Enterprises today can’t go a day without talking about (or hearing about) AI.
Every organization is exploring its options and opportunities with the technology, particularly generative artificial intelligence (AI). Amazon Web Services (AWS), which strives to be an AI innovator, today furthered that mission with the release of several new gen AI tools at AWS Summit: New York.
“Generative AI has captured our imaginations,” Swami Sivasubramanian, AWS VP of database, analytics and machine learning (ML) said in a keynote at the event. “This technology has reached its tipping point.”
AWS is ultimately working to democratize access to generative AI so enterprises can “reimagine experiences and bring new products to life,” he said.
Stronger BedrockAmazon’s fully managed foundation model (FM) service Amazon Bedrock provides access to models from leading companies via a simple application programming interface (API). This allows customers to quickly find and test models, secure and customize them, then integrate and deploy them into production.
Today, AWS announced the expansion of Bedrock, adding Cohere as a provider and incorporating the latest FMs from Anthropic (Claude 2) and Stability AI (Stable Diffusion XL 1.0). Bedrock also now includes a feature that allows user to create fully-managed agents without the need for expertise.
“Generative AI has the potential to transform every application, business and industry,” Sivasubramanian said in a statement. He added that advancements across data processing, compute and ML are “expediting the shift from experimentation to deployment.”
Cohere’s Command text generation model is trained to follow user prompts and return summaries, copy and dialogue. The tool can also extract information and answer questions. Meanwhile, the company’s text understanding model Embed can be used for search, clustering or classification tasks across more than 100 languages.
Anthropic’s Claude 2, for its part, can take up to 100,000 tokens in each conversational task prompt, “meaning it can work over hundreds of pages of text, or even an entire book,” according to Amazon. The tool can also write longer documents such as memos and stories “on the order of a few thousand tokens.”
Furthermore, Stability AI’s models can create text, images, audio, video and code through simple text instructions. The model’s most recent iteration features enhanced image and composition detail that can be used to generate more realistic creations for films, television, music and instructional videos, according to AWS.
The integration of Stable Diffusion provides Amazon Bedrock customers “access to cutting-edge resources and advances our goal to activate humanity’s potential with AI,” said Emad Mostaque, Stability AI founder and CEO.
A supportive AI agentA new capability Agents for Bedrock allows users to build fully-managed agents “in a few clicks” and without the need for coding, according to AWS. This can accelerate the development of gen AI applications manage and perform tasks by making API calls to company systems.
The tool allows models to understand user requests, break tasks into step-by-step guides and even carry on conversations to collect additional info, fulfill requests and generate orchestration plans.
Insurance company Travelers is an early adopter of the Agents tool. “The potential that AI brings is huge,” said Mano Mannoochahr, Travelers’ chief data and analytics officer. With direct access to models from AWS and others, “Amazon Bedrock provides the potential for quick and easy experimentation, development and deployment.”
Tony Grout, chief product officer at sales enablement platform Showpad, meanwhile, emphasized the importance of bringing value to buyer interactions.
“This requires tailoring information and improving how sellers engage with buyers based on their unique needs,” he said.
His company uses Amazon Bedrock to experiment and push new models to production “so we can ensure that every conversation is empathetic, authentic and builds trust with buyers.”
AWS Entity Resolutions unifies data, workflowsEnterprise records are often siloed in dozens, if not hundreds, of different apps, channels and data stores. Linking them is often a “home-grown” process involving complex data pipelines and external partner integrations — and deriving the most impactful insights can be cumbersome.
To support customers in this area, AWS has announced the general availability of AWS Entity Resolution. The ML-powered analytics service helps organizations analyze, match and link related records stored across different apps, channels and data stores. Users can generate customizable workflows to join related consumer, business and product information.
This allows businesses to better understand how their data is related, matched and linked to support decision-making, establish a clearer and more efficient supply chain, craft more relevant marketing campaigns and improve financial decisions, according to AWS.
The company also plans to integrate LiveRamp and TransUnion and provide interoperability with the Unified ID 2.0 open source framework.
“AWS customers across every industry rely on accurate data for daily decision-making to drive better business outcomes,” said Dilip Kumar, VP of AWS Applications.
With AWS Entity Resolution, he explained, organizations can match records and link flexible and scalable workflows that easily connect to existing applications.
New generative BI capabilitiesBusiness intelligence (BI) is critical in modern enterprise. To help enterprises explore their data and discover and share insights, AWS introduced Amazon QuickSight Q in 2020. The tool allows any user to ask questions about their data in natural language and provides interactive dashboards, paginated reports and embedded analytics.
Today, the company announced that it is furthering QuickSight Q with new large language model capabilities available through Amazon Bedrock. These will allow users to create visuals, fine-tune and format those visuals and create calculations with natural language and without needing to know specific syntax.
QuickSight Q also has a new feature, Stories, which helps users generate, customize and share “compelling visual narratives” with natural language prompts, according to AWS.
Enhanced search, improved training timeAWS has also launched a preview release of a vector engine for Amazon OpenSearch Serverless. This provides a simple, scalable, high-performing similarity search capability in Amazon OpenSearch Serverless.
Users can build ML search experiences and generative AI applications without having to manage underlying vector database infrastructure, AWS says.
Furthermore, the general availability of Amazon EC2 P5 instances addresses customer demand for high performance scalability in AI/ML and high performance computing (HPC) workloads.
The next-generation GPU instances are powered by the latest NVIDIA H100 Tensor Core GPUs and will provide a reduction of up to 6 times in training time (from days to hours), a performance increase that will lower training costs by 40%, Amazon claims.
Comments