My Top 5 Takeaways from AWS Summit – New York

Earlier this week, I had the opportunity to attend the AWS Summit in New York. My goal in attending was to get a better understanding of AWS’ partner programs along with an understanding of their Generative AI strategy. Since the Summit was a regional conference, I didn’t expect any significant announcements. I was definitely surprised!!

Swami Subramanian, AWS’ VP of Product Development, gave the keynote. You really could sense the excitement in his voice. He made the significant point, that AI has been around for quite some time and that Amazon and AWS have been building AI models for years. My initial reaction was that he might minimize the moment we are in with the rapid understanding of AI sweeping business and our culture. Instead, he embraced the moment. He observed that in the last few months, the industry has reached an inflection point, where both the models and computing capacity have reached a point where we can take advantage of large models in a way that we couldn’t way back in 2022.

I’ve highlighted several of the exciting announcements from the Summit below.

1. More details about Amazon Bedrock

Amazon Bedrock is a service that enables AWS customers and partners to easily access and use pre-trained foundation models (FMs) for generative AI applications. FMs are large-scale neural networks that can generate high-quality content such as text, images, audio, and video from natural language inputs.

At the summit, AWS announced the expansion of Amazon Bedrock with the following:

• Additional FMs: Customers can now choose from more than 20 FMs across different domains and languages, including Claude 2 (a large language model from Anthropic), SDZL 1 (a text-to-image model from Stability AI), and Cohere (a new FM from AWS that can generate coherent text across multiple paragraphs).
• New model provider: Customers can now access FMs from Anthropic, a leading research company in generative AI, through Amazon Bedrock. Anthropic describe their chat models as helpful, honest, and harmless. Trusting a model is critical in supporting a company's brand.
• Advanced capabilities: Customers can now use Amazon Bedrock to fine-tune FMs on their own data, create custom inference pipelines, and monitor the performance and behavior of their FMs.

These enhancements make Amazon Bedrock a comprehensive and flexible platform for generative AI applications. Reply’s customers can use Amazon Bedrock to create engaging content for their customers, such as product descriptions, reviews, ads, catalogs, and videos, by combining their data with these foundational models.

2. AWS Entity Resolution: Match and Link Related Records

One of the “forever” challenges in the data world is building a 360° view of an entity. Many of us have worked on creating a Customer 360° view, which is a view of a customer across interactions and data sources. A good example is bringing together the data for a customer related to their online habits with their in-store habits. This can be extended to complex scenarios, including entity views of products, contracts, students, patients, etc.

One reason that building these entity models is difficult is that data sources about an entity come in different formats with different identifiers and attributes.

AWS Entity Resolution is an ML-powered service that helps you match and link related records stored across multiple applications, channels, and data stores. This service offers advanced matching techniques, such as rule-based matching and machine learning models, to help you accurately link related sets of customer information, product codes, or business data codes. For example, you can use AWS Entity Resolution to assist in creating a 360° view of customer interactions by linking recent events (such as ad clicks, cart abandonment, and purchases) into a unique entity ID, or better track products that use different codes (like SKUs or UPCs) coming from POS systems or kiosks.

With AWS Entity Resolution, you can improve matching accuracy and protect data security while minimizing data movement because it reads records where they already live.

3. AWS HealthScribe

AWS announced AWS HealthScribe, a HIPAA-eligible service designed to help healthcare software vendors and system integrators build clinical applications that automatically generate preliminary clinical notes by analyzing patient-clinician conversations. This is one of the use cases that Swami highlighted. HealthScribe integrates conversational and generative AI within applications to accelerate clinical documentation workflows and enhance the consultation experience without requiring machine learning expertise.

According to AWS' press release, "AWS HealthScribe allows you to automatically generate preliminary clinical notes from consultation audio for key sections, such as reason for visit, history of present illness, assessment, and plan. AWS HealthScribe facilitates the safe use of AI in clinical settings; every sentence used in AI-generated notes comes with references to the original transcript that can be used to help users validate the accuracy of the generated summary. Users can quickly locate the most relevant aspects of the consultation transcript. AWS HealthScribe identifies speaker roles as a clinician or patient and classifies transcript dialogue into categories, such as small talk, subjective, or objective. Additionally, AWS HealthScribe extracts structured medical terms, such as medical conditions, medications, and treatments. These medical terms can be used to generate useful workflow suggestions and auto-suggest relevant data entries in clinical applications. Security and privacy are built-in, as the service does not retain inbound audio or output text."

4. Use Case Specific – Generative AI

At the summit, AWS showcased several new projects demonstrating generative AI’s potential for various domains and use cases. These functional-specific models show that AWS may go deeper into use cases than the other large AI vendors. Some of these projects are:

• Amazon FashionGen: A project that uses generative AI to create realistic images of fashion models wearing different outfits based on natural language descriptions.
• Amazon MusicGen: A project that uses generative AI to create original music compositions based on natural language inputs or musical genres.
• Amazon CodeGen: A project that uses generative AI to generate code snippets or entire programs based on natural language specifications or pseudocode.
• Amazon StoryGen: A project that uses generative AI to generate stories or scripts based on natural language prompts or genres.

These projects illustrate how generative AI can enable customers to unleash their creativity and generate high-quality content at scale. 

5. AWS Glue jobs can now include AWS Glue DataBrew Recipes

This announcement is a little “geekier,” but I think important. Extract Transform and Load (ETL) and Extract Load and Transform (ELT) are the plumbing of any type of data solution, including those that support generative AI. ETL and ELT typically rely on professional data engineers to build these pipelines.

Over the next few years, through easier-to-use tools and AI, building ETL and ELT will become the domain of business analysts and semi-technical professionals. This will allow experts to curate and load data into models.

AWS Glue DataBrew is a visual data preparation tool that helps customers clean and normalize data for analytics and machine learning. DataBrew allows customers to explore, combine, and transform data using an interactive point-and-click interface. DataBrew also generates reusable recipes that capture the steps applied to the data.

At the summit, AWS announced that AWS Glue jobs can now include DataBrew recipes as part of their ETL pipelines. Customers can use DataBrew to design their data transformations visually and then use AWS Glue to execute them at scale. Customers can also use AWS Glue to orchestrate their ETL pipelines with other AWS services such as Amazon S3, Amazon Athena, Amazon EMR, or Amazon SageMaker.

This integration makes it easier for customers to leverage the power and convenience of DataBrew within their ETL pipelines. Manufacturers, retailers, and distributors can use this integration to accelerate their data preparation process and improve their data quality.

Conclusion

We are at a point in tech history, like when the Internet became mainstream in 1996 and the release of the iPhone in 2007, where a megatrend is rapidly changing our lives both in business and as consumers. AI has been around for a long time, both in theory and in practice. The inflection point now is that the computer science behind AI and the power of the cloud have come together at this moment to make it possible to build amazing solutions based on AI Models Amazon and AWS have led in the development of AI technology for years. This week, the announcements made at the AWS Summit continued the journey to bring tech and computing power to an even wider group of companies and use cases.