LinkedIn emplea cookies para mejorar la funcionalidad y el rendimiento de nuestro sitio web, así como para ofrecer publicidad relevante. Si continúas navegando por ese sitio web, aceptas el uso de cookies. Consulta nuestras Condiciones de uso y nuestra Política de privacidad para más información.
LinkedIn emplea cookies para mejorar la funcionalidad y el rendimiento de nuestro sitio web, así como para ofrecer publicidad relevante. Si continúas navegando por ese sitio web, aceptas el uso de cookies. Consulta nuestra Política de privacidad y nuestras Condiciones de uso para más información.
A somewhat longer version of my Frontiers talk about technology and the future of the economy, with additional material pitched to an audience of Internet operators at Apricot 2017, in Ho Chi Minh City, Vietnam on February 27, 2017
The future is full of amazing things. On my way to Vietnam, I spoke out loud to a $150 dollar device in my kitchen, asked it if my flight would be on time, and then asked it to call a Lyft to take me to the airport. A few minutes later a car showed up. And in a few years, that car might well be driving itself. Someone seeing this for the first time would have every excuse to say “WTF?” That of course is an expletive that stands for What’s the Future? But a lot of people are reading the news about technology and the economy and are feeling a profound sense of unease. They are also asking themselves WTF? But in a very different tone of voice.
They read that researchers at Oxford University project that up to 47% of human tasks, including many white collar jobs, could be eliminated by automation within the next 20 years.
They’ve heard that self driving cars and trucks will put millions of people out of work.
They’ve seen calls for Universal Basic Income, with the assumption that there will be nothing left for humans to do once corporations outsource all the work to machines. While I think Universal Basic Income is an intriguing idea, I don’t think we need it because there will be nothing left for humans to do. There’s plenty to do. The problem is that
Our economy has the mistaken idea that the goal of technology is to maximize productivity, even if that means treating people as a cost to be eliminated.
Even leaving aside the obvious problem of injustice and inequality, this is the stuff of revolutions. Andy Macafee, the author, with Erik Brynjolfsson, of the Second Machine Age, once said to me, talking of the fear that robots will take over, “The people will rise up before the robots do.” We’re already seeing this in the US in the rise of Donald Trump to the presidency.
We’ve seen this happen before. In England, back in 1811 and 1812, a group of weavers waving the banner of a mythical figure named Ned Ludd staged a rebellion, smashing the steam powered looms that were threatening their livelihood. The weavers were right to be afraid. The decades ahead were grim, as machines replaced human labor, and it took time for society to adjust.
But those weavers of the Luddite rebellion couldn’t imagine that their descendants would have more clothing than the kings and queens of Europe, that ordinary people, not just kings and queens, would eat the fruits of summer in the depths of winter, luxuries brought from all over the world.
They couldn’t imagine that we’d tunnel through mountains and under the sea, that we’d fly through the air, crossing continents in hours, that we’d build cities in the desert with buildings a half mile high, that we’d put spacecraft in orbit, that we would eliminate so many scourges of disease! And they couldn’t imagine that their children, grandchildren, and great grandchildren would find meaningful work bringing all of these things to life!
Victorian England eventually rose to the challenge. Instead of sending children to work in the factories, they learned to send them to school. They shortened the work week and paid higher wages. And society became more prosperous. What is our failure of imagination? Why are we using technology to put people out of work rather than using technology to put people *to* work on the jobs of the future? What is our equivalent of sending kids to school instead of to work in the factories or the chimneys? What is our equivalent of the wonders that the industrial age brought to the world?
It isn’t technology that wants to eliminate jobs. Here’s what technology really wants. Nick Hanauer, a technology investor who was the first non-family investor in Amazon, put it best when he said: “Technology is the solution to human problems. We won’t run out of work till we run out of problems.” Are we done yet? Are we done yet?
Some global grand challenges technology can help us to solve - Climate change. - Rebuilding and rethinking the infrastructure by which we deliver water, power, goods, and services like healthcare. - Dealing with the “demographic inversion” — the lengthening lifespans of the old and the smaller number of young workers to pay into the social systems that support them. - Income inequality. “The people will rise up before the robots do.” - Displaced people. How could we use technology to create the infrastructure for whole new cities, factories, and farms, so people could be settlers, not refugees?
What does all this have to do with you? You are internet operators, after all.
To explain, let’s start with this guy. If you remember, starting about 20 years ago, everyone was focused on Linux as a desktop alternative to Microsoft Windows. Linux has made reasonable strides on the desktop since then, but it’s been far outstripped by the fundamental changes that Linux enabled for a new model of computing. In 2002, I started doing a talk called The Open Source Paradigm Shift, which began with a simple question: how many of you use Linux? Depending on the audience, maybe half the hands would go up, or sometimes, just a few. Then I’d ask, “How many of you use Google?” And every hand in the room would go up. People were still thinking of the computer industry through the lens of the dominant desktop paradigm, which was already on its way to changing.
In the old paradigm, software was an artifact, something that was made and then left to run. In the new paradigm, applications like Google and Amazon were ongoing business processes, with people hidden inside them, just like von Kempelen’s famous Mechanical Turk, the 19th century hoax that pretended to be a chess playing automaton but actually had a human hidden inside. Take the people out of an application like Google and Amazon, and they stop working.
The fundamental design pattern of internet applications is that they are hybrids of human and machine. Many of you are inside the machine, at the lowest levels of the stack. Without you, the entire infrastructure of the future would not work.
JCR Licklider, the legendary ARPA program manager who helped to fund much of the early work on the internet, foresaw this possibility in 1960. It took much longer than he thought, but it is clear that he was right.
These programmers at Pivotal look an awful lot like these workers in a Victorian sweatshop to me!
I’m really kidding, though, as is illustrated by these statistics. Low wage employers like McDonalds and Walmart are the new sweatshop, but something different is happening in tech. McDonalds has 440,000 employees and 68 million “monthly users,” while a company like Snap serves 100 million monthly users with only about 300 employees. How can that be?
At the same time, the algorithm is also the new shift boss. This is the inverse of the situation we talked about earlier, when programmers are the managers, and programs are the workers. There is a class of program now that is the equivalent of a middle manager, directing human workers. The algorithm is the new shift boss. You can see this clearly in a company like Uber or Lyft, where the managerial programs tell people where the demand is, track the performance of the job and how to pay for it, and even solicit feedback from the customers about the human worker’s performance.
This is where things get tricky. When you’re writing a program that serves users directly, measuring user satisfaction is all you have to worry about. But when you’re writing programs that will also manage human workers, you have a real responsibility to make sure that those managerial algorithms are taking care of their workers. Companies like Uber have set their algorithm to optimize for only one factor: passenger pickup time. Until recently, drivers have been considered a throwaway commodity. That’s a big mistake. It’s becoming increasingly clear that many Uber drivers are being paid less than a living wage.
So too are old economy companies, which are also using algorithms to manage their workers. At most multinational low wage employers, scheduling software is increasingly used to manage workers, telling them when to show up, and firing them when they don’t. Part of the appeal of services like Uber is that they put the control about when to work in the hands of the workers.
Fortunately, even in the old economy, some companies are starting to take notice, and realize that the scheduling algorithms need to take the human workers into better account. For example, in 2014, Starbucks ended the dreaded “Clopen” - in which an employee, who might live an hour away, was assigned to close the store at 11, and reopen it at 4 or 5 am. Many companies still follow this practice, though.
I like to hope that the current state of people management algorithms is somewhat akin to the state of search engines before google, crowded with ads and not doing a very good job of satisfying all of the user needs of the ecosystem. This is a screenshot of Altavista from 1996. I’m old enough to remember how bad search was back then.
Google came along and not only made algorithms that were focused on better search quality, but also algorithms focused on better ad quality. That’s what ride sharing companies need to do today - improve their algorithms to manage their human workers not just for passenger pickup time and customer experience, but also to make sure that the drivers themselves have a good experience and can make a good wage.
It’s easy to blame technology for the problems that occur in periods of great economic transition. But both the problems and the solutions are the result of human choices. As we saw during the first industrial revolution, society suffers if the fruits of automation are used solely to enrich the owners of the machines, and workers are treated as a cost to be eliminated, or as cogs in the machine, to be used up and thrown away. But as I mentioned earlier, Victorian England figured out how to do without child labor, with reduced working hours, and guess what, society became more prosperous. We saw the same thing here in the US during the 20th century.
It’s important to remember, that, as the labor organizer David Rolf, points out, “God did not make being an auto worker a good job!” We made choices as a society to share the fruits of machine productivity more widely. We can do that again.
One of the most powerful and important digital lessons comes from Amazon, which over a period of a few years transformed itself from being simply an e-commerce site into the platform on which many of the world’s businesses rely, creating a new “cloud computing” industry in the process. What most people don’t realize is that Amazon restructured itself internally to use its own platform before it made it available to others.
As you think about digitalization, it’s important to remember that, as Deloitte’s Josh Bersin put it, “Doing digital is not the same thing as being digital.”
Some of the most important work on 21st century management technique is coming out of the DevOps movement. Understanding how to keep systems working at scale is incredibly difficult, and it is hard for people outside the industry to grasp just how important the techniques developed in our community have become.
One of the things that is essential to understand is that, as Tom Steinberg, the founder of the UK non-profit MySociety, once wrote, “Good governance and good policy are now inextricably linked to the digital.” Digitalization is no longer an option. We have to get it right.
Working with the new United States Digital Service, the White House developed its new college scorecard program using these techniques. As Haley van Dyke, Deputy Director of the USDS told the story, the USDS team that built the software was raked over the coals for not incorporating features requested by White House staff. They replied that they had tested these features on thousands of students, and none of them used them, so they took them out of the product. When staffers protested that they wanted their features in the product, Cecilia Muñoz, the Director of the White House Domestic Policy Council, backed up the USDS team, saying “ “This isn’t just how we should be developing software. It’s how we should be developing policy.”
This is not typically how governments work. Massive programs are enacted based on some one’s theory of change The implementation of that program is outsourced via a contracting system looking for the lowest bids Governments have expertise in policy, not in implementation Implementation is what makes the difference Measurement of success and failure, when it happens at all, is in timescales of years or decades
One of the most useful documents outlining how to apply the best practices of the technology industry to government is the UK Government Digital Service design principles. It’s unfortunate that after the change in government, the UK GDS has been sidelined, with a return to business as usual. I worry that the same thing may happen in the US.
My wife, Jennifer Pahlka, founder and executive director of the nonprofit Code for America, which works to bring government services into the digital age, and who was the co-founder of the United States Digital Service, likes to say that government has to move “from apps to ops.” In moving to digital, government has too often simply recreated its old paper-based processes rather than reinventing them. It’s essential to rethink services in light of what is now possible.
The greater speed and scale of the systems we use today make a compelling argument that the tools government relies on are far, far too slow. As former IBM fellow Jeff Jonas noted, “Would you cross the street with information that was five seconds old?” Yet government statistics for managing the economy are usually years old, while hedge funds and other financial players are no longer forecasting, but, as Google chief economist Hal Varian calls it, “nowcasting.”
Many information problems we face today are analogous to the Facebook “fake news” controversy. Frankly, I think Facebook and Google have a far better record of policing themselves in the public interest than government does with, say, bad actors in the financial system.
There have been many calls for Facebook to use human fact checkers to eliminate fake news. Users post 7 billion pieces of content to Facebook a day, and most stories that go viral do so in a matter of hours. Expecting human fact checkers to catch fake news is like asking workers with picks and shovels to build a modern city. At internet scale, we now rely increasingly on algorithms to manage what we see and believe. Those algorithms are the tools of human judgment, not a replacement for it. Just like the giant machines we see on the skyline of a growing city, they are the tools of human effort.
But as with every new technology, one of our grand challenges is figuring out how to make sure we understand and manage the consequences. As more and more systems are run by algorithms, we have to ensure that we understand what it takes to understand what is going on in the black box. We have to ensure that these algorithms are not controlled by only a few giant corporations but becomes the common heritage of mankind. And we have to ensure that every data/CS training program has ethics and security embedded through out the curriculum.
Cybersecurity is another of our great challenges. We will need to apply Machine Learning. The DARPA Cyber Grand Challenge asked for the development of AI to find and automatically patch software vulnerabilities that corporate and government IT teams just aren’t able to keep up with. The problem is that an increasing number of cyber attacks are being automated, and as one knowledgeable friend of mine, who wishes to remain anonymous, remarked, “It takes a machine to get inside the OODA loop of another machine.” (OODA: Observe, Orient, Decide, Act.) Image from https://www.cybergrandchallenge.com/press#photogallery
There’s a fascinating story about the birth of commercial jet aviation. Britain was poised to lead the way with its pioneering aircraft, the DeHavilland Comet. But suddenly, for no reason, the planes began to fall out of the sky. First they blamed bad weather and pilot error, but when it happened in clear skies, they realized something was wrong. An analysis determined that it was metal stress fractures. DeHavilland had tried to design the planes so they could resist all stress, but a young engineer from Boeing realized that the right answer was to allow tiny cracks, but to make sure that they couldn’t propagate too far. This is very similar to the “fail soft” concept from computer security. Robust systems don’t prevent all error; they recover from it quickly and gracefully.
Something like an autopilot gives us a useful model for thinking about the general rules for regulating algorithms. Regulations Must focus on outcomes, not on rules. Must operate at the speed and scale of the systems it is trying to regulate. Must incorporate real-time data feedback loops. Must be robust in the face of failure.
And in the case of some algorithms Must address the incentives that lead to misbehavior. Must be constantly refined to meet ever-changing conditions.
Financial systems, and information systems like Google and Facebook are in this latter category, since they are effectively battlefields between the service provider and those who want to game the system.
Digitalization is coming to the real world. Matt Cohler, one of Facebook’s earliest employees and now a venture capitalist, noted one key point about his investment in Uber. He said “the smartphone is becoming a remote control for real life.” In the first wave of the internet, we digitized media. We are now deep into the process of digitizing every real world service. The opportunities are enormous.
This is a key takeaway from on-demand apps. Don’t get all caught up in whether or not being an Uber driver is better or worse than being a licensed taxi driver. Think about how technology can transform real world processes. We have to reinvent processes, not just duplicate them.
In this regard, it’s worth noting that speech interfaces may have an even bigger impact than the smartphone. We’ll soon not just be talking to household devices like the Amazon Echo, but to our cars, our homes, and our appliances. And just as the smartphone expanded the impact of the internet far beyond the PC – and with it, the security and operations challenges of keeping it running – speech will accelerate the pervasiveness of computing throughout our society.
We have to stop thinking that the internet of things is about devices. It is about pervasive networking. I like to point out that it is Uber, not Nest, that teaches us the most about the internet of things. Because Uber relies on the phone, not on a dedicated device, it is easy to overlook the fact that it depends entirely on a network of sensors, connecting the driver and passenger in real time. The data from those sensors is used to completely rethink the workflow of on-demand transportation. IoT isn’t about the device, it’s about what the system of connected devices makes possible.
Uber has been very problematic in its management practices. But as an application, it teaches us something very important about the future. We had connected taxicabs before Uber. They just stuck a credit card reader in the back along with a television screen to show ads. It took a radical rethinking of the possibilities lying latent in digital technology to realize that a smartphone in the hands of both drivers and passengers meant that it would be possible to completely change the way transportation was summoned, who provides it, and how payment is collected. What business processes are you simply recreating in the digital era, when you should completely reinvent them? Aaron Levie, founder of another internet startup, box.net, made this admiring comment about Uber that has always stuck with me. “Uber is a lesson in building for how the world should work instead of optimizing for how the world does work.”
Perhaps the most important thing that Uber teaches is that the real opportunity is not in using technology to replace people, but to augment them so they can do something that was previously impossible. How crazy is it that even here in Vietnam, you can use the Uber app to summon a motorcycle to take you where you want to go? How WTF? Would that have been only a few years ago!
The third reason is that economic transformation takes time and effort. It’s easy to underestimate the amount of economic activity that’s involved in simply updating our technology! Goldman Sachs Investment Research estimated that it won’t be until 2060 that all existing vehicles will be upgraded to full autonomy, because unlike mobile phones, which are typically turned over every year or two, vehicles remain in use for an average of 20 years! But even pure digital transformation takes time. We’re 25 years into the commercial internet era, and online advertising is just on the cusp of passing television. Amazon, for all its success is only 20% the size of Walmart. There’s a long way to go! As Jeff Bezos likes to say, “It’s still day one!”
The “self-driving airplane” has been with us for decades. I am not a pilot, yet the pilot of this plane had no worries about letting me take the controls. He was able to trust the airplane’s autopilot because the desired outcomes are well known. Aviation has been extensively studied, and we know what success looks like. I sat with my hands on the controls, and felt them being moved by the autopilot, responding to minor changes in the air currents, keeping the plane on course.
Air travel became much safer, and the market for it expanded. Tourism became a much larger worldwide industry. The rise of computerized logistics also brought about our world of on demand delivery. What Fedex started, companies like Amazon are intent on finishing.
But Uber lost the plot when they started talking about self driving cars. Rather than crowing about how they’d finally get rid of those pesky drivers, they should have been talking about an experiment that they’ve run since 2014, delivering flu shots. “Sure, we won’t always have drivers. But just imagine how many other jobs we can restructure and make more magical and on demand once the transportation is even cheaper and more convenient!”
Zipline, a California startup working in Rwanda, is a great example of reinvention. It shows how two of the latest technologies, on-demand and drones, can utterly transform how we think about healthcare delivery. They have been doing a pilot project, delivering blood and critical medicines to isolated local clinics. The country has poor, often impassable roads, and lacks developed hospital infrastructure. Postpartum hemorrhage is a major cause of death. By drone, blood can reach any corner of the country in 15 minutes or less.
We should be thinking about how technologies like on demand and self-driving cars would let us reinvent public transportation and the shape of our cities, not regulating them as a threat to incumbent 20th century industries! The key to making good use of new technology is to keep your eyes fixed on the fitness function of government, which is the greatest good for all of society. Embrace the future. Don’t fight it.
WTF can also stand for “Welcome the future!”
Thank you very much.
I’ve been running an event called the Next:Economy Summit on how we can put what I call WTF technologies to work building a better world.
Mark Zuckerberg and Priscilla Chan’s announcement that they are funding an initiative to cure all disease within their children’s lifetime is a great example. It’s hard to imagine that AI won’t play a major role in achieving that ambitious goal.
Al is already being used in both clinical practice and in research. Deep mind is sifting through millions of eye scans.
And the White House Precision Medicine Initiative has already helped blaze that trail, with a vision of tailoring treatment to each individual patient.
AI will play a huge role in precision medicine. I heard recently from one startup that of the over 1 million full human genomes that have been sequenced, only 49000 have been interpreted. That’s a job for AI.
And already AI is being used to analyze millions of radiology scans at a level of resolution and precision impossible for humans, to keep up with the flood of medical research at a level that can’t be accomplished by a human practitioner.
Even more exciting, we’re studying the brain and how it works, with the prospect of creating prosthetics that give their users a sense of touch, that allow direct brain control of devices, and even “brain prosthetics” to deal with neurological diseases!
Augmented reality and telepresence are two other technology tools that allow us to rethink how we deliver digital services in the 21st century. A company called Pristine lets a remote specialist be called in to review emergency room patients, seeing the same thing as the on-premise physician. Another company called Augmedix uses Glass to let a remote observer take physician’s notes, rather than having the doctor focused on the computer instead of the patient. You can imagine how technologies like these could be used to upskill community health workers, giving them access to expertise on demand as they need it, while preserving the human touch. Put this together with on-demand transportation, and you could reinvent house calls for the 21st century. Just as we used industrial age technology to augment human workers so that they could do tasks that were previously impossible, so too this should be our goal for digitalization. Augment and empower people to do things that were previously impossible.
How about Climate Change? Climate change is for our generation what World War II was for our parents and grandparents, a challenge that we must rise to or suffer dire consequences. Already in data centers, AI is radically increasing power efficiency. How do we rethink and rebuild our electric grid to be decentralized and adaptive? How do we use autonomous vehicles to rethink the layout of our cities, making them greener, healthier, better places to live? How do we use AI to anticipate ever more unpredictable weather, protecting our agriculture, our cities, and our economy?
And God knows, we are already seeing the scourge of war, and the massive refugee crisis that it has set in motion. Do we just accept that as the cost of doing business? Or can we solve for that, rebuilding as the US helped rebuild Europe after the scourge of World War II?
We also made choices to invest in the future. That golden age of postwar productivity was the result of massive investments in in roads and bridges, universal power, water, sanitation, communications. Louis Hyman, author of Borrow: The American Way of Debt, pointed out that we went from 10% of homes in the US having electricity in 1930 to 60% ten years later, simply by putting idle capital and human ingenuity to work. After World War II, we committed enormous resources to rebuild the lands destroyed by war, but we also invested in basic research. We invested in new industries: aerospace, chemicals, and yes, computers and telecommunications.
Charts like this one, from Max Roser’s Our World in Data, documenting the march of progress during the 20th century, are what we aspire to.
I want to leave you with the wise words of Spiderman: “With great power comes great responsibility.”
You are building the infrastructure of the future. I think you know that better here at Apricot than we do in Silicon Valley. You see every day what power the internet has to make a better world, and you’ve spent much of your lives working for that outcome. I hope I’ve given you a taste of where all this is going and why it continues to matter.
“…47 percent of jobs are
“at risk” of being automated in the next 20 years.” Carl Frey and Michael Osborne, Oxford University “The Future of Employment: How Susceptible Are Jobs to Computerisation?”
It isn’t technology that wants
to eliminate jobs “Technology is the solution to human problems. We won’t run out of work till we run out of problems.” Nick Hanauer
Some global grand challenges technology
can help us solve • Climate change. • Rebuilding and rethinking the infrastructure by which we deliver water, power, goods, and services like healthcare. • Dealing with the “demographic inversion” in developed countries— the lengthening lifespans of the old and the smaller number of young workers to pay into the social systems that support them. • Income inequality and income insecurity. • Displaced people. How could we use technology to create the infrastructure for whole new cities, factories, and farms, so people could be settlers, not refugees?
“The hope is that, in
not too many years, human brains and computing machines will be coupled together very tightly, and that the resulting partnership will think as no human brain has ever thought and process data in a way not approached by the information-handling machines we know today.” - J.C.R. Licklider, Man-Machine Symbiosis,1960
We are building a global
brain, composed of all of us, augmented and connected by technology. You are responsible for the health of its nervous tissue.
2. Many of today’s workers
are programs. Developers are actually their managers Every day, they are inspecting the performance of their workers and giving them instruction (in the form of code) about how to do a better job
This is where things get
problematic Algorithms can no longer just be optimized for the user. They have to take into account the needs of workers too. Companies like Uber are far behind the curve in realizing this.
It was no accident that
the rescue of healthcare.gov in the fall of 2013 was carried out by a team many of whose key players were Site Reliability Engineers from Google.
5. Government too is being
reshaped by the digital “This isn’t just how we should be developing software. It’s how we should be developing policy.” Cecilia Muñoz, Director, White House Domestic Policy Council
How governments typically work •
Massive programs are enacted based on some one’s theory of change • The implementation of that program is outsourced via a contracting system looking for the lowest bids • Governments have expertise in policy, not in implementation • Implementation is what makes the difference • Measurement of success and failure, when it happens at all, is in timescales of years or decades
Users post 7 billion pieces
of content to Facebook a day. Expecting human fact checkers to catch fake news is like asking workers to build a modern city with only picks and shovels. At internet scale, we now rely increasingly on algorithms to manage what we see and believe.
But we do have a
responsibility to get the transformation right! The great question of the 21st century will be “ Whose black box do we trust?” John Mattison
“It takes a machine to
get inside the OODA loop of another machine.” Observe Orient Decide Act A DARPA Cyber Grand Challenge team
7. We need a new
approach to regulation One that is akin to an airplane autopilot, focused on outcomes, not rules. One that is based on real-time feedback loops. One that is tolerant of failure, not one that tries to prevent it.
Regulation in the age of
algorithms Must focus on outcomes, not on rules. Must operate at the speed and scale of the systems it is trying to regulate. Must incorporate real-time data feedback loops. Must be robust in the face of failure. Must address the incentives that lead to misbehavior. Must be constantly refined to meet ever-changing conditions.
Economic Transformation Takes Time and
Effort Goldman Sachs Investment Research estimates it won’t be till 2060 that all vehicles are autonomous. The AI transformation itself will drive the economy for decades to come.
“The New Deal’s Reconstruction Finance
Corporation helped light up America — moving it from 10 percent of homes having electricity in 1930 to more than 60 percent a decade later. [We] utterly transform[ed] the economy in about five years, by using idle capital.” –Louis Hyman Borrow: The American Way of Debt