With generative AI already changing industries and businesses around the world, the race is on for the next big commitment to this technology.
Having played a leading role in helping to build and establish the systems behind many enterprise use cases for generative AI, Amazon Web Services was eager to boost its credentials at the recent AWS re:Invent 2024 event, with a series of updates and new releases alike.
More powerful, with more options
“What Amazon does really well is taking technologies like generative AI and applying them at scale to real-world business problems,” Vasi Philomin, vice president of Generative AI at AWS, told us in a group briefing at re:Invent.
“We don't allow ourselves to just take technology and do technology for technology's sake; we want to apply real problems and do it at scale.”
Philomin points out how Code Whisperer, which he helped create and is now known as Amazon Q Developer, was created at a time of great buzz around generative AI and has now helped fuel a major push within AWS to create services. Truly useful and intuitive.
“We could have created another chatbot and launched it,” he recalls of the launch, “that would have been the easiest thing to do, but we did something else, which was take a step back and think carefully about what companies do.” Do you need? “It’s not just a chatbot, it’s much more powerful.”
Philomin describes how AWS AI customers are typically builders or developers, so they want to build products and crave new services and tools that allow them to create something entirely new.
“We want to bring it all together so our customers have choice,” he says, adding that choice is another crucial factor, as AWS seeks to provide access to more models than any other provider.
“Having one model doesn't help you build apps; you need workflows that work across models that help you build apps,” he adds, highlighting the role of guardrails in establishing company-wide policies on which apps are built. with the AWS Bedrock platform. You can do or not do, for example talk about competitors or mention politics.
Forward-thinking agents
AI-powered agents will play a big role in the next generation of enterprise technology, and Philomin describes AWS's work on multi-agent collaboration as “probably the most forward-thinking release” in recent memory.
“I think generative AI in general will go in that direction, where the abstraction that companies will choose to expose their business logic, which is driven by generative AI, they will expose it as agents.”
“To me, an agent is a digital worker who is given a lot of resources, and that worker can go and solve a problem that you give them,” he notes, “an agent automates work, not just chats with you , but it really does.” something behind that.”
“It's a reflection of the real world, where humans work together in a certain way to accomplish things,” he adds, noting that it's similar to hiring a new employee and then getting them up to speed on the culture, principles, leadership and culture. of your company. teaching them the tools you use to get the job done – just a digital version.
Philomin also outlines a broader vision, in which a general supervisor, or “super agent,” will sit above specialized sub-agents, since “the idea is that the supervisor will coordinate them… and solve a much larger problem.” . , not simply using them independently, but having them collaborate with each other.”
“You can imagine now, in the future, where this is going: you can set up different types of organizational structures, with agents, and then see how they perform as a group.”
So if agents assume even these organizational positions, where does that leave the human worker? Philomin notes that “the question that comes back again and again is: is generative AI replacing people? “I don't believe it at all.”
“In my opinion, what is going to happen is that you are going to leave the mundane, boring, undifferentiated work to AI, and what you are going to focus more on is doing the work that you really want to do; it's true in all industries”. .”
This includes making software developers more productive by leaving menial code (“the boring stuff,” as Philomin calls it) to an AI, as well as highlighting how Kiva robots in Amazon fulfillment centers have taken on “menial jobs.” and undifferentiated”, such as lifting heavy objects. shelves, but without replacing human workers, who moved on to other tasks, with the end result of an increase in overall human recruitment.
“It's just a great starting point,” says Philomin, “humans aren't going to compete against AI, they're going to compete against other humans using AI… you're going to be at a disadvantage if you don't want to use a tool, it's the same as “If you said I don't want to use a smartphone: you're going to be less productive than someone else.”
As generative AI infiltrates our everyday tasks, what are the concerns? As with any new technology, an initial cycle of hype will likely be followed by widespread adoption, which in turn requires regulation and scrutiny.
“We have to find the balance between not stifling innovation and at the same time facing challenges,” says Philomin.
“Responsible AI is a challenge we have been working a lot on: AI regulation must be done well, it is absolutely necessary, but if each country makes its own regulation and there is nothing in common, it will stifle innovation.”
“We are working with the regulators to make sure that they at least talk to each other and that there is a common agreement.”
“It's not that the challenges are difficult to overcome, it's just that we need time to overcome them… it's a fairly new technology, and it's also evolving; there's a lot of research behind the scenes.”
AWS has therefore firmly established its place at the forefront of the enterprise AI era, and Philomin concludes by noting that the company is well prepared to maintain this place in the future.
“Everything has a reason why we launch it,” he notes, “Amazon thinks about the whole thing: we think end-to-end, that's why it takes us time… and that's what sets us apart from everyone the rest, who is it?” thinking about one thing and not the whole.”
“When a technology of this magnitude happens once in a lifetime, I think it could be as big, if not bigger, than the Internet… I think the potential is there, now we have to execute on it.”