Garman told WIRED before the event that Amazon will also introduce a range of tools to help customers discuss generative AI models that he says are often too expensive, unreliable and unpredictable.
These include a way to increase the capabilities of smaller models using larger models, a system for managing hundreds of different AI agents, and a tool that provides proof that a chatbot’s output is correct. Amazon builds its own AI models to recommend products on its e-commerce platform and other tasks, but it primarily serves as a platform to help other companies create their own AI programs.
While Amazon doesn’t have a ChatGPT-type product to advertise its AI capabilities, the reach of its cloud services will give it an advantage in selling generative AI to others, says Steven Dickens, CEO and principal analyst at HyperFRAME Research. “The breadth of AWS will be something interesting,” he says.
Amazon’s own line of chips will help make the AI software it sells more affordable. “Silicon is going to have to be a key part of any hyperscaler’s strategy in the future,” says Dickens, referring to cloud providers that offer hardware to build larger, more capable AI. He also notes that Amazon has been developing its custom silicon longer than its competitors.
Garman says an increasing number of AWS customers are moving beyond demonstrations to creating commercially viable products and services that incorporate generative AI. “One of the things we’re really excited about is moving customers away from having AI experiments and proofs of concepts,” he told WIRED.
Garman says many customers are much less interested in pushing the boundaries of generative AI than in finding ways to make the technology cheaper and more reliable.
A recently announced AWS service called Model Distillation, for example, can produce a smaller model that is faster and less expensive to run while still having similar capabilities to a larger one. “Let’s say it’s an insurance company,” Garman says. “You can take a whole set of questions, feed them into a really advanced model, and then use that to train the smaller model to be an expert at those things.”
Another new cloud tool announced today, Bedrock Agents, can be used to create and manage so-called AI agents that automate useful tasks such as customer service, order processing, and analytics. It includes a master agent who will manage a team of AI subordinates, providing reports on how they are working and coordinating changes. “You can basically create an agent that says you’re the boss of all the other agents,” Garman says.