Two chips, a new AI assistant & AI safeguards — Amazon's latest launches could change the way we work with AI

The headline launch from the Amazon Web Services (AWS) stables was Amazon Q, the company's new generative AI cloud assistant, tailor-made for individual needs of various businesses. "Amazon Q is a new type of a generative AI cloud assistant designed to work for you, at work," said Selipsky, in his keynote. "Q lets you answer questions quickly, with natural language interactions that let you chat, create content and take action — all informed by understanding your system, your data repositories and your operations," he said.

The AWS CEO added that the new AI assistant was also designed to understand "rock-solid security and privacy," which were critical in determining permissions to access data that vary across multiple roles and hierarchies within companies. "If a user doesn't have permission to access something without Q, they cannot access it with Q either," said Selipsky, adding that he truly believed this is going to be transformative. "We want lots of different kinds of people who do different kinds of work to benefit from Amazon Q," he said.

Signaling the launch of the next-generation AWS-designed chips, 'Trainium2' and 'Gravitron4', Selipsky pointed that both AWS processors would go a long way in artificial intelligence (AI) and machine learning (ML) training at a faster pace, while consuming lesser energy. "The AWS Trainium2 is our purpose-built chip for generative AI and ML training. It is optimised for training foundation models with hundres of billions to trillions of parameters, and is 4 times faster than (its predecessor) AWS Trainium," he said.

Simiarly, the newly launched AWS Gravitron2, Selipsky claimed is the company's "most powerful and energy-efficient processor till date", even as AWS has declared it to be 20% faster than Gravitron3, with speeds seeing a 40% and 45% spike for database applications and large Java applications, respectively.

What many regard as the most timely and much-needed product launch in these times was also unveiled in Selipsky's keynote address. 'Guardrails' is AWS' latest generative AI application safeguard that can align itself with various companies' respective responsible AI policies. Its launch comes at a time when there is considerable debate over whether AI companies must self-regulate content or be subject to a policy-level regulation of AI output.

In a nutshell, 'Guardrails' has the ability to configure harmful content filtration based on each organisation's responsible AI policies. "An important component for Responsible AI is promoting interactions between consumers and applications where the output stays in line with a company's guidelines. The easiest way to do this is to place limits on what information can be returned," Selipsky said.

He added that Guardrails helps safeguard one's generative AI applications with responsible AI policies. "Using natural language descriptions, you can choose what you'd like the AI model to avoid. For example: a bank can configure an online assistant to refrain from providing investment advice; or an e-commerce site can ensure its online assistant doesn't use hate speech or insults," he said.