Exclusive: Mark Zuckerberg on the next 10 years of Facebook

By nearly any measure, Facebook has had a remarkable year. More than 1.65 billion people use the service every month, making it the world’s largest social network by a considerable margin. Its advertising business has grown significantly faster than analyst expectations, powered by sophisticated targeting capabilities that rivals struggle to match. And in April, CEO Mark Zuckerberg laid out an ambitious 10-year vision that places the company at the frontier of computer science, making aggressive moves in bringing artificial intelligence and virtual reality to the mainstream.

Facebook takes flight

Read our account of Aquila’s inaugural flight

And yet what Zuckerberg talks about most these days, in meetings with world leaders or at his live Town Hall Q&A sessions, is basic internet connectivity. In August 2013, Facebook announced the creation of internet.org, the company’s sometimes controversial initiative to bring online services to underserved areas. Since then, Facebook’s connectivity efforts have expanded greatly. It released open-source blueprints for telecommunications infrastructure in an effort to drive down data costs. It’s testing Terragraph, which augments terrestrial cellular networks with new millimeter-wave technology that delivers data 10 times faster than existing Wi-Fi networks. And it continues to expand its Free Basics program despite setbacks. (In India, regulators banned the program, arguing that because Facebook has the final say over which services can be part of Free Basics, it violates net neutrality principles.)

Facebook believes Aquila could be a powerful tool in bringing internet access to the entire world

Then then there’s Aquila, a wild project to design, build, and launch an aircraft that can beam down internet access to millions of people from 65,000 feet in the air. If Facebook succeeds at its ultimate goal of keeping Aquila drones aloft and delivering data for 90-day periods, the company believes it will have a powerful new tool in bringing internet access to the entire world. (Google, which has undertaken a similar effort called Project Titan, seems to agree. The companies are collaborating to address regulatory issues raised by the drones.)

On the occasion of Aquila’s first successful test flight, Facebook invited me to its headquarters in Menlo Park to meet Zuckerberg and discuss the aircraft, connectivity, and how those efforts will power the next generation of Facebook services. Because while Facebook’s connectivity mission begins by ensuring the entire world has internet access, it doesn’t end there. Zuckerberg believes that you can’t have the best artificial intelligence or virtual reality services until you bring the internet to everyone who doesn’t have it — and dramatically improve the internet for everyone who already does.

Facebook asked that we keep the discussion to the company’s next 10 years — a fair trade, I thought, for a chance at asking Zuckerberg about far-future versions of Facebook. And so we talked about the obstacles Facebook faces in building internet infrastructure, the role the company hopes to play in developing AR, and why VR is likely to define the next generation of computing. We also talked about Zuckerberg’s year-long effort to build an AI that powers his home, and the trouble with using bots to operate your toaster.

But we didn’t cover everything. There are still some basic questions about the manner in which Facebook hopes to provides the internet to the world. Who builds the infrastructure? Who pays for the data? And then there’s the question that dogs all of the company’s connectivity initiatives: to what extent is Facebook providing internet access, and to what extent is it simply providing access to Facebook? After all, connecting 7 billion people to the internet should be hugely beneficial to Facebook’s core advertising business. The fact that Facebook presents its internet efforts as an unalloyed public good, and not (also) a strategic necessity for the company, can fuel suspicions that Facebook has ulterior motives.

Whatever else Facebook does going forward, Zuckerberg said, it starts with a connected world

But in our discussion, Zuckerberg spoke with deep conviction about the benefits that programs like Free Basics were already bringing to the world’s poor. As I was on my way out, he stopped me to ask if he could say a bit more about why he’s spending so much of his time on connectivity projects. “The internet really does bring so many opportunities to people,” he said. “There are all these studies that show that for every 10 people that get on the internet, about one person gets lifted out of poverty, and almost one new job gets created. If you’re talking about 4 billion people who are not on the internet, spreading internet connectivity is clearly one of the biggest things we can do to improve the quality of life for so many people around the world.”

Zuckerberg added that direct investment in education and health is critical. “But I think one thing that a lot of people miss is that if you’re living in a town that doesn’t have a good school, then the internet actually is your best bet for getting a good education,” he said. “And if you’re living in a place that doesn’t have a good doctor, then the internet is your best bet for being able to look up what a condition might be.”

Whatever else Facebook does going forward, he said, it starts with a connected world. “If we make progress on this, it will be one of the great things that our generation can do to improve lives around the world.”

Zuckerberg and I spoke for more than 45 minutes at Facebook’s headquarters, accompanied by a member of his communications team. Below, you can read the interview, edited for length and clarity. You can also use the table of contents to skip to specific subjects we discussed. And for an exclusive account of the launch of Aquila, check out our story “Facebook takes Flight.

Bringing the world online

Casey Newton

You’re making huge investments in bringing the world online. What are the big challenges ahead?

Mark Zuckerberg

So, the mission of the company is to make the whole world more open and connected, and at some point that means helping to connect people to the internet so they can connect to all the people they care about using services on the internet. When we started off in a college dorm room — you don’t think about such things. You’re just trying to build something useful, and figure if a small portion of the people who are already on the internet use it, then that’s really good. But 10 years into the journey of building Facebook, we connected about a billion people — and we realized that our real mission wasn’t to connect one-seventh of the world, it was to connect the whole world. And now we’re this bigger company and have these resources that we can use to solve some of these more fundamental technology problems, like getting the more than half of the world who’s not on the internet onto the internet. So we started looking at the problem and trying to understand, what are all the obstacles that are preventing people from being on the internet?

“Our real mission wasn’t to connect one-seventh of the world, it was to connect the whole world”

For the 4 billion people who are not on the internet today, there are three major obstacles. The first is availability of networks. That is, even if they had a phone, they [can’t] get a signal because there’s no fiber, or there’s no mobile broadband network where they are. That’s about 1.6 billion people. To go after [them], you need to build really affordable new kinds of networks through solar-powered planes, satellites, that kind of work.

[Then], there’s affordability of the network. There’s another billion or so people where technically there may be a signal that they can pick up, but they can’t afford to use it. There are really two solutions to that: one is to make data cheaper, and the other is to use less data. We’re working on both of those. In terms of using less data, Facebook’s apps are some of the most used apps in the world. Just making our own apps more efficient, and making the frameworks that we use open-source — so other big apps can use them [too] — goes a long way. In terms of the cost of data, we’re working on stuff like TIP Telecom Infra Project, making open-source software available for running telco networks — like our Open Compute project, which we’ve done internally. It saved us billions of dollars, and if we can save the telco industry billions of dollars, then in a competitive market that will be passed along to consumers in the form of lower data prices.

And then the third pillar, which is in a lot of ways the biggest, is about a couple billion people. [For them], the issue is not availability, or affordability, but awareness. They can pick up a signal on a phone, and they can actually afford it, but maybe they didn’t grow up with the internet, or they haven’t used a computer. If you ask them, “Do you want to pay for a data plan to use the internet?” the first question you get back is, “What do I get with that?” That’s where programs like Free Basics come in, so people can get access to some free services. What we see is that as soon as people see the value of the internet, within a month, half of people who use Free Basics end up paying for full access to the internet. So, those are the three big issues: availability, affordability, and awareness, and we have big projects on all those.

“Those are the three big issues: availability, affordability, and awareness, and we have big projects on all those”

Today, the big milestone is around availability. If you want to [connect] these 1.6 billion people who don’t have a network, [you can’t use] the same solutions today that are unaffordable for telcos. The approach of stringing fiber and building out these ground base station towers, that’s very expensive. And if you’re getting into a more rural area, or even a less dense suburb on the outskirts of a city, that’s not necessarily going to be economically sustainable for telcos. So we needed to build something that was cheaper. And what we came to was building these solar-powered planes, which are basically cellphone towers in the sky. They’re pretty cheap to build, and they can fly for months at a time, and beam down internet to people.

You know, if you told me when I was getting started with Facebook that one day we would build planes, I would’ve told you that you were crazy. But this is now an important thing for us to do to fulfill this mission of connecting everyone in the world. We’re bringing our “move fast” culture to this problem, and I’m really impressed with our team’s work. We’re not a plane company, but within a little more than two years of having this idea and starting to build the team, we now have a plane which has flown for an hour and a half. [It’s] just a really big milestone on this path to connecting everyone.

From left: Mark Zuckerberg, Facebook founder and CEO; Jay Parikh, global head of infrastructure and engineering; Kathryn Cook, technical program manager for Aquila; Yael Maguire, head of Connectivity La (Image courtesy of Facebook)

Launching Aquila

So June 28th, Aquila had its first test flight out in Yuma. [Read our full account of the launch here.] What can you tell me about the experience of being there? What were you thinking about as all of this was happening?

Well the first thing is, it worked, right? [laughs] I mean, a lot of times these things don’t work the first time. And you know, a lot of the team was really nervous about me coming. I just felt like this is such an important milestone for the company and for connecting the world, that I have to be there. It’s worth it for me to wake up at 2AM and fly down to Arizona, to the middle of the desert, to be there for this even if it didn’t work. But it did, and that was amazing. It was this incredibly emotional moment for everyone on the team who’s poured their lives into this for two years. [We’re] on the path to breaking the world record for the longest UAV flight in history — because it needs to be able to stand there for months in order to be this base station in the sky.

“It’s worth it for me to wake up at 2AM and fly down to Arizona… even if it didn’t work. But it did.”

So, it was a very emotional moment. And then the other things that strike you — it’s big. We were trying to get a picture to capture the scale of it, and we couldn’t. [Zuckerberg points to a somewhat awkward photograph of Facebook executives standing in front of Aquila, which barely fits in the frame.] This is fine, but this doesn’t really give you a sense that this thing has a wingspan bigger than a 737 — yet it’s so light, because it’s carbon fiber. It’s a pretty amazing thing to look at.

Then I’d say [what] was a little bit counterintuitive is it flies really slowly. We’re all conditioned to think that things in the sky move quickly, right? Most times when people are designing planes, they’re designing them to get people or things from a place to a place, so there’s no real advantage to moving slowly. But if your goal is to stay in the air for a long period of time, then you want to use as little energy as possible, which means going as slowly as you physically can while not falling out of the air. I think [Aquila] goes about 25 or 30 miles an hour, with a power output which is around 2,000 watts. To put that in perspective, the Tour de France is happening right now. A good cyclist will output 400 watts. So you’re talking about five cyclists [being] all of the power that this huge plane consumes — not only to fly, but to beam down connectivity at really fast speeds that break records as well.

Give me a sense of how far along the Aquila project is. What’s next for it?

This was the first test, and it was really successful. We intended to fly it for 30 minutes, and it was doing so well that we actually kept it up for 90 minutes. One thing that really excited the team and me is that the design ended up being even more efficient than we’d guessed, and it used even less power to stay up than we thought. We did pick up a lot of telemetry and data, and there were some things that we want to tweak. We’ll do another test flight, and at some point soon we’ll test the solar-powered plane with the communications payload in it. That way we can test it actually beaming down connectivity. We’ve tested that piece separately — that’s a lot easier to test without the plane. That whole side of things is really awesome, too. We have this laser communications system which can beam data at tens of gigabits per second, which is 10 times faster than anything that’s existed before.

At some point we’re going to put these pieces together and there’ll be a couple more test flights. But the milestone that we’re really going after is, can this break the record for the longest endurance UAV? Once you have that, then I think we’re ready to build thousands of them, and work with telcos and governments all around the world to connect people on the outskirts of cities, and rural areas, and in disaster zones, and wherever you can’t get traditional connectivity to today.

What are the implications of drones like Aquila in emergency situations?

A few years ago there was the Ebola epidemic in western Africa, and one of the big things that people needed was mobile connectivity and data. Facebook paid to get a bunch of connectivity from satellites there. But in the future I think what you’d really want to do is send a plane over there, and then tens of gigabits should be enough for the hospitals and [emergency] centers. That that’s going to be an important use case in the future.

What do you hope the Aquila project will have achieved in a few years?

“We’re not an aerospace company. But I guess we’re becoming one.”

It’s going to be cheap to build these, and I think the future is going to be thousands of solar-powered planes on the outskirts of places where people live. That’s going to make connectivity both much more available and cheaper. And, I think, it can help play an important role in closing this gap [and] getting more than a billion people online. This is an early milestone, but it’s a big one. It’s not something you necessarily expect Facebook to do, because we’re not an aerospace company. But I guess we’re becoming one.

And how will you get so many planes in the air?

We’re not planning on building a network. What we want to do is prove that this works, and then figure out ways to license the technology or give it away so that telcos and governments and nonprofits and crisis areas can offer this. That’s the vision. I think it’s really exciting.

Terragraph, and improving connectivity

This year Facebook also introduced Terragraph, which is designed to supplement local cellular networks with millimeter-wave technology. What effects do you hope Terragraph has when it goes live in San Jose, CA, later this year? And assuming people love it, how will you think about scaling it?

I think 5G and Terragraph in general are going to be very important for both increasing bandwidth and decreasing latency. That’s important for some of the other things I think we’re going to want in the future, [like] live video, where you can broadcast to people and then interact with them live no matter where they are in the world in just a second. Virtual reality is an important one that we think about — virtual reality is all about presence and feeling like you’re there with the person. In order [for that] to work, the network needs to be able to download a huge amount of information instantaneously, and there can’t be latency.

“In order for VR to work, the network needs to be able to download a huge amount of information instantaneously.”

I think when most people think about virtual reality, [they] think about games or 360-degree video experiences. But I think about things like a doctor doing surgery remotely, [which] is already happening today in some cases. You want the best doctor in the world to be able to do a surgery no matter where it is. Increasingly, we’re using micro-robotics and different [devices] that doctors are controlling, but [you also] need high bandwidth and low latency. I think that’s a really exciting future use case that can improve health outcomes and safety, that is based on connectivity and virtual reality. That’s the kind of thing I get excited about.

And then there are more day-to-day use cases like, you go to the Warriors game or a concert or sports event, and the connectivity is overwhelmed because you have that one base station for that area. There’s a spike in density and there isn’t a match between how much bandwidth is available and how many people are there trying to pull it. In the future, when you have things like Terragraph that can produce and have much higher throughput of bandwidth, I think the days of going to a sports stadium or a concert and not being able to post your photo or video online are done. That’s a more day-to-day thing that people can look forward to.

The lessons of Free Basics

You also mentioned Free Basics, and it strikes me that everything we talked about today is going to require some kind of partnership with the government. Given some of the debate that Free Basics has raised around net neutrality, I want to know what you’ve learned from that debate, and how it’s informed how you’re going to approach your other connectivity projects.

You’re right that all of these things will involve partnerships with the industry, and telcos, and governments — and nonprofits, in a lot of cases. With Free Basics, we worked with all of them. Overall Free Basics has been very successful, right? It’s live in 42 countries, and we’ve connected more than 25 million people [through internet.org]. That’s a good start. You hear a lot about companies offering some Wi-Fi hotspots or something, but we’re talking about tens of millions of people getting online. That’s a big deal, and it’s growing quickly.

“You hear a lot about companies offering some Wi-Fi hotspots … we’re talking about tens of millions of people.”

We have had setbacks in a couple of countries, India of course being the biggest. With a billion people in India not online, it’s the most important [country] to get right. We’ve learned a lot about how we need to interact with governments and the political system and regulators, and build support in order to have these things work. And I think we’ll take those lessons forward on the future work we’re doing in Free Basics, which by the way is continuing to roll out around the world. One day, once we’ve shown that it’s a successful program around the world, I hope that we’ll get another chance to come back to India and offer it there, too.

But [partnerships are] also going to be important for all these other things [too]. Self-flying, solar-powered planes [have] a lot of problems that are similar to self-driving cars, and I think we’ll have to work out what the framework is for that.

AI, and Zuckerberg’s robot butler

Let me ask a little bit about AI. At the start of this year you wrote about wanting to build an AI to manage your home. How’s that project going, and how is it informing the way you think about AI being useful in Facebook’s own products?

It’s going well, and I’m planning on doing a demo soon. I have it to the point where I can control everything in the house. It can control the lights, the temperature, the doors. It can make me toast.

How does it make you toast?

Well, the real question is not how it makes me toast, that’s actually easy — you hook up an internet-powered [device] to the power and you trigger it. The real question, which is actually the more challenging AI problem, is when to make me toast. In the morning I’m writing all these emails, and I go for a run, and my meetings start at different times, and so I’ve built this whole thing which figures out from where I am — Am I out? Am I on a run? — when’s the right time to [make me toast] and then tell me. That still needs some work, but all the stuff is working.

One of the things that’s been pretty fun is using all the AI tools that we’ve developed at Facebook so far, just seeing how much progress we’ve made on image recognition and face recognition and object recognition and voice recognition and language understanding. The bot framework that the Messenger team is building has been really cool. If I hadn’t done this home AI project, I probably wouldn’t have had a reason to get into the code on those, which now I have. Even if I’m asking sometimes silly questions, [I get to] know more of the people, and I use more of the code. That, I think, is helpful. At home, I think the biggest impact has been that I’ve wrested control of the thermostat from Priscilla. [Everyone laughs]

“This isn’t magic, right? It’s math and statistics and pattern recognition”

At F8, you talked about AI as a way to build systems that are better than humans at perception — seeing, hearing, and understanding language. Facebook can already describe the contents of photos and translate my posts into many languages at once. As you think a decade ahead, where else do you think AI can improve in terms of perception? And which do you find most exciting?

I think AI breaks out into two big categories. There’s pattern recognition, and then there’s this broader, amorphous category of general intelligence and unsupervised learning which, frankly, we don’t understand that much about. Most of the things that we’ve made progress with today are complex pattern recognition systems. There’s a lot of debate in society about, “Is AI going to do good or harm?” But I think pattern recognition will be extremely helpful in so many ways, from being able to tell a blind person what is in an image, to being able to read something to someone who is deaf or just can’t turn the volume on on their phone right now. Or, being able to diagnose diseases better, or better identifying drugs that might treat a disease, or drive cars more safely. This stuff is really going to save people’s lives.

It’s important to keep in mind that this isn’t magic, right? It’s math and statistics and pattern recognition and using a lot of data in different data sets. It’s not going to get away from us — it’s all stuff that will just improve people’s lives. The thing that’s probably most exciting to me is that when one team makes progress on making these pattern-matching systems better, and we figure out how to make deep neural nets converge at 37 layers — something that people hadn’t done before — then we publish that paper and other companies and teams can take that advance and now their systems are diagnosing diseases better, or driving cars better. That’s part of what’s so exciting. It’s really important to have this perspective that we’re still in the phase of pattern matching. At a very basic level, this is really good, and not something that we should be afraid of pushing further.

Bots of the future

Increasingly, AI is moving into action — bots like the ones you introduced at F8 are performing real-world tasks. But some people, myself included, have complained that conversational bots can feel slower than other interfaces. How do you see conversational interfaces evolving over the next several years?

Well, let me tell you where we came from — which is [every month], many millions of people will message a business page on Facebook to ask questions about that business. Increasingly, businesses are using their Facebook page as their main presence, and if you want to reach a business, messaging them on Facebook is a great way to do that. But you can imagine that having that business respond might take hours, or in some cases, a day. So one of the first applications and things that we started looking at was researching, can we build an AI system that looks at the answers that a business is giving to people and start to predict for a certain category of questions what the answers will be? We realized that we actually could in a lot of cases. And as a business responds faster, people message it more. So one thing that we’re starting to think about is if we can move businesses away from manual responses to automated ones, then that actually is dramatically faster than what people have had before.

“I don’t think conversation is the best way to interact with everything.”

Is some tailored UI going to be better than a conversational interface? Maybe in some cases. I don’t think conversation is the best way to interact with everything, but I think this could literally be 10 or a 100 times better than the way people interface with businesses today. And qualitatively, too, whether it’s faster or not, I don’t know anyone who likes calling businesses for customer support or making reservations, because it’s uncomfortable, and you have to give your full attention, and it’s synchronous. Just being able to send a message, whether it’s you talking to it, or sending a voice clip, or typing something quick, and getting the answer back as soon as possible, whether it’s a couple of seconds later or a couple of minutes later — I think that’s a big improvement in people’s lives. That’s what I see and that’s what I’m excited about.

I’m somewhat ashamed to admit to you that I primarily use Facebook M in cases where I would otherwise have to pick up a phone.

I think that’s probably not very different from other people.

Facebook’s bet on virtual reality

This spring, Facebook showed off demonstrations of “social presence” — bringing people together from across the world in a virtual space. It feels very sci-fi to me, but you’ve got a working demo. Can you talk about the implications of this kind of technology, especially once you reach super high resolutions and low latency?

VR is important for a couple of big reasons. The first is that our mission is to give everyone the power to share their experience as best as possible, and to help everyone understand what’s going around the world. Technology allows that to happen at increasing levels of richness. If you go back 10 or 15 years, most of what people shared online and most of what you read was probably text. And now almost everything has photos, and photos are becoming a very primary way that people communicate. And as the networks get better, we’re getting videos, too. Now you can post a video without it taking minutes to upload, you can watch a video without it taking 30 seconds to buffer, [and] it’s starting to be a good experience. So videos are starting to be one of, if not the main, ways that people interact and consume content online.

“Our [lived] experience isn’t video. It’s not a small 2D frame, it’s this 3D world.”

But that’s not the end of the line — there’s always more, and our [lived] experience isn’t video. It’s not a small 2D frame, it’s this 3D world. You want to be able to capture that, and you want to be able to map it out digitally, whether that’s in a game where you’re creating the landscape, or in a video where you’re capturing it. I think this is just the next logical extension in how we express and experience the world. I think you’ll do that in VR with a headset, and you’ll do that in interfaces like Facebook and Messenger and WhatsApp, in the way that we have 360-degree videos in News Feed today. It’s not necessarily the full experience of having a headset, but it’s still really good and gives you the sense of what being there is.

The other big trend for VR is that every 10 or 15 years, there’s a new major computer paradigm. These are usually defined by their interface, and how they interact with the world. If you go back to the ’80s, we had these big servers that you almost needed a degree to be able to use. In the ’90s, we got desktop computers which were pretty clunky, but they were efficient enough that they made work better, so people used them primarily as a productivity tool. But they were kind of using a mouse over here to change something over there. Then we got the web, which made it so that data could flow to multiple different computers more easily. We got phones, and phones are really the first mainstream computing platform. People really like it, but that’s not the end of the line, either. There’s going to be something more.

Just like now with phones, we have them with us all the time [and] they’re more natural [than computers] in so many ways, VR and AR are going to be even more natural. All of these limitations with the phone, you have to take it out of your pocket, it’s kind of awkward — you won’t have that. You’ll just have glasses, or eventually contact lenses. [You’ll be able to] look around and see different things and interact with [by] using your hands and reaching out and grabbing stuff. That, I think, is the next logical step for how we’ll interact with this. We’re really excited about pushing both of those trends forward.

The challenges of augmented reality

We’ve seen a lot of your plans for VR, but not much about AR. What role do you want Facebook to play in advancing AR, and might it involve your new hardware division?

We’re working on AR research. The reason we started working on VR [first] is the technology is ready. You can build a VR headset that is affordable. And people are really using them — people use them week after week, and spend a lot of time with them and have all these different experiences. You can make that because it works with phones. Even Rift, which is a bit more expensive, can use a lot of [phone] components, and that’s why it’s hundreds of dollars and not thousands or tens of thousands of dollars.

[With] AR, there’s still more science questions that need to be worked out, and I’m optimistic that we’ll have the answers to that pretty soon. Within the next five or 10 years, we’ll have versions of that that are maybe where Rift is today, for AR. But, there’s no doubt in my mind that VR is easier to build and bring to a lot of people, and that’s why we’re are starting here on the product side. But we’re also researching AR very seriously, too.


Originally published at www.theverge.com on July 21, 2016.