A sweeping guide to the technologies and talents driving disruption in our office cultures, hiring practices, and economic opportunities
1. Offices without walls
VR could kill the company headquarters
VR could kill the company HQ. As VR headsets become faster and lighter, companies will replace real-world conference rooms with virtual workspaces that are more inspiring and infinitely flexible.
In May, Andrew Bosworth, Facebook’s head of virtual and augmented reality, tweeted a 9-second video clip that reveals the future of remote work. Shot from the perspective of a person sitting at a desk, the video shows two monitors floating at eye level. Everything behind them—a bookshelf, closet, guitar propped against a wall—appear digitally muted, as if to minimize distraction.
The video’s point of view turns to the right to find a third monitor floating next to the desk. A hand reaches out, pinches it between two fingers, and drags it into the workspace. The existing eye-level screens adjust outward as the new monitor settles between them.
“This is real footage using prototype headsets,” wrote Bosworth, noting that remote work’s future will rely on “mixed reality concepts” that “allow people to switch between real and virtual worlds.”
It’s all but inevitable that some version of Bosworth’s clip will come to define the next chapter of knowledge work. Existing technology already makes fully remote offices practical, but prior to 2020, few people took advantage. US Census data shows that in 2019, only about 6% of people worked from home full-time. Covid-19 has boosted that number to 42% with video conference calls, shared documents, and collaborative workflow software largely keeping productivity stable. In the short term, many will return back to offices once it’s safe to do so. But millions of newly minted work-from-home professionals are eager for a tech breakthrough that will make their kitchen tables feel and function more like the real-world offices they left behind.
No longer sci-fi
There are already simplified versions available of our future VR-enhanced WFH setups:
FLOATING SCREENSThe company VSpatial fills your workspace with floating screens and allows you to collaborate with colleagues who appear before you as floating heads.
VIRTUAL CONFERENCE ROOMSGlue rents sleek conference rooms where you and your coworkers can gather in VR.
REAL-WORLD INTEGRATIONSMeetinVR lets you merge real and virtual elements during virtual collaboration.
The promise of offices set in VR and AR worlds goes beyond isolated productivity. Yes, you’ll be able to create a 360-degree wall of monitors that wrap you in a cocoon of pdfs, spreadsheets, and chat boxes. More significantly, you’ll be able to sit down with colleagues and pass these digital files back and forth. Your workspace will grow and shrink to fit your team. If you’re giving a presentation, you’ll load a meeting room with a stage. Whiteboards and video players will appear and vanish as needed, and your digital assistant will circulate meeting notes afterward.
By the end of the decade, you’ll wear a headset no heavier than a pair of thick-frame glasses, and through it, you’ll work side-by-side with coworkers who—in the real world—are sitting miles away in their own home offices. You’ll use your hands to pluck software tools from the cloud and open them up on the table between you, and you’ll pass documents back and forth in a way that feels as natural as real life.
The avatars these companies use now skew cartoony, but in time, they’ll become more lifelike. Their mouths and arms will sync perfectly to the words and gestures of their owners, and the experience will become as convincing as reality.
The Oculus Quest points to VR’s exponential growth curve
Facebook’s latest headset, the Quest 2, is a leap forward in every way. In just two years, they created a lighter, cheaper headset with sharper resolution, and more powerful processing. The next question is whether there will be enough users to justify a company’s decision to go virtual.
Again, the Quest 2 provides a clue. On its Q3 2020 earnings call, Facebook announced that pre-orders for the new console were five times higher than they’d been for the original Quest. And on that same call, Mark Zuckerberg announced the company was working toward a milestone of 10 million active VR units, a point at which he believes developers will prioritize VR over other platforms and set off an explosion of new software.
The digital migration is already occurring with video conferencing and cloud-based workflow management programs, but as virtual worlds become more powerful, today’s tools will come to seem archaic. The efficiencies of remote work will pile up, and offices as we know them will cease to exist.
2. Self-taught bots
Your robot coworkers may learn to code themselves
Robots may learn to code themselves. AI allows machines to learn organically and share their knowledge with other machines, busting open their potential both in and out of factories.
Remember the first time you shot a basketball? If you were a kid playing on a standard 10-foot hoop, you probably lobbed an air ball. So you tried again, and your second shot hit the left side of the backboard. You kept adjusting your aim and power until finally the ball dropped through the net.
This process of trial-and-error is one of the primary strategies humans use to learn new skills, and as it turns out, robots can use a similar technique called “reinforcement learning.”
In a 2020 proof-of-concept robotics experiment, researchers from Google Brain, part of the Alphabet X R&D division, wired a robot for deep reinforcement learning and gave it a basic understanding of how to stand up after falling down. Researchers then challenged the robot to walk on its own, and after a few hours of twitching and stumbling, it became ambulant. The researchers published their findings in a paper titled, “Learning to Walk in the Real World with Minimal Human Effort.”
This kind of sophisticated machine learning will allow us to hand complicated tasks over to robots. And, despite our fear of job displacement, we’ll find new, better ways to work. While the legend of John Henry tells us we’ll battle machines to the death, history says otherwise: Breakthrough technologies create more jobs than they destroy, according to a report from the Mckinsey Global Institute. The internet, for instance, created 2.6 jobs for each one it eliminated, and as MIT researchers noted in a 2020 report, roughly two-thirds of today’s jobs didn’t even exist in 1940.
Practice makes perfect, even for robots
Illustration by Anthony Calvert
Go ahead, take a shot!
Drag to shoot
Adjust the dial to tweak the robot's shot and try again.
So where does AI-powered robotics technology lead us? Imagine two hopeful scenarios.
In the first, you’re the proud owner of a new home-assistant robot. Out of the box, your bot seems brainless: It can’t navigate the hallways of your house, operate your fancy coffee, or grip the scooper to clean your cat’s litter box.
With a traditional robot, those tasks would need to be coded into the software, which explains why the world’s current fleet of motorized machines are primarily engaged in predictable factory work. But with deep reinforcement learning, your new bot is as curious as a child and ready to pick up new skills.
Creating home-bots like this is the goal of Google’s Everyday Robots Project, which also falls under Alphabet X. “Our moonshot is to see if we can make robots as helpful to people in the physical world as computers now are in the virtual world,” wrote project leader Hans Peter Brondmo in a recent blog post. But to get there, he says, the technology has to move beyond coding: “We have concluded that you have to teach machines to perform helpful tasks; you cannot program them.”
Here’s reinforcement-learning scenario two: Imagine a manufacturing plant needs to thread 10 million bolts from one box into 10 million nuts in another. It’s a perfect job for a robot. But rather than ordering a new machine or waiting for a long and laborious coding project, the foreman simply shows one factory bot what a successful bolt-nut combo looks like. Then they leave the machine a box of loose bolts and nuts to practice with overnight.
By the time work begins the next day, the robot will know exactly how to thread a bolt into a nut. Better yet, the bot will be able to export the new bolt-nut code to other robots; with a quick, wireless file transfer, an entire factory of bots will learn to perform the task. If the company happens to have factories in other parts of the world, they can instantly activate a room full of bolt-nut bots, too.
With more complicated tasks—like wiring a complex circuit board, for instance—multiple machines can work on the problem at once and share information as they go. This is called “distributed learning” (it’s also referred to as “cloud robotics”), and Japan’s Fanuk, an industrial robot company, is developing hive-mind machines to do this right now.
Roll all the reinforcement-learning applications together and you start to see the full potential. When somebody teaches a home-assistant robot to weed a garden or mix a whiskey sour (two jobs that single-purpose bots can already perform), you’ll be able to upload the code to your own bot. And if a factory wants its fleet of robots to lace tennis shoes in the morning and string rackets in the afternoon, that won’t be a problem.
As machines grow more adaptable, entrepreneurs will have an easier time launching new products, and we’ll automate time-consuming tasks like taking out garbage and painting bedroom walls. Robots will take on the world’s rote labor, and in doing so, they’ll free humans up for the more ambitious and creative projects we do best.
3. Hiring without borders
Offices could recruit international talent without language barriers
Offices could recruit international talent without language barriers. Emerging translation technology promises to work so proficiently that you might not even know when someone’s speaking another language.
The world’s first translation machine debuted in 1954, and the headline from the Washington Times Herald explained it like so: “Robot Brain Translates Russian into King’s English.”
The robot brain in question could translate 250 words, which is about half a percent of the average 20 year-old’s vocabulary. It wasn’t exactly a market-ready tool, but as a proof of concept, it kickstarted a race to build on technology that might one day unite cultures and bridge language barriers.
Nearly 70 years later, we’re almost there. Translatron, a new language-decoding system from Google, promises to translate speech in real-time, complete with context that closely matches the speaker's intent. Better still, Translatron will retain the speaker’s voice and vocal inflection. Combined with other emerging AI technologies in video, you’ll soon be able to sit at a virtual conference table talking with people who speak half a dozen languages, yet all you’ll hear is your own native tongue.
Translation technology has evolved considerably in the time since the original robot brain. A decade ago, Google Translate rolled out an app you could carry on your smartphone, and in 2016, it switched over to an AI-based neural network that allowed it to translate full sentences. This upgrade allowed the app to quickly rearrange chunks of words into grammatical, natural-sounding translations.
That was a breakthrough moment. Prior to Google’s neural network upgrade, translation technologies had basically worked on a word-by-word basis, like foreign students doing language homework. But the new technology performed closer to a human translator, and key to the improvement was the AI’s ability to learn over time. The more people use Google Translate, the more accurate it becomes. In a November 2016 announcement post, Google Translate’s director of product wrote, “With this update, Google Translate is improving more in a single leap than we’ve seen in the last 10 years combined.”
As it exists today, Google Translate understands spoken word, and using your phone’s camera, it can instantly decode foreign signs and handwriting. It’s lightyears ahead of the 1954 robot brain. And yet—it hasn’t yet eliminated the language barrier.
The shortcomings of AI-based translators are substantial enough that companies aren’t generally willing to use them to form business relationships. They still make mistakes, and as with human translators, they introduce a time lag that makes it difficult for the listener to fully grasp the vocal context of each sentence as it’s spoken. The current tools are sufficient for establishing superficial diplomatic relationships or traveling in a foreign country, but they can’t yet facilitate the nuanced, complicated conversations that drive daily business.
Translatron will change that. The new technology will fundamentally reprogram the architecture of machine translation.
Speech-to-speech translation as we now know it involves three computational steps. With Translatron, this process is reduced to one. Rather than translate text, the technology rearranges spectrograms, the visual representations of sound, so that they convert instantly from one language to another.
This single-step process has a few advantages. One, it will make translations more accurate by avoiding the problem of compound errors that occur through multiple steps. Two, it will virtually eliminate the time lag between source speech and translation. And three, it will retain the speaker’s voice and vocal inflection. Translations will be as smooth as if the person were actually speaking another language.
Instant translations from farm to table
Illustration by Anthony Calvert
Toggle to listen in:
The real magic happens when you combine Translatron with other emerging video technologies that essentially “translate” moving images. For example: NVIDIA, a Silicon Valley technology company, is currently working on AI that could adjust video in real time, so everybody will appear to be making eye contact with everybody else on a live call. The logical next step would be to sync the mouths of foreign speakers so that the lips move in time to the output languages, rather than the speaker’s natural tongue.
You’ll be interviewing a native Mandarin speaker virtually without ever realizing they doesn’t speak English. Similarly, if you’re pitching a panel of potential customers in France, you’ll be able to clearly explain your service without losing any critical context. New companies will form with employees who live globally. They’ll speak Portugese, Swahili, Japanese, and English, yet everybody will experience daily meetings in their own native tongue.
Over the years, machine-translation technology has made great strides in lowering the language barrier. Now, we can finally imagine a world where the barrier simply doesn't exist.
4. AI with ethics
Companies could code artificial intelligence with human morality
Companies could code artificial intelligence with human morality. As AI becomes increasingly powerful, machine ethicists will step in to keep algorithms from causing harm.
In 1997, IBM’s Deep Blue, a computer capable of analyzing 200 million board-game moves per second, beat the world chess champion in a six-game tournament. In 2011, IBM’s Watson beat a panel of trivia experts in Jeopardy, and in 2016, A Google DeepMind computer defeated the world’s best living player of Go, a complicated game with more board configurations than there are known atoms in the universe.
Those were childhood years of artificial intelligence, where the world’s smartest computers just wanted to play games. But now AI is entering adolescence, and as it slouches toward omnipresence, the engineers who build the hardwired brains are starting to grapple with thorny issues around how to integrate thinking machines into society without causing harm to the humans they serve.
Today’s AI powers banking, cars, and navigation apps, and soon it will exist in everything with a battery or power cord. The companies building these technologies are asking the hard questions: Is AI making decisions that align with our human values? Are our algorithms and autonomous machines concerned with issues like safety and income equality? Can machines be ethical?
Paving the way In 2016, the Partnership on AI formed to roadmap the future of ethical AI. The organization consists of:
19 academic institutions
including Harvard’s Berkman Klein Center and the ACLU
61 research non-profits
including Human Rights Watch and Association for Computing Machinery
20 leading AI companies
including Apple, Intel, Microsoft, Google, Amazon, and Facebook
The organization points to a future where ethicists work side-by-side with coders, helping to ensure that when machines think, learn, and create, they do so with broad social values in mind. The Associated Press already uses AI to write standard sports and corporate earnings stories, but as AI journalism becomes more complex, PAI’s effort will become increasingly important. Which sources should AI trust for reporting? Similarly, ethicists will help guide AI-based disaster plans for autonomous vehicles (where should a crashing car aim if the brakes fail?) and govern the behavior of personal-assistant robots in the home (how do you ensure the machine doesn’t rip up floor tiles to clean the cracks in between?).
While sci-fi movies present rogue AI as killer robots, the likelier scenario is machines will do harm accidentally as a byproduct of trying to complete the task we’ve given them. And researchers already know that poorly defined tasks can quickly run afoul of sober morality. A few years back, Microsoft released Tay, a chatbot experiment that was supposed to learn communication skills by listening to social media chatter. Almost immediately after Tay was went live, it began writing racist, sexist, and homophobic tweets. The lesson was clear: Without the proper guardrails, AI risks appealing to our worst instincts.
Establishing these guardrails isn’t easy, but that’s where ethicists come in. In 2020, researchers from Pennsylvania's Harrisburg University wrote a paper describing an algorithm that could predict whether someone was a criminal with 80% accuracy simply by analyzing a picture of their face. A useful tool for police? Maybe. But it was also a threat to civil liberty. Experts in ethical AI—including those from Google, Facebook, Apple, and Microsoft—responded to the paper with a harsh rebuke: In the hands of the criminal justice system, they argued, the tool would create “dangerous feedback loops” that deepen racial profiling. The paper outlining the algorithm was ultimately pulled from publication.
PAI sets broad industry goals, such as building algorithms that can detect fake video. It offers training tools that allow coders to practice building AI with achievement objectives like “avoiding negative side effects.” And to help coders learn from past mistakes, it maintains an AI incident database: Each time a trading algorithm fuels a market crash or an autonomous car kills a pedestrian, the details go into the database for further study.
The field of ethics in AI is still young, but as it develops, companies will begin deploying in-house oversight teams to scrutinize datasets and work alongside coders. Artificial intelligence will continue maturing into adulthood, and as it does, we’ll be there to ensure it grows up responsibly.
5. Space race redefined
The world population could virtually double
The world population could virtually double. Billions of people still don’t have internet, but thanks to the new tech-world space race, that’s about to change.
Take a quick inventory of your internet usage. Alongside your work activity, factor in services like banking, shopping, investing, education, and video streaming. Now bundle everything together, put a price tag on it, and multiply that by 4.7 billion—the world population of internet users.
That unfathomably large number is the value of the web, and it’s expected to grow multiple times over as tech companies bring billions more people online by launching thousands of satellites into space.
Nearly 40% of the world is currently without internet, but a new tech-company space race will change that by putting a constellation of WiFi routers in earth’s orbit. They’ll circle the globe and beam internet to the top of every mountain and bottom of every valley. Virtual education and new employment opportunities will lift people out of poverty, new forms of financing will emerge to support overlooked entrepreneurs, and billions of brains will plug in to speed the world’s rate of innovation.
As of now, Elon Musk’s privately held SpaceX boasts the most mature network of satellites. After launching two test units in 2018, SpaceX put hundreds of satellites into orbit, and for $99 a month, it offers a beta-version of satellite-based internet.
Amazon is the other massive player, and in July of 2020, it received FCC approval to launch 3,236 low-orbit satellites. Before expanding beyond borders, the company’s Project Kuiper will target underserved areas in the US. “We have heard so many stories lately about people who are unable to do their job or complete schoolwork because they don’t have reliable internet at home,” said Amazon’s senior vice president, Dave Limp, in a press release. “Kuiper will change that. Our $10 billion investment will create jobs and infrastructure around the United States that will help us close this gap.”
A CONSTELLATION OF CONNECTION SpaceX plans to deploy 12,000 new satellites by 2022
1 circle = 100 units
Current SpaceX Satellites
Satellites planned for launch by end of 2021
Operating at a rate of 120 satellites per month, the speed is staggering. The previous record-holder for largest commercial satellite production, Iridium, deployed six a month.
With other well-funded newcomers, like OneWeb and Canada’s Telesat, joining the space race, there’s a clear and massive expansion of digital markets. By 2030, the UN expects the world population to hit 8.5 billion, with 90% of those ages 10 and up predicted to be online. That amounts to hundreds of thousands of new entrepreneurs, hundreds of millions of new workers, and literally billions of new consumers.
In their 2020 book, The Future is Faster Than You Think, Steven Kotler, a two-time Pulitzer nominee, and Peter Diamondis, founder of Singularity University, predicted that space-based internet would be unprecedented in its power to improve conditions around the world. “As the population online doubles, we’re likely to witness one of the most historic accelerations of technological innovation and global economic progress yet seen,” wrote the co-authors. “The whole planet is just a few years away from becoming the largest innovation lab in history.”
To their point, Morgan Stanley expects the space industry to be worth $40 trillion by 2040, and internet services will make up about 40% of it. The internet has already proven to be the biggest job and wealth engine in all of human history. Soon, the entire world will be able to take advantage.