“The AI ​​revolution has already happened, now we have to build it”: Dave Levy, AWS

The world of technology has a new fad every now and then. If a few years ago it was smartphones, social media, or the cloud, it's true that today it's hard to imagine our daily lives without them. But in between, there have also been colored mirrors, from Facebook's "metaverse" that no one used to Apple's $3.5 million virtual reality headset. Where does generative artificial intelligence , which has revolutionized the tech industry in the last two years, fit in?

This is a question many analysts are asking. And while there's consensus that the AI ​​that exploded with OpenAI two years ago has transformed much of office work , there's uncertainty about what changes it will bring to the world of work. No one is entirely clear about what will happen with a technology that grows by leaps and bounds week by week. And that, too, generates a certain amount of panic .

“When the iPhone came out, users didn't quite understand how to touch a screen without buttons. At times, they got lost. That's why there was a physical button in the middle that took you to the home screen: so that with just a press, you felt at home ,” a former Apple executive who worked for 12 years on a project to help the public sector adopt what was then a revolutionary technology, recalls in an interview with Clarín.

Today, this executive does similar work but with the implementation of artificial intelligence, also in the same area: this is Dave Levy , vice president of Amazon Web Services (AWS) for the Public Sector.

AWS is Amazon's cloud computing division , which not only provides storage (like what we use in Google Drive, iCloud, or Microsoft Outlook) but also services for developers. It's a business aimed at developers rather than end users. Levy heads the area that collaborates with governments, universities, hospitals, and organizations linked to the public sector, as well as the aerospace and satellite industries.

For example, in Argentina, the company has worked on the digitalization of the Córdoba government, in addition to being behind the development of Boti , the chatbot for the City of Buenos Aires (among other projects). The cloud (which is, after all , "someone else's computer" ), questioned a decade and a half ago, is now a standard. In 2022, for example, it was a key tool for migrating and safeguarding data from Ukraine at the beginning of the Russian invasion.

At a summit held in Washington, DC, where the company offers talks, updates, and training, Levy spoke with Clarín about the current state of the cloud in relation to artificial intelligence and highlighted a key component that isn't discussed much: data storage.

─The entire industry is talking about AI these days, and this sometimes makes it a little difficult to separate real advances from the smoke. Where do we stand?

─Look, I do believe that AI is going to transform the user experience. We're already seeing it: it's real, it's already happening. Perhaps the big change will happen when I'm retired, although it's happening pretty quickly. I think, if we have to separate the "signal from the noise," the signal here is that it's going to allow us to do things we didn't think possible.

─Well, we have to do performance evaluations every year. For example, I manage 10,000 global employees. Of course, I'm not going to replace that job with an AI, but it can help me design reports, provide me with a template, and help me be more efficient with that very heavy work. I imagine it will free us from the heavy lifting and probably make us more productive, although the truth is we can't give the answer to where all this is going today.

─There are also doubts about AI in various fields, from copyright to job replacement.

─Yes, that's natural. When a new technology comes out, there are doubts. Let me tell you something, I worked at Apple when the iPhone was just coming out, back in 2007. Back then, we weren't used to pressing a screen: we used physical buttons. And the iPhone had one in the middle; it had one for many generations, do you remember?

─Yes, in fact he maintained it for several generations.

─Exactly, but that wasn't a technological or engineering issue; it was so that, when the user felt lost, they had a safe way to return to the beginning , to "home." So, the point of that button was to make the user feel safe. That's what happens when new systems are introduced, and these are the kinds of problems AI developers are facing today: how do we help users feel safe with this technology? That's where the work needs to focus.

─Some analysts point out that the three major revolutions of the last 25 years were the iPhone, the cloud, and now AI. Do you think those are the milestones?

─Well, many compare it to Gutenberg 's printing revolution. That's how big the change happening before our eyes could be. But at the same time, we're at such an early stage that it's very difficult to know where we'll be in 10 years. Today, there are already systems that assist with programming, which is leading many users who have never written a line of source code in their lives to give instructions to an AI and have a program. Perhaps having just a good idea is enough to carry it out ? It could be, but we'll have to wait and see. But it will surely be a major revolution.

Discussions about AI are often associated with the vast computing power required to operate: when a user makes a query to a chatbot, all that processing happens on the servers of the company providing the service, such as OpenAI (ChatGPT), Google (Gemini) or Anthropic (Claude) , to name a few.

However, one key factor is left out of the equation: data storage and the cloud . First, because all of this isn't processed on users' devices, like a phone or laptop, but on the servers of large companies. And second, because in order to operate, these language models rely on immense amounts of databases ( data lakes ), and all of this information is stored somewhere.

─How important is data storage in this new world of generative AI?

─It's critical . First, you have to have all the information in one place. Second, it has to be structured so it's accessible. Storage used to be very difficult to access; you had to have a huge number of disks stacked up, and it was also very expensive. When AWS started, it cost 15 cents a month. At the time, we thought it wasn't realistic, but it was, and over time, it became a more accessible standard. It's probably not as hotly debated today, but storage is absolutely critical and one of the foundational elements.

─With generative AI systems, aren't we creating too much information, much of it worthless?

─At one point, it was said that data is the new oil. I still believe that. We produce information that perhaps isn't useful yet, but it could be useful in the future. I also believe that, as humanity, we've always produced a lot of information. I recently visited Greece, which has a culture that dates back more than two thousand years. It would be great to have more information about their architecture and culture! We have books, there's information, but I'm sure we're missing out on things. I think the human condition is one of producing vast volumes of information.

─In your keynote talk, you said, “The AI ​​revolution is over: it’s time to start building.” Do you reaffirm that?

─Well, I would reconfirm it with some nuances. I would say that AI is here, the revolution has already happened. If we had any doubts a few years ago, I think we've already dispelled them: it's here to stay. What I meant was: let's start putting the blocks in place. Let's think in terms of, precisely, storage: let's put the data where it belongs. Let's try to understand how much computing power we need to perform a task.

─That's another discussion: the large amount of resources needed to process an AI prompt.

─Exactly. In many cases, we work with clients and tell them they might be paying for more processing power when, in reality, it could be cheaper. All of this needs to be considered because computing is very expensive. Why use more or pay more? That's what needs to be adjusted, now that we know the generative AI revolution has begun. My call is this: let's start building.

─What doubts or questions do you have about AI?

─Well, I think, in reality, perhaps the time for asking questions about AI is over. Now it's time to safely do concrete things with it.