Is Chinese Censorship Really About Censorship?

comment 1

Take a look at Capital And Main for the podcast version of this, and the original appearance of these ideas, among far more interesting ones. And to it…

For most Westerners, information technology in China is headlined by the fact that it is censored. If you travel to China, you will find that you can’t load a Google map, check your Instagram likes, send your tweets, or even read articles from the New York Times.

After the riots in 2009 in the western province of Xinjiang, Facebook was blocked, along with other social media services. Chinese citizens like Wang Yi have been sentenced to labor camps for retweeting comments critical of the Communist government, and dissidents have met similar fates. The Communist Party in China maintains a tight rein over information flows, and it is doing its best to control the fluid and unpredictable nature of social media.

But other Western technology companies have struggled in China, too: Uber, once notorious for its ability to muscle a sizable share out of even the most intractable markets, merged with Didi Kuaidi, the Chinese competitor, after finding myriad challenges establishing itself in the country.

Amazon, arguably the Western world’s most successful modern company, *lost* market share from 2011 to 2016, while Alibaba, and other Chinese competitors continued to grow in the double and triple digits.

What, then, if the censorship piece—which looks, from a Western, free information, democratic lens, to be a political decision—is not only political, but a business decision, as well?

Consider this: The Chinese market is 1.4 billion people, or 18% of the world’s population. As culture, content, and commerce have spread thanks to the Internet and international travel, companies that want to succeed in the long-term have to take a global view.

Nobody understands this better than the Chinese government, which has invested hugely in mining natural resources in Chad, Angola, and other parts of sub-Saharan Africa, while dramatically expanding its telco footprint in the Middle East and Northern Africa.

The customer base for a company in the 21st Century is not simply those where the company is headquartered. It is, rather, the whole world.

At the just-concluded World Economic Forum in Davos, Jack Ma, the charismatic founder of Chinese tech conglomerate Alibaba, said, “I think globalization can not be stopped. Nobody can stop globalization. Nobody can stop trade.”

That is, except China.

The math is simple: Sina Weibo, Baidu, and Tencent—competitors to Twitter, Google, and Facebook, respectively—have an 18% bigger global market of potential customers than their American counterparts, so long as the latter are banned in China.

Didi Kuaidi merged with Uber, invested in Lyft, Ola (the Uber of India) and Grab (the Uber of Southeast Asia). Imagine operating as a global company, and having your government actively supporting you not just through tax incentives and diplomacy, but by outright banning your competitors. Whether this is the explicit intention of the Chinese government or not, it is incredibly effective.

Whatever your point of view on Sino-American relations, this muscular form of interventionist government—of business development as a form of realpolitik—deserves as much prominence in the discussion around China and the future of technology as political censorship does. It has led to a marketplace where Chinese companies have a structural advantage against their competition.

To be sure, political control of business fits tightly into the narrative of a Communist government, and it would be naive to assume that it is *not* about censorship. But Communist governments have censored media in the past. This one feels different: like censorship with business development attached.

As for what the West can do in response: Perhaps the answer lies less in igniting a trade war, as President Trump seems bent on doing by slapping tariffs on Chinese-made solar panels and threatening similar action on other products, and more in fostering an entrepreneurial environment that leads to the creation of truly irresistible goods and services. After all, soft power is contagious.

There are now tens of thousands of youth in China using virtual private networks, or VPN’s, to post and share on Instagram, a Facebook company. And no wonder. That’s where the soccer players and movie stars and fashion icons are. China is in no position to replicate that.

Perhaps free and open society supports a different kind of business, and perhaps that kind of business has something to offer that others do not. Over the long haul, China may have no choice but to move in this direction.

Thanks to Rick Wartzman at Capital and Main for editing and publishing me on this one!

Self-Driving Game Theory

Leave a comment

Take a look at Capital And Main for the podcast version of this, and the original appearance of these ideas, among far more interesting ones. And to it…

Whether it takes 15 years or 50, it is almost certain that self-driving cars will be on the road within my lifetime. Waymo, the Alphabet company, is at 3 million test miles driven. Uber, BMW, Cruise (in partnership with General Motors), Mercedes, Volvo, Nissan, Ford, and others have also been logging hundreds of thousands of miles with no one behind the wheel. Tesla drivers, meanwhile, have notched 300 million miles using the company’s Autopilot autonomous driving feature.

At the exponential pace of software adoption, the technology is nearly ready for the road. But technological hurdles are one thing; cultural barriers are bound to prove far trickier to overcome.

A 2017 study by Deloitte found that three-quarters of Americans do not trust driverless vehicles. The American Automobile Association found that 54 percent of drivers feel less safe even sharing the road with fully autonomous cars. But how safe is safe enough?

To better understand this question – the main question causing vexation among insurers and others – think about the psychology of the driver. Implicit in my decision to get on the road and drive is an understanding that there are other drivers on the road, as well. And every interaction with another moving vehicle is an exercise in game theory: that is, intuitively modeling what I expect the other driver to do as a function of what I’m going to do. Yet how does game theory work with a self-driving car? Am I supposed to anticipate what an algorithm would do?

Self-driving cars tend to come in fleets. Each one is automatically part of a network of others that share its software. This allows the machine learning processes to run at the rate of *all* the driving data captured by all the cars on the road, instead of just one.

This also means that the cars can talk to each other at the speed of a super-fast wireless connection. In an interaction between two self-driving cars, the game theory is not between them, but between the entire fleet and any other entity, since they have perfect information about each other. Imagine being a driver, then, and approaching an intersection where all the cars know exactly what the others will do. Of course, each self-driving fleet will have its own software, and thus far, there is no sign that there will be any integration between different company fleets. So, the likelier scenario is: You approach an intersection, and you don’t know if the cars are talking about you or not. And if they are, what are the saying? Now consider a split-second decision at high speeds, or in confined space. Unnerving, right?

Each year, more than 30,000 Americans die and many more are injured in car accidents, the vast majority of which are caused by human error. Driverless cars could eliminate 90% of these deaths and injuries, according to experts. But these numbers—impressive as they are—may not matter very much.

The fear of a road full of self-driving cars is a fear that machine game theory, even if it is explicitly designed to avoid accidents, is not perfectly compatible with human empathy. And in the moments of inconsistency, what will the robot do? For example: If there is a chance to protect two pedestrians, but it requires potentially injuring a rider, which does a fleet of self-driving cars choose? What if you’re the one sitting in the passenger seat? Colloquially, ‘what if the software malfunctions?’ gets at the same point. Human malfunction is reasonably predictable. I know that accidents happen when a driver is drunk, distracted, sleepy, or overly upset. But what might cause a machine to crash—literally? Being unable to relate on a human level makes it much scarier, even though the reality may be much safer. Solving the social resistance to self-driving cars (beyond concerns that they could cost millions of jobs) will not come down to safety statistics, or even to miles driven without incident. Those don’t actually matter nearly as much as policymakers and technologists think. Signaling, and communicating a clear sense of fairness – and if not empathy, then something that rhymes with it – will matter much more.

Thanks to Rick Wartzman at Capital and Main for editing and publishing me on this one!

The Rise Of Startups… Or not

Leave a comment

They call it a Cambrian moment. These days, every city worth its salt comes with a startup ecosystem: co-working spaces and accelerators mark emerging hubs across the United States and around the world. Since The Social Network, high school students have grown up with the legend of the college dorm room startup as a central narrative. From healthcare to insurance to apparel to food, it seems as though there is a startup for every piece of our economy today, and 10 more behind each one. They seem to be everywhere. 

In recent years, the number of new venture capital firms has exploded, raising to hundreds of new fund formations a year. My Limited Partner friends tell me they see 300+ new fund pitches per year. Accelerators are rising follow-on funds, athletes are raising side project funds, and scout programs are launching as standalone platforms to fund the early stage. Meanwhile, the late stage has similarly continued on a great fundraising run: there have been multiple multi-billion dollar funds closed, Softbank has committed $200B into the ecosystem, and the middle eastern sovereign capital pools are investing heavily into tech. And there seems to be no end to growth in the space: as global yield stays low, hundreds of billions of assets are looking for a home, and finding promise in the global tech sector. Startups. What’s more, after the global financial crisis in 2008, large corporations had shrunk, Millennials were graduating into uncertain job markets. Youth unemployment was startlingly high, from Spain to Iran to South Africa, and everything in-between. Necessity is the mother of invention. Of course, startups rushed in to fill the void.

Or have they? Take a look at this chart, published by the Kauffman Foundation in this great research report (pdf link) about new business creation in the United States:


It shows that the rate of business owners nationally has actually fallen by over 20% since 1996, with an even more precipitous decline in 2008. So first of all, more startups weren’t created after 2008. Dramatically fewer were. And while new business formation has rebounded since 2008, it is still lower than it was 30 years ago. But take a look at the chart below, from the U.S. Census Bureau, and you’ll see that, in fact, new business formation in the United States is at a 40-year low!


So what gives? I conclude a few things:

First of all, the word startup has taken on a special connotation, and implies the formation of a unique type of company. I’m not convinced by this formulation: 7-Eleven was venture-backed. Blue Bottle Coffee was a venture-backed company. The fastest growing company over the last five years, by revenue, was a Utah-based bootstrapped multi-level marketing company called Younique Products. It was founded in 2012, and had $400 million of revenue by the end of 2016. Seriously. Look it up. So– any business can grow fast and deliver 20x, 30x, etc. Any notion that there is a certain *type* of company that is venture fundable is flatly wrong. But this mindset is particularly relevant among venture capitalists lately. My friend Satya captured it well: 

while there is more money than ever in VC there is also more risk aversion and less independent conviction. stranding many companies that are building solid businesses in “out of favor” markets or in markets where “venture scale returns” are not a straight line path.

— Satya Patel (@satyap)

November 28, 2017


Word. In fairness to venture capitalists, liquidity has been particularly hard to find lately, which may be affecting how they think about deploying their capital. The hive mind is more intense in venture capital today than it has been in my 7 years in the business. Valuations have been blown out of proportion among “fundable” companies, while those with promising but early trajectories, those with ambitious but workmanlike metrics, are perennially struggling to raise capital. My favorite quotation to capture this phenomenon is: deals these days are badly undersubscribed until it is badly oversubscribed. 

Second of all, the United States is in a period of possibly the most intense consolidation in the innovation and technology ecosystems since the 1870′s –– when, in an effort to entice European businesspeople and workers to the United States, *massive* government contracts were granted to corporations to fund expansion of the United States project, leading to the passage of the Sherman Antitrust Act in 1890. Look at the top companies in the United States by market capitalization: if you take out oil companies and investment corporations, you have Apple, Google, Microsoft, Facebook, Amazon. Semil Shah covered how their presence affects startup investing (, but it’s worth noting that the phenomena that have lead to these companies’ massive, immense success is not isolated to them. The network effect, and the focus on consumer surplus as a means of crowding out competition, is endemic to the software-enabled innovation company today. 

Competition and capitalism are actually not compatible, as Peter Thiel might say. And network effects, focused on consumer surplus, scalable demand generation, and very thin aggregation of crowds, are a *really* good manifestation of capitalism. Why start your own private practice, when economies of scale make lifestyle better to be part of a network? How do you open a small business when a big network-based company is offering a cheaper service? How do you find an edge in technology spaces that require data, when the incumbents have all the network effects and incredible expertise in attacking adjacencies?

Finally, given both of the former two points, there is a broad-based misunderstanding about startup formation today. People say: “Cloud-based services make it easier than ever”. “Pre-seed capital invests earlier”. “More funds than ever before”. But the truth is, the *vast* majority of capital that invests in startups comes after the critical period that matters most for startup founders: the “friends & family round”… the “just bootstrap it” round… As we all know, most Americans can’t afford an unexpected expense of $400. And as I’m sure most of us can intuit, the ability to raise a “friends and family round” is not evenly distributed according to talent. So taking a year, 6 months, or even a quarter, to quit your job and start a new business is simply a nonstarter for most Americans today. In a world where the small business bank loan has all but evaporated (particularly post-2008), many Americans – and others around the world – aren’t starting companies because they just can’t afford to. When we talk about diversity in tech, I’m interested in hearing about the structural impediments to startup success. Ultimately, many of them can be boiled down to access to capital. Large, urban coastal centers are attracting capital and resources aplenty, while mid-sized and middle-of-the-country towns are not keeping apace. Within those urban coastal centers, wealthy, well-connected, mostly white males are raising more and more capital at higher and higher valuations, while other demographics are not keeping apace. Even within the demographics that overindex in startup formation, there is a very strong sense of “haves” and “have-nots” which is intensifying, even in the midst of a lot of noise about diversity and startups. This is pushing inequality further, slowing our ability to innovate writ large, and keeping too many Americans outside of equity participation in this country. Economists across the political spectrum will agree that new job growth comes from new business formation. And businesspeople of all stripes will tell you that equity upside is the most important input for building wealth. This isn’t just a matter of building cool technology – though that is impacted, as well – this is a matter of how nations can grow to be healthy, and supportive of their citizens. I was pleased to read Sam’s post American Equity yesterday, as its clear he recognizes the same issue I do. The central planning and policy-prescriptions for this issue should range from ideas like his, to the varying universal basic income (UBI) concepts, to investing in incentives for banks to lend for small business formation. All of this is good.  But the market driven approaches will need to supplement these – frankly, the central planning approaches will need to supplement the market driven approaches. And this starts with recognizing where we *actually* are. I’ve spoken to too many people in the tech world who say there are “too many funds” or “not another app”.

Quite the opposite, if you ask me.

About Proprietary Networks

Leave a comment

They are the among the only lasting advantages in business – particularly in the fast-moving technology environment. And the question about how to grow one is one of the most common pain points for someone early in their career, particularly if they did not get lucky and choose the right company to work for, where the network was built-in for them.

So here’s a word: most people wrongly focus on the “network” side of the equation before the “proprietary” side. It’s relatively easy to guess somebody’s email address. And it’s a toss up who will respond to a cold email. It’s relatively easy to meet a luminary – attend the right conference, hang around after she speaks, you may get 5 minutes. But if you do this before the “proprietary” part, you’re getting it backwards.

Proprietary, in this context, means you are known *for something*. You are useful, have an expertise, or even simply a well-researched interest somewhere. If you reach out to me as someone generally interested in getting ahead, why do you stand out over any other person who shares that desire? But if you are yourself an expert, or have a truly differentiated point of view, then it’s a different game – you’ll often find they actually want to meet you as much as you want to meet them.

If you want to build out your network, step one: stop networking. Practice thinking independently; reading, writing, and building. Become the most interesting person in your field.

“Insurance Is Sold, Not Bought.”

Leave a comment

In an influential 1979 behavioral economics paper, Daniel Kahneman and his colleague Amos Tversky developed “Prospect Theory” as a way to make sense of decision-making. The summary of the paper was, if I may, that humans do not make optimal decisions, which normative (”should”) frameworks suggest, but instead have irrational aversion to certain losses, and minimize the probability of other losses.

Here’s an example: 


Here, Kahneman (via Thinking, Fast and Slow) outlines the cases where a decision-maker irrationally minimizes potential losses versus maximizing potential gains. As the theory goes, humans sometimes choose to be risk-averse, even it results in an unfavorable outcome. The opposite also applies: we sometimes are risk-seeking even if it results in an unfavorable outcome. The intersection of actuarial science and behavioral economics purports to find itself in the bottom right box. You choose to buy insurance, which is $1 more than the statistically-adjusted cost of “going for it” because of the fear of losing $10,000. At least, that’s how the model suggests it would be, right?

It is illegal to drive without auto insurance in all 50 states of the country today. The penalties range per state, but it is on the order of $150 to $1000 for a first offense, and in some states even a month’s imprisonment! And yet, in 2012, 1 in 8 motorists were uninsured.* That number is high enough to suggest that the bottom right quadrant of the matrix above does not apply as cleanly as we would hope it to. Some drivers, of course, flatly can’t afford a monthly payment, or have a risk calculus that combines the risks relating to auto insurance with other financial risks in their lives. That is to say, it’s impossible to know what everybody’s short term financial needs are; walk a mile in another man’s shoes, right? But nonetheless, many drivers are simply risk seeking in a way that breaks even this probability distortion model. Of course, the whole point of prospect theory is that actual decision-making includes a variety of probability distortions and mental shortcuts. And in this case, even when it is mandated by law, with a financial penalty not only for accidents, but also for non-adherence, double digit-percentages of drivers still go without auto insurance. Mental shortcut, indeed. Health insurance, which is (for now) mandated by federal law, has comparable adoption numbers. The uninsured rate for health insurance was 11.9% in 2012.**

The insurance policies that people buy often fall into two categories. One: mandated by law, with scary non-adherence implications; two: mandated as a means to accessing something else you really want, like a house or an apartment. Other than that, many consumers choose to wing it, either because there are short term cash needs which appear more pressing, or because they believe it “wouldn’t happen to them”. These behavioral quirks in our psychology are really important for insurance tech companies to study, and learn well. Insurance tech is very “hot” right now. Many startups are thinking about the great ways that usage-based insurance, flexible and tied to smaller, more granular actions, can be more efficient, empowering, and consumer-friendly. They’re right. But we think we’re safer, healthier, and better protected than we are. So, a caution in designing your business model to account for what will likely be expensive customer acquisition. A caution in designing technocratic policies, especially if you’re going to model consumer behavior. Insurance is sold, not bought.

MAYA: Most Advanced Yet Acceptable

Leave a comment

Two anecdotes stuck out to me in Derek Thompson’s Atlantic post about “The Four-Letter Code to Selling Just About Anything.”

First, the concept that people prefer the image of themselves in the mirror to that in photos. I certainly do, and often feel like I look lopsided in photos. But that makes sense, since I’m seeing the mirror image of what I’m used to seeing. I imagine the same applies to the phenomenon of hearing one’s voice (which happens far less often). If you are used to hearing it come from your own mouth, with the baritone coming from your lungs, the nasally effect manifesting in your own nose, et cetera, that is the familiar sound and so something close, but not quite there, is jarring.

Second, the clustering of popular names: Derek describes research done by Stanley Lieberson which concludes that “Most parents prefer first names for their children that are common but not too common, optimally differentiated from other children’s names. This helps explain how names fall in and out of fashion, even though, unlike almost every other cultural product, they are not driven by price or advertising. “

Our subconscious does more work than we give it credit for, and drives the ‘facts’ which we accept in product design, brand development, fashion, food, and even what constitutes ‘beauty’. As per this article, we all share a programmed yearning for familiarity, but perhaps not at the conscious level. That familiarity, developed in our early evolution, establishes trust. But too familiar, and our conscious brain kicks in, as we also want individuality, and the ability to be unique in our expression and decisions. MAYA, or “most advanced yet acceptable” is Raymond Loewy, the iconic 20th century industrial designer’s, framework for describing this exact concept.

I always used to find it funny that so-called “creative types” dressed in similar ways, or how uniform the “alternative kids” trope was across American grade schools, or how certain typefaces fall into favor (or out of it, Comic Sans) in the name of “good design”. Why is the Scandanavian aesthetic, and mid-century modern furniture, found in so many Millennial households today?

Two easy conclusions come to mind.

First, beauty and style are often not so subjective as we let ourselves believe, but in fact follow a form that maps to the nature of our social psychology and neurology. We value belonging and the feeling of independence, and a healthy tension between those in a brand creates conditions for a consumer to fall in love with an aesthetic.

Second, given a formal structure to beauty and style, one can actually change the conditions for beauty, so long as she hews to the frameworks that motivate people. Recommendation and matching engines are not new in Internet history: digital advertising and e-commerce businesses have used behavioral psychology to design experiences that will optimize access to the consumer wallet. Pandora (and now Spotify, Apple, and other streaming radio services) can pick something I will probably like well, at this point. But these achievements only goes so far as curation. I wonder: can a software program design an aesthetic from scratch? 2016 was a landmark year for Artificial Intelligence interfaces taking market share in the consumer landscape, from customer service to scheduling, from health coaching to search. I have yet to see an AI-designed aesthetic that has succeeded, but the above suggests that the conditions for it are within reach.

It seems inevitable to me, then, that we will soon be able to develop technology can automatically design a sound, an image, or even a space, to appeal to the subjective mind. I wonder, as that time approaches, whether there is a greater premium on actually being truly unique in brand and design, rather than simply derivative, rather than anchoring in the familiar. My colleague Morgan wrote a thoughtful meditation about how odd truly transformative inventions look at first.

Timing and Infrastructure

Leave a comment

Over the holiday I was catching up on the unread articles in my browser (time to switch to Pocket/Instapaper… there were almost 100) and I bumped into this one by Bloomberg collecting venture predictions for the most important trend of 2016.

I really liked Rebecca’s comment: “2016 was the year the internet quietly sped up″ and haven’t been able to get it out of my head. In the venture community, we often credit Amazon Web Services – and the rise of cloud computing more generally – as an inflection point in the startup industry, as a founder could build a technology company for the cost of a subscription to EC2, instead of having to buy and maintain their own servers and manually include CPU, memory, PCI components, et cetera. This is the difference between $100,000+ for your own hardware and $1000+ for access to a subscription. The floodgates burst open with startup activity, and the world was never the same. Crazy as it sounds now, perhaps it will have been Jeff Bezos’ most valuable contribution to the technology community during this period.

When the 200 millionth 3G handheld device shipped in 2007, days before Steve Jobs went on stage to announce the iPhone, another inflection point occurred: the birth of mobile as we know it today. Most people ascribe all the value to the iPhone 1, incorrectly. Yes, touch screens and the concept of the app were transformative, but it was actually the iPhone 2 that I find most interesting, and that actually represents the true inflection point: the iPhone 3G. In that moment, connection speeds were fast enough that ‘real-time’ was within reach for applications, GPS and turn-by-turn navigation arrived (at scale) on a handheld device. And it was only in this context that UberCab made sense. And only in the release, the next year, of the iPhone 3GS, where the camera finally was usable, where Instagram made sense.

LTE, whose adoption reached interesting scale some time in late 2009/early 2010, took the evolving speed characteristics to the next step, whereby a piece of media wouldn’t have to buffer for long periods to load, where the expectation was instantaneous and continuous media. It’s only in this environment – a matter of telecommunications standards having been adopted widely enough, where true shopping and banking on the phone were not unreasonable, where constant multimedia communication made unit economic sense – where Spotify and Snapchat were possible.

As 4G adoption takes hold, and live video streaming and realtime video chat explode on Facebook, Snapchat, and the new players Marco Polo, House Party, Tribe continue their blistering growth, we have our resilient and fast-improving infrastructure to thank for the rewiring of our user experiences with these devices.

5G (the G simply stands for generation, FYI) is around the corner, and upload and download speeds are north of 50 Mbps for many connections today, such that the “always-on” connection that we have come to expect from Wifi will be possible with standard mobile devices as well. In this sense, it may be that the Internet of Things was indeed a bit early, but not simply because the killer use cases hadn’t arrived, as many have speculated, but actually because the infrastructure wasn’t ready yet. I also believe the internet of Things is a massively correct trend, but we’re not thinking small enough or big enough – stay tuned for a post on the latter, but as a teaser, consider the notion of data centers that are themselves mobile (read: self-driving cars/trains, UAVs). I’m imagining a dynamically updating, fully mobile cloud that moves data packets between the physically closest points, across a mesh, or *incredibly* cheaply, as the servers themselves move humans and cargo. Imagine what kind of application layer that technology will enable. Distributed computing on crack!

I’m endlessly excited about companies that are accelerating users up the Internet access curve, and about companies that are steepening the internet access curve itself. While we quarrel over social psychology and culture as the driving forces of adoption, the infrastructure layer may be the most important precondition for timing an information technology movement.