XL v0.8-preview
・Titan Embeddings G1 - text v1.2
・Titan Text G1 - Express v1-preview



Click on each model to see usage fees, token limit, available languages, supported use cases, etc.



Among the above list, only Claude v2 supports chat in Japanese. According to

Rakuda Benchmark , which evaluates AI's Japanese language ability, Claude v2 has a performance that is between GPT-3.5 and GPT-4.



However, at the time of writing the article, Claude was 'currently unavailable' and it was not possible to confirm its performance.



Usage fees are set in detail for each model, so it would be a good idea to select the model you want to use according to your purpose and then check the pricing page . For example, in the case of Claude, it was $0.01102 (about 1.65 yen) per 1000 input tokens & $0.03268 (about 4.88 yen) per 1000 output tokens.

in Software,   Web Service, Posted by log1d_ts