Here is a selection of the top Day 3 releases from AWS re:Invent 2023!
Amazon Titan Image Generator in Amazon Bedrock
The Amazon Titan Image Generator foundation model has been released within Amazon Bedrock to generate high quality images using natural language prompts. It supports various methods of editing the image after creation including configuring image dimensions and changing backgrounds. What is particularly interesting is that you can customise the model with proprietary data to be able to generate images within certain brand design guidelines!
Amazon OpenSearch Zero-ETL Integration with S3
A preview of Amazon OpenSearch zero-ETL integration with S3 has been released. This one looks super powerful and simplifies the querying of operational data and logs that are stored in S3. Up until now, it’s been a bit of a complex process to set up querying of semi-structured data in S3 using Athena, so this new integration looks to be incredibly useful, saving massive amounts of time!
Easier Model Comparison in Amazon Bedrock
Comparing models in Amazon Bedrock is now much easier, with evaluation tools now available in preview. This will allow experimentation with the available foundation models much easier in a playground environment, which will speed up your time to business value.
Responsible AI Innovation with Guardrails
AWS released new tools and capabilities to enable responsible AI innovation with guardrails for Bedrock, for example, tools to redact PII supplied as input and content filters. This will help tremendously with building out applications fast but maintaining protections and safeguards.
Vector Engines for AWS Services
Building out GenAI applications requires some sort of vector engine to store vector embeddings so it makes sense that a bunch of vector engines were released for various AWS services, including Amazon OpenSearch Serverless, Amazon DocumentDB, Amazon MemoryDB for Redis. Being able to store this data in existing data stores is very powerful and makes your querying much more efficient than having to query multiple data stores and trying to consolidate all the information manually.
A great blog post describing all the new Amazon AI capabilities is here.
