The recent AWS re:Invent 2023 event was, unsurprisingly, dominated by AI and machine learning, with the company unveiling a host of new products and updated services aimed at helping customers get the most out of the emerging technology.
Amazon Web Services (AWS) likes to say it invested heavily in AI before the current hype took off, but unlike many others it can actually deliver on its promises, with companies like Amazon Q and Bedrock laying the foundation for enterprises around the world . world to realize the potential of AI in their business.
To learn more, I spoke with Swami Sivasubramanian, the company's Vice President of Data and AI – and the man tasked with ensuring AWS's AI promises are delivered in the real world.
Generative AI “fascination”
“We are all fascinated by generative AI, but when we play with these chatbots we have to say: okay, what does this mean for my business, how can I give value to it, to create new customer experiences, or to automate internal processes” , says Sivasubramanian.
He notes that AWS is “unique because we innovate at every layer of the stack,” with the company not only working on the infrastructure needed for AI through the new GPU instances, Graviton4 and Trainium2 chips and more, but also building the software and layers.
Sivasubramanian played a major role in re:Invent 2023: delivering the keynote speech on the second day. Over the course of two hours, he delved into the reasoning behind AWS's AI push, while also revealing more about its plans and new products.
A key theme of his keynote was the 'symbiotic relationship' between people, data and AI, particularly emphasizing the importance of human input in future AI models.
“The biggest thing you'll see are the differentiators as people really move from hype cycle to hype cycle, and that's how you engage people in the loop to make these models work for certain use cases.”
“Every Gen AI app is ultimately not just about sending directions to the models,” he notes, “ultimately, we as humans are the ones creating these ideas!”
Sivasubramanian describes this relationship as one of “intelligence augmentation,” where humans play a crucial role in helping AI models learn and grow, with developers playing an absolutely key role.
“This is an exciting time,” he says, “the abstractions are getting higher and higher, which means that as we moved from assembly language programming to C, and from C to Java, we could actually get more builders who were able to build these applications.”
“Identifying who can build innovative applications will really give more people the opportunity to build… (and) exposing these kinds of interfaces and interactions will really enable new types of developers as well.”
But this expansion and development must also be linked to finding the right use cases for generative AI.
“While the fascination (with generative AI) still exists, some are still trying to connect the dots between their business needs and customer needs,” warns Sivasubramanian.
He highlights that many customers are now often realizing that one AI model will not rule all, and are instead now looking for more customization without the pressure of being locked into a single provider – something AWS appears to be offering with its Bedrock platform.
“We were the first cloud company to come out and say, 'Hey, do you want the right tool for the right job?' That works very well for companies,” he noted.
2023 has undoubtedly been the year that generative AI went mainstream, but many companies are still struggling to figure out how to best use it – and AWS is no different, says Sivasubramanian.
“What's happening with generational AI is very similar to what happened with the start of AWS (…) there's a huge demand for people wanting to build their own models, for a huge number of companies wanting to actually leverage these models, built in and adapted for different applications.”
“It's not a challenge, but an exciting opportunity for us to enable our customers' success with these innovations – and we're off to a great start! We want to ensure that these customers will be successful for years to come, just as they did at AWS.”
“This is really a golden age of machine learning right now, and there is a real symbiosis between these three (data, AI and humans) (…) and they need to come together so that all of these can continue to emerge and flourish. ”