AWS CTO outlines the four steps to unlock GenAI’s potential

Published on the 20/09/2023 | Written by Heather Wright

AWS CTO outlines the four steps to unlock GenAI's potential

Choice, flexibility, democratisation…

Generative AI will transform industry, business and applications across four key categories, but there are also four key steps required to unlock the potential of the technology.

That’s according to Rada Stanic, AWS chief technologist, who says AWS believes generative AI will transform ‘every single industry, every business and every application’.

But Stanic, speaking at Tech Leaders 2023 recently, admits the technology isn’t new. AWS has been using AI for more than 25 years for personalisation on its Amazon platform, where 4,000 purchases a second are made; in robotics-powered shipping processes from its fulfilment centre and billions of Alexa interactions each week.

“We are really not strangers to this technology”

The proliferation of data and availability of compute capacity to train and fine-tune the models, along with constant innovation, has helped create what she calls an inflection point. 

“We are really not strangers to this technology. It is just that it is now starting to shape every business, every industry and every application.”

While it’s ChatGPT that has captured the imagination of many, and brought generative AI to the masses, Stanic notes that it is just one ‘humble, simple example of how technology will help transform everything’. 

“We see four key categories where generative AI will truly transform industries and businesses,” she says.

Enhanced customer experience through virtual assistants and chatbots is one area Stanic sees transformation. 

While self-service options for customers have been touted, others in the industry have noted the potential too for AI bots to ‘listen in’ on agents’ conversations and feed through relevant data and resources to support the conversation. 

Stanic says employee productivity is also a key area, noting that for developers, AWS has launched Code Whisperer, a code companion that takes plain language inputs and creates code. 

“We’re not talking about replacing humans. We’re really talking about improving the productivity and helping them focus on more creative tasks,” Stanic says. 

Bolstering creativity and content creation and improving business operations via document processing, process optimisation, cybersecurity and data augmentation round out the top four categories for transformation, according to Stanic. 

On the creativity front she notes the example of Canva’s launch of a text to image feature, powered by a large language model (LLM) from Stable Diffusion, late last year. 

“ChatGPT and OpenAI have captured everyone’s imagination and shown what’s possible. But while we don’t know what the future holds, we do know that there won’t be a single LLM to rule them all,” she notes. 

“So choice and breadth is very, very important, and also the flexibility to build,” she says.

AWS, which claims to have 100,000 customers already using AI, has been positioning itself to enable customers to use a wide range of models – both Amazon’s own offerings and those from vendors including Anthropic and Stability AI. 

“The second point is providing a platform that can perform and perform in a cost-efficient way, because you need to have capacity to train and build these models and to fine-tune them. But the adoption will only be as strong as the cost effectiveness of that platform.”

Stanic says democratising the technology and making it easy for people to develop applications powered by generative AI is also necessary to unlock the full potential of the technology. 

“Everyone is talking about the skills shortages, so how can we make this accessible and usable by as many people as possible?”

Security forms the fourth requirement.

“For us at AWS we have this saying: ‘It’s a day zero job’. So before allowing and helping our customers to do anything with any kind of technology, we make sure that it’s secure. 

“In the context of generative AI that means first of all protecting and securing intellectual property for all those model providers who will be critical to providing the breadth of choice. 

“It’s also about protecting customers’ data. Making sure it doesn’t leave the context of their virtual private environment and that it’s also not used to train these models that they are using. That’s so very important.”

She highlighted several offerings AWS has debuted in recent months, including Amazon Bedrock which allows any developer to make a single API call to an LLM deployed in the AWS environment and to use the output of the query to embed it in the application.

For more complex problems, such as booking flights or processing insurance claims, Agents for Bedrock will enable developers to orchestrate a series of simpler tasks to solve more complex outcomes.

And for the health sector, AWS has launched AWS Health Scribe, using GenAI-powered speech recognition to produces transcriptions and clinical notes for medical professionals.

“With all the focus on responsible AI we have actually built that into the service. So for every AI generated sentence in this transcript there is a reference to the original script to ensure accuracy, which is one small example of responsible AI framework.”

Post a comment or question...

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.


Thank you! Your subscription has been confirmed. You'll hear from us soon.
Follow iStart to keep up to date with the latest news and views...