Amazon’s Werner Vogels: Enterprises are more daring than you might think

The longtime chief technology officer talked with Protocol about the AWS customers that first flocked to serverless, how AI and ML are making life easier for developers and his “primitives, not frameworks” stance.

Amazon CTO Werner Vogels

"We knew that if cloud would really be effective, development would change radically."

Photo: Amazon

When AWS unveiled Lambda in 2014, Werner Vogels thought the serverless compute service would be the domain of young, more tech-savvy businesses.

But it was enterprises that flocked to serverless first, Amazon’s longtime chief technology officer told Protocol in an interview last week.

“For them, it was immediately obvious what the benefits were and how you only pay for the five microseconds that this code runs, and any idle is not being charged to you,” Vogels said. “And you don't have to worry about reliability and security and multi-[availability zone] and all these things that then go out of the window. That was really an eye-opener for me — this idea that we sometimes have in our head that sort of the young businesses are more technologically advanced and moving faster. Clearly in the area of serverless, that was not the case.”

AWS Lambda launched into general availability in 2015, and more than a million customers are using it today, according to AWS.

Vogels gave Protocol a rundown on AWS Lambda and serverless computing, which allows customers to build and run applications and services without provisioning or managing servers. He also talked about Amazon CodeWhisperer, AWS’ new machine learning-powered coding tool, launched in preview in June; how artificial intelligence and ML are changing developers’ lives; and his thoughts on AWS providing customers with “primitives” versus higher-level managed services.

This interview has been edited and condensed for clarity.

So what's the “state of the state” on AWS Lambda and how it's helping customers, and are there any new features that we can expect?

You'll see a whole range of different migrations happening. We've had folks from Capital One that migrated old mainframe codes to Lambda. [IRobot, which Amazon announced plans to acquire on Friday], the folks that make Roomba, the automatic [vacuum] cleaner, have their complete back end running as serverless because, for example, that's a service that their customers don't pay for, and as such, they really wanted to minimize their costs yet provide a good service. There's a whole range of different projects happening and whether that is pre-processing images at some telescope deep in Chile, all the way up to monitoring Snowcones running in the International Space Station, where they were in Lambda on that device as well and actually can do processing of imagery and things like that. It's become quite pervasive in that sense.

Now, the one thing is, of course, if you have existing code, and you want to move over to the cloud … moving over to a virtual machine is easy — it's all in the same environment that you had on-premises. If you want to decompose the application that you had, don't want to do too many code changes, probably containers are a better target for that.

But for quite a few of our customers that really want to start from scratch, but sort of really innovate and really think about [what] event-driven architectures look like, serverless becomes quickly the sudden default target for them. Mostly also because it's not only that we see significant reduction in cost for our customers, but also a significant reduction in their carbon footprints, because we're able to do much better packing on energy than customers would be able to do by themselves. We now also run serverless on our Graviton processors, so you'll see easily a 40% reduction in cost in energy usage.

For me, serverless means that our customers don't have to think about security, reliability, managing performance, managing scale, doing failover — all those kinds of things — and really controlling costs.

But always I'm a bit ambivalent about the word “serverless,” mostly because many people associate that with when we launched Lambda. But in essence, the first service that we launched, S3, also is really serverless. For me, serverless means that our customers don't have to think about security, reliability, managing performance, managing scale, doing failover — all those kinds of things — and really controlling costs. And so, in essence, almost all services at AWS are serverless by nature. If you think about DynamoDB [a serverless NoSQL database], or if you think about Neptune [a graph database service] or any of the other services that we have, most of them are serverless because you don't have to think about sort of provisioning them, managing them. That's all done for you.

Can you talk about the value of CodeWhisperer and what you think is the next big thing for or the future of low-code/no-code?

For me, CodeWhisperer is more an assistant to a developer. There's a number of application areas where I think machine learning really shines and it is sort of augmenting professionals by helping them, taking away mundane tasks. And we already did that, of course, in AWS. If you think about development, there's CodeGuru and DevOps Guru, which are both already machine-learning services to help customers with, on one hand, operations, and the other one sort of doing the early security checks during the development process.

CodeWhisperer even takes that a step further, where if you look how our developers develop, there's quite a few mundane tasks where you will go search on the web for a … piece of code — how do we do [single sign-on] login into X, Y or Z? Most people will just cut and paste or do a little translation. If that was in Python and you need to actually write it in TypeScript, we may do a translation on that.

There's a lot of work, actually, that developers do in that particular area. So we thought that we could really help our customers there by … using machine learning to look at the complete base of, on one hand, the AWS code, the Amazon code and all the open-source code that is out there, and then do a qualitative test on that, and then include it into this body of work where we can easily help customers by just writing some plain text, and then saying, “I want a [single sign-on] log-on here,” and then the code automatically appears. And with that, we can do checks for security, we can do checks for bias. There's lots of other things that are now possible because we're basically assisting the developer in being more efficient and actually writing the code that they really want to write.

When we launched Lambda, I said the only code that will be written in the future is business logic. Well, it turns out we're still not completely there, but tools like CodeWhisperer definitely help us to get on that path because you can focus on what's the unique code that you need to write for the application that you have, instead of the same code that everybody else needs to write.

People really like it. It's also something that we continuously improve. This is not a standing-still product. As we look at more code, as we get more feedback, the service improves.

If I think about … software developers, it's one of the few jobs in the world where you can be truly creative and can go to work and create something new every morning. However, there's quite a bit of heavy lifting still around that [that] sort of has nothing to do with your creativity or your ability to solve problems. With CodeWhisperer, we really tried to take the heavy lifting away so that people can focus on the creativity part of the development job, and I think anything we can do there, developers like.

In your tech predictions for 2022, you said this is the year when artificial intelligence and machine learning take on the undifferentiated heavy lifting in the lives of developers. Can you just expand on that, and how AWS is helping that?

When you think about CodeWhisperer and CodeGuru and DevOps Guru … or Copilot from GitHub … this is just the beginning of seeing the application area of machine learning to augment humans. Whether there is a radiologist somewhere that is late at night looking at imagery and gets help from machine learning to compare these images or whether it's a developer, we're really at the cusp of how machine learning will accelerate the way that we can build digital systems.

I was in Germany not that long ago, and there the government told me that they have 80,000 open IT positions. With all the scarceness in the world of labor, anything which we can do to make the life of developers easier so that they're more productive, that it makes it easier for people that do not have a four-year computer science degree to actually get started in the IT world, anything we can do there will benefit all the enterprises in the world.

What's another developer problem that you're trying to solve, or what are developers asking AWS for?

If you're an organization like AWS or Amazon or quite a few other organizations around the world, you make use of the DevOps principle, where basically your developers also have operational tasks. If you do operations, there's information that is coming from 10 or 20 different sides. There's log files, there's metrics, there's dashboards and actually tying that information together and … analyzing the massive amounts of log files that are being produced by systems in real time, surfacing that to the operators, showing that there may be potential problems here and then give context around it because normally these log files are pretty cryptic. So what we do with DevOps Guru, for example, is provide context around it such that the operators can immediately start taking action, looking for what [the] root cause of particular problems are. So we're looking at all of the different aspects of development and operations to see what are the kind of things that we can build to help customers there.

At AWS re:Invent last year, you put up a slide that read “primitives, not frameworks,” and you said AWS gives customers primitives or simple machines, not frameworks. Meanwhile, Google Cloud and Microsoft are offering these sort of larger, chunkier blocks such as managed services where customers don't have to do the heavy lifting, and AWS also seems to be selling more of them as well.

Let me clarify that. It mostly has to do also with sort of the speed of innovation of AWS.

Last year, we launched more than 3,000 features and services. And so why are we still looking at these fine-ingrained building blocks? Let me go back to the beginning of AWS — when we started then, how software companies at that moment were providing infrastructure or platforms was basically that they would give developers everything [but] the kitchen sink on Day One. And they would tell you, "This is how you shall develop software on this platform." Given that these platforms took quite a while to develop, basically what you operate is a platform that is already five years old, that is looking at five years back.

Werner Vogels gives his keynote at AWS re:Invent 2021. Photo: Amazon Web Services, Inc.

We knew that if cloud would really be effective, development would change radically. Development would indeed be able to scale quicker and make use of multiple availability zones and many different types of databases and things like that. So we needed to make sure that we were not building things from the past, but that we were building for how our customers would want to build in 2025. To do that, you don't give them everything and tell them what to do. You give them small building blocks, and that's what I mean by primitives. And all these small building blocks together make a very rich ecosystem for developers to choose from.

Now, quite a few, especially the more tech-savvy companies, are more than happy to put these building blocks together themselves. For example, if you want to build a data lake, we have to use Glue [a serverless data integration service], we have to use S3, maybe some Redshift, Kinesis for ingestion, Athena for ad hoc analytics. I think there's quite a few customers that are building these things by themselves.

But then there's a whole category of customers that just want a data lake. They don't want to think about Glue and S3 and Kinesis, so we give them a service or solution called Lake Formation. That automatically grabs all these things together and gives them this higher-level component.

Now the fact that we are delivering these higher-level solutions, for example, some customers just want a backup solution, and they don't want to think about how to move things into S3 and then do some intelligent tiering [so] that if this data isn't accessed in two weeks, then it is being moved into cold storage. They don't want to think about that. They just want a backup solution. And so for that, we provide them some backup. So we do have these higher-level services. It's more managed-style services for you, but they're all still based on the primitives that sit underneath there. So whether you want to start with Lake Formation and later on maybe start tweaking things under the covers, that's still possible for you. While we are providing these higher-level components, where customers need to have less worry about which components can fit together, we still provide the underlying components to the developers as well.

Is quantum computing something that enterprise CTOs should be keeping their eye on? Do you expect there to be an enterprise use for it, or will it be a domain just for researchers, or is it just too far out to surmise?

There is a back-and-forth there. If I look at some of the newer developments, it's clearly research oriented. The reason for us to provide Braket, which is our quantum compute service, is that customers generally start experimenting with the different types of hardware that are out there. And there's typical usage there. It's life sciences, it's oil and gas. All of these companies are already investigating whether they could see significant speed-ups if they would transform their algorithms into things that could run on a quantum machine.

Now, there's a major difference between, let's say, traditional development and quantum development. The tools, the compilers, the software principles, the books, the documentation for traditional development — that's huge, you need great support.

In quantum, I think what we'll see in the coming four or five years, as I listen to the Amazon researchers working on this, [is that] much of the work will not only go into hardware, but also how to provide better software support around it, such that development for these types of machines becomes easier or even goes at the same level as traditional machines. But one of the things that I think is very, very clear is that we're not going to be able to solve new problems necessarily with quantum computing; we're just going to be able to solve old problems much, much faster. That's why the life sciences companies and health care and … companies that are very interested in the high-performance compute are experimenting with quantum because that … could accelerate their algorithms, maybe by orders of magnitude. But, we still have to see the results of that. So I'm keeping a very close eye on it, because I think there may be very interesting workloads and application areas in the future.


Judge Zia Faruqui is trying to teach you crypto, one ‘SNL’ reference at a time

His decisions on major cryptocurrency cases have quoted "The Big Lebowski," "SNL," and "Dr. Strangelove." That’s because he wants you — yes, you — to read them.

The ways Zia Faruqui (right) has weighed on cases that have come before him can give lawyers clues as to what legal frameworks will pass muster.

Photo: Carolyn Van Houten/The Washington Post via Getty Images

“Cryptocurrency and related software analytics tools are ‘The wave of the future, Dude. One hundred percent electronic.’”

That’s not a quote from "The Big Lebowski" — at least, not directly. It’s a quote from a Washington, D.C., district court memorandum opinion on the role cryptocurrency analytics tools can play in government investigations. The author is Magistrate Judge Zia Faruqui.

Keep ReadingShow less
Veronica Irwin

Veronica Irwin (@vronirwin) is a San Francisco-based reporter at Protocol covering fintech. Previously she was at the San Francisco Examiner, covering tech from a hyper-local angle. Before that, her byline was featured in SF Weekly, The Nation, Techworker, Ms. Magazine and The Frisc.

The financial technology transformation is driving competition, creating consumer choice, and shaping the future of finance. Hear from seven fintech leaders who are reshaping the future of finance, and join the inaugural Financial Technology Association Fintech Summit to learn more.

Keep ReadingShow less
The Financial Technology Association (FTA) represents industry leaders shaping the future of finance. We champion the power of technology-centered financial services and advocate for the modernization of financial regulation to support inclusion and responsible innovation.

AWS CEO: The cloud isn’t just about technology

As AWS preps for its annual re:Invent conference, Adam Selipsky talks product strategy, support for hybrid environments, and the value of the cloud in uncertain economic times.

Photo: Noah Berger/Getty Images for Amazon Web Services

AWS is gearing up for re:Invent, its annual cloud computing conference where announcements this year are expected to focus on its end-to-end data strategy and delivering new industry-specific services.

It will be the second re:Invent with CEO Adam Selipsky as leader of the industry’s largest cloud provider after his return last year to AWS from data visualization company Tableau Software.

Keep ReadingShow less
Donna Goodison

Donna Goodison (@dgoodison) is Protocol's senior reporter focusing on enterprise infrastructure technology, from the 'Big 3' cloud computing providers to data centers. She previously covered the public cloud at CRN after 15 years as a business reporter for the Boston Herald. Based in Massachusetts, she also has worked as a Boston Globe freelancer, business reporter at the Boston Business Journal and real estate reporter at Banker & Tradesman after toiling at weekly newspapers.

Image: Protocol

We launched Protocol in February 2020 to cover the evolving power center of tech. It is with deep sadness that just under three years later, we are winding down the publication.

As of today, we will not publish any more stories. All of our newsletters, apart from our flagship, Source Code, will no longer be sent. Source Code will be published and sent for the next few weeks, but it will also close down in December.

Keep ReadingShow less
Bennett Richardson

Bennett Richardson ( @bennettrich) is the president of Protocol. Prior to joining Protocol in 2019, Bennett was executive director of global strategic partnerships at POLITICO, where he led strategic growth efforts including POLITICO's European expansion in Brussels and POLITICO's creative agency POLITICO Focus during his six years with the company. Prior to POLITICO, Bennett was co-founder and CMO of Hinge, the mobile dating company recently acquired by Match Group. Bennett began his career in digital and social brand marketing working with major brands across tech, energy, and health care at leading marketing and communications agencies including Edelman and GMMB. Bennett is originally from Portland, Maine, and received his bachelor's degree from Colgate University.


Why large enterprises struggle to find suitable platforms for MLops

As companies expand their use of AI beyond running just a few machine learning models, and as larger enterprises go from deploying hundreds of models to thousands and even millions of models, ML practitioners say that they have yet to find what they need from prepackaged MLops systems.

As companies expand their use of AI beyond running just a few machine learning models, ML practitioners say that they have yet to find what they need from prepackaged MLops systems.

Photo: artpartner-images via Getty Images

On any given day, Lily AI runs hundreds of machine learning models using computer vision and natural language processing that are customized for its retail and ecommerce clients to make website product recommendations, forecast demand, and plan merchandising. But this spring when the company was in the market for a machine learning operations platform to manage its expanding model roster, it wasn’t easy to find a suitable off-the-shelf system that could handle such a large number of models in deployment while also meeting other criteria.

Some MLops platforms are not well-suited for maintaining even more than 10 machine learning models when it comes to keeping track of data, navigating their user interfaces, or reporting capabilities, Matthew Nokleby, machine learning manager for Lily AI’s product intelligence team, told Protocol earlier this year. “The duct tape starts to show,” he said.

Keep ReadingShow less
Kate Kaye

Kate Kaye is an award-winning multimedia reporter digging deep and telling print, digital and audio stories. She covers AI and data for Protocol. Her reporting on AI and tech ethics issues has been published in OneZero, Fast Company, MIT Technology Review, CityLab, Ad Age and Digiday and heard on NPR. Kate is the creator of RedTailMedia.org and is the author of "Campaign '08: A Turning Point for Digital Media," a book about how the 2008 presidential campaigns used digital media and data.

Latest Stories