Generative artificial intelligence is a rapidly advancing technology that is merging hype and reality. It’s poised to have a huge impact on next-generation cloud technologies.

Companies are now expanding managed services offerings to address evolving customer needs, adapt to advancements in AI and navigate the growing complexity of multi-cloud environments. This expansion may also involve redefining services to better align with these changing landscapes. Amazon Web Services Inc. made a few important announcements over the past few months relating to new capabilities and products relating to Bedrock, CodeWhisperer, Cohere, Trainium and Inferentia.

“Bedrock hit [general availability] just sometime back, and we have added more models to Bedrock. Since the announcement earlier this year, we also added Cohere,” said Bratin Saha (pictured), vice president and general manager of AI and machine learning at AWS. “We are really excited at how customers are starting to use them to use the models in Bedrock. In addition, CodeWhisperer … relaunched the customization capability that allows customers to customize it for their own code basis and styles of coding. We are also very excited by what customers can do with Tranium and Inferentia.”

Saha spoke with theCUBE industry analyst John Furrier at Supercloud 4, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how AWS is advancing generative AI technology and offering customers a range of tools and models to build and deploy AI applications.

Background work

The combination of CodeWhisperer’s customization capability and purpose-built hardware infrastructure makes Bedrock the easiest and most efficient way to build generative AI applications, according to Saha. In essence, Bedrock makes it easier for companies to deploy models to production by handling the heavy lifting of training, optimizing, reducing inference cost and managing scalability while also allowing for the use of proprietary or fine-tuned models.

“There might be situations where customers actually build their own proprietary models or take a model and fine-tune it and then deploy it. There are two options there. One, if you go do it with Bedrock, you actually get your own private copy of the model,” Saha said. “When you want to fine-tune the model because you’re getting your own private copy of the model, your data doesn’t get into the base model. And that is where we have a number of features that we have added in SageMaker so that we can reduce the cost by an order of magnitude.”

Choice and flexibility

Choice and flexibility are important in managing infrastructure and integrating with APIs in the large language world, according to Saha. Different approaches, such as prompt optimization, prompt engineering, fine-tuning and specialized models, can be used based on the specific use case to achieve the desired accuracy level and reduce development costs.

AWS emphasizes the need for choice and flexibility in its models, as well as the importance of handling data and adapting data modeling to changes in embedding patterns and context, Saha added.

“We do a lot of that now where we are looking at the end-to-end, not just the model performance, and ensuring that we have the right data architecture to make sure that the data is being handled in the right way,” Saha said. “All of the basic security privacy stuff but also things like bias protection and toxicity and all of that needs to be eradicated.”

Watch the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Supercloud 4:

Photo: SiliconANGLE