Generative AI Technology Leads AWS Agenda at re:Invent 2023

In an effort to upstage Microsoft in the AI space, AWS CEO Adam Selipsky invited NVIDIA CEO Jensen Huang and Dario Amodei, co-founder of Anthropic, to share the stage at AWS re:Invent 2023. Selipsky committed to integrating generative AI across AWS solutions and guardrails to secure customer data.

See Also: Live Webinar | Generative AI: Myths, Realities and Practical Use Cases

In a 150-minute keynote speech Tuesday, Selipsky announced several key partnerships and ventures with Salesforce, NVIDIA and Anthropic but did not discuss future investments or acquisition goals.

While Selipsky did not bring up the saga of OpenAI CEO Sam Altman’s firing by the company’s board and subsequent reinstatement last week, he did make a dig about major OpenAI investor Microsoft by splashing a screenshot highlighting that Microsoft had temporarily restricted the use of OpenAI tech because of security and privacy concerns.

Selipsky then announced a guardrails program for responsible AI and reiterated the company’s commitment to security and privacy of data. The new set of guardrails for Amazon’s Bedrock tool will help organizations keep a check on language that a model can use.

AWS-Salesforce Partnership to Focus on CRM

Selipsky gave a shoutout to AWS’s partnership with Salesforce and its plans to make the Salesforce Einstein AI platform available with Amazon Bedrock, the company’s foundational AI model. Businesses will now be able to use solutions such as Salesforce Prompt Builder, part of the Salesforce Einstein Copilot Studio, to create custom prompt templates grounded in their data.

“Together, we will make it easy for developers to securely access and leverage data and gen AI technologies to drive transformation for their organizations and industries,” Selipsky said. AWS will bring Salesforce products onto its marketplace. Salesforce will now support Amazon Bedrock as part of its open model ecosystem strategy.

NVIDIA to Power AWS AI Supercomputing

Huang, founder of NVIDIA, announced that AWS will become the first hyperscaler to bring NVIDIA GH200 Grace Hopper Superchips to the cloud. The platform will be available on Amazon EC2 instances supported by AWS Nitro System and hyperscale clustering, enabling customers to scale to thousands of GH200 Superchips.

“We will host NVIDIA DGX Cloud, the company’s AI factory on AWS infrastructure. DGX Cloud on AWS will accelerate training of gen AI and LLMs that can reach beyond 1 trillion parameters,” Huang said.

NVIDIA and AWS also will partner on Project Ceiba to design the world’s fastest GPU-powered AI supercomputer – an at-scale system.z. NVIDIA will use this supercomputer, which features 16,384 NVIDIA GH200 Superchips and capable of processing 65 exaflops of AI, to advance its next wave of generative AI innovation. While AWS also announced its own Trainium2 chip for training AI models, its partnership with NVIDIA is seen as a blow to other hyperscalers.

AWS Previews Anthropic-Powered Amazon Q AI Assistant

AWS announced the launch of its own AI assistant – Amazon Q – a new type of Anthropic-powered assistant that will “work for you at work. Initial applications for AWS developers are available for preview.

Amazon Q can leverage data from as many as 40 data sources – including M365, Slack, Salesforce, Dropbox and Amazon S3 – while maintaining extremely good privacy and security to stop unauthorized access, Selipsky said. “Trained on 17 years of knowledge, Amazon Q will be vastly different from existing competitive products. We are putting sufficient guardrails of responsible AI to address the security and data privacy concerns,” he said.

AWS is trying to shake off the perception in the market that it is a laggard in its AI vision compared to Google, with its recent $4 billion investment in Anthropic, and Microsoft, with the recent management upheaval at OpenAI.

Selipsky reiterated the company’s commitment to addressing the shortage of AI-skilled human resources and said its AI Ready initiative will help train 2 million people by 2025. The company has announced eight free courses to help people upskill in AI and generative AI. “We also have a gen AI scholarship program with online learning platform Udacity and a new collaboration with Code.org designed to help students learn about generative AI,” he said.