Thursday, June 22, 2017

The Services Revolution: Why Social Networks Turned Into an Instituion

Last month I gave a talk (pdf) on innovation timing at OpenWay Club. The presentation covered, among other topics, the unfolding technology revolution in services. The talk drew on several key sources, including the work of Oliver E. Williamson, a Nobel Prize winner in economics from UC Berkeley, Cesar Hidalgo's book "Why Information Grows", and our book with Max Shtein "Scalable Innovation."

My goal was to show that new technologies have fundamentally changed the nature of services because they commoditized "specificity" and "recurrence." (see figures below). That is, in a networked digital world knowing your customers and interacting with them on a regular basis is dramatically less expensive than in a "stand alone brick-and-mortar" world. To illustrate the main points, here's a screen shot of a relevant page from Hidalgo's book (with my annotations) and several slides from the talk.

(The recent purchase of Whole Foods by Amazon is another example of the shift to Groceries-As-Service model, where Amazon leverages its customer insights into recurring retail sales.)

Even more importantly, the new service models have become a major global institution because they addressed the fundamental issue that plagued service businesses since ancient times. Douglas C North (Nobel Prize in Economics, 1993), described the problem in game theory terms:
In the world of personal exchange (recurring-specific - ES), it pays for parties to an exchange to cooperate, because the parties have personal knowledge of the other players and there is the possibility for repeat dealings between the parties. But in a world of impersonal exchange, it pays for the parties to defect, ceteris paribus. With impersonal exchange, the world is one in which there is not an iterated game.... One does not know anything about the other players, and indeed there are a large number of players.
That is, in traditional transactions players on both sides have incentives to cheat because they don't know each other personally or through a personal network. Therefore, in 1999 North suggested that to make the global marketplace efficient and scalable a new model had to be invented:
...we are going to have to devise institutions de novo that attempt to confront and deal with worlds of impersonal exchange.
Remarkably, new service models, such as Airbnb, Uber, Amazon, Alibaba, Instaply and others provide a glimpse of the institutions to come. Since identities of sellers, buyers and recommenders are known, parties are less likely to cheat; therefore, the number and quality of transactions shows rapid growth.  Although the solution is not perfect, it is a lot more efficient than all attempts to introduce global regulations. It's exciting to see how social networking technologies are redefining the rules of commerce and provide a working alternative to law.

Monday, March 20, 2017

Lunch Talk: Jay Kaplan: Crowdsourcing Cybersecurity (at Stanford)

Entrepreneur Jay Kaplan, co-founder and CEO of Synack, describes how the idea of creating a cybersecurity service for enterprise businesses by crowdsourcing hackers went from sounding like a long shot to launching as a venture capital-backed startup. Kaplan, previously a senior analyst at the National Security Administration, talks about the virtues of government work and the nuances of “white hat” hacking.

Direct link to Youtube.

tags:network, security, enterprise, control

Friday, March 10, 2017

LunchTalk: Alan Burdick: "Why Time Flies" (at Google)

Alan Burdick is a staff writer and former senior editor at The New Yorker. His most recent book, "Why Time Flies: A Mostly Scientific Investigation," was published in January by Simon & Schuster.

“In his lucid, thoughtful, and beautifully written inquiry about time — what is it, really? Did we invent it, or does it invent us? - Burdick offers nothing less than a new way of reconsidering what it means to be human.” (Hanya Yanagihara, author of A Little Life and The People in the Trees)

Saturday, March 04, 2017

Creative Solution of the Day: the Publication Dilemma

Typically, inventors face a disclosure dilemma: on the one hand, you want to explain your idea to a potential investor or a customer; on the other hand, you don't want to explain it because the idea can be easily stolen. Researchers face a similar dilemma when they consider publishing their results that  might have valuable commercial implications.

In the 1840s, Samuel Colt used the US patent system to overcome the dilemma:
When Samuel Colt, of revolver fame, was trying to sell the U.S. government a system of naval mines, he had to establish that his device was original without giving away its secret. His imaginative solution was to submit the plan to the Patent Office, obtain a confirmation of its originality, and then withdraw the application before the patent was granted, thereby avoiding the publication of the patent specifications.*
The Colt's approach exemplifies a powerful problem-solving technique often called "Separation in Time." According to the principle:
- you perform the useful action first — in the Colt's case: explaining the invention via a patent application — at the time when your potential customer needs to be convinced;
- then you perform a reverse action — withdraw the patent application — at a different time, so that the competition doesn't learn about the idea.

Snapchat provides the most recent example of a successful application of the "Separation in Time" principle along the lines of Samuel Colt's solution. That is, a Snapchat picture or a post is published for a short period of time to a limited group of subscribers; then, the post disappears, so that the information doesn't leak out to the general public. Clearly, the technique can be used for a broad variety of "limited offers."

* Source: Alex Roland, "Secrecy, Technology, and War: Greek Fire and the Defense of Byzantium, 678-1204." 1992.

tags: dilemma, problem, solution, social, separation

Tuesday, February 28, 2017

Lunchtalk: TED - Why humans run the world

A TED talk by Yuval Harari, the author of The Sapiens. From the talk description:
Seventy thousand years ago, our human ancestors were insignificant animals, just minding their own business in a corner of Africa with all the other animals. But now, few would disagree that humans dominate planet Earth; we've spread to every continent, and our actions determine the fate of other animals (and possibly Earth itself). How did we get from there to here? Historian Yuval Noah Harari suggests a surprising reason for the rise of humanity.

Tuesday, February 21, 2017

Stanford CSP, BUS 152 - Session 5, Quiz 1

Background: A major shift in business and technology strategy (aka pivot) seems to be inevitable during the life time of both startups and large companies. The change requires the team to make critical decisions under conditions of uncertainty.

Please listen to the podcast above and answer the following Questions:

1. Why LoudCloud was too early to the market? What were the key decisions for LoudCloud/Opsware in executing a pivot?

2. Did Lytro develop a new technology? Please describe briefly its pivot in terms of the 4Q diagram covered during our Session 4 (Feb 13, 2017). Using the same terms, describe the team's original mistake.

3. (Optional). Imagine that you are the CEO of Twitter. Your user base and revenues are not growing fast enough to compete with Facebook and Snapchat. You have $1B and 2 years to execute a pivot. Describe your key decisions and reasoning behind them.

tags: bus152, quiz, 4q diagram,

Thursday, February 09, 2017

Stanford CSP, BUS 152 - Innovation Timing. Session 4, Quiz 1


Recently, a number of technology companies introduced Augmented Reality headsets that enable users to overlay images from the real world with information generated by a computer, including text, video, graphics, etc. Microsoft HoloLens  is one of the most advanced projects in the field, backed by a major industry player. For example, in a 2016 TED Talk, Alex Kipman demonstrated a head-mounted 3D hologram computer that lets users interact with "magical objects" directly, which eliminates, among other things, the need for displays, keyboards, joysticks, and other typical computer accessories.

Please read the wikipedia article referenced above, watch the TED video, and answer the following questions:

1. Is HoloLens a new technology? Explain.

2. Imagine that it is year 2040. Do you see HoloLens-like devices dominating specific fields of human-computer interaction, e.g.

a) Education (certainly, highly likely, maybe, absolutely not). Explain.
b) Gaming (certainly, highly likely, maybe, absolutely not). Explain.
c) Construction and Industrial Services (certainly, highly likely, maybe, absolutely not). Explain.
d) Healthcare (certainly, highly likely, maybe, absolutely not). Explain.

3. (Optional) In your opinion, what are the top three bottlenecks that may prevent Augmented Reality devices, like HoloLens, from becoming widely popular among consumers during the next 3-5 years? Explain.

Friday, February 03, 2017

Lunch Talk: Superintelligence

A panel discussion with leading AI experts and business leaders about the challenges and opportunities presented by Superintelligence.

Panelists: Bart Selman (Cornell), David Chalmers (NYU), Elon Musk (Tesla, SpaceX), Jaan Tallinn (CSER/FLI), Nick Bostrom (FHI), Ray Kurzweil (Google), Stuart Russell (Berkeley), Sam Harris, Demis Hassabis (DeepMind).

00:00. Yes, No, It’s complicated
03:10. Timescale (Elon at 5:45)
07:07. How to slow it down
14:04. Risks and mitigations (Elon at 32:14)
37:00. Upsides (Elon at 51:18)
52:44. Democracy 2.0
54:14. Bad guys
56:43. Democratising AI (Elon)

lunchtalk, intelligence, problem, system,

Wednesday, February 01, 2017

Stanford CSP, BUS 152 - Innovation Timing. Session 3, Quiz 1

Background: Over the last decade, AI-based technologies succeeded in solving various problems that before were considered impossible to solve using computational methods. In one recent example, Stanford researchers "have trained an algorithm to diagnose skin cancer." In another example, AI bot easily outplayed humans in poker.

The significance of the latter development is that the algorithm successfully handled a problem with imperfect information:
Poker requires reasoning and intelligence that has proven difficult for machines to imitate. It is fundamentally different from checkers, chess, or Go, because an opponent’s hand remains hidden from view during play. In games of “imperfect information,” it is enormously complicated to figure out the ideal strategy given every possible approach your opponent may be taking.

Given that innovation fundamentally involves decision-making with imperfect information, we may want to consider how AI will impact broader innovation processes in our society.

1. Assume that AI decision-making services are widely available. In your opinion, which segments of the society will start using such services first: consumer or enterprise? Explain your reasoning and give approximate calendar time estimates for each segment.

2. Consider Kahneman's System 1 vs System 2 approach to human decision-making (e.g. as discussed during our Session 2). Will a wide adoption of AI services improve or worsen people's ability to use "System 2 thinking"? Explain.

3. In your opinion, will AI-based decision-making services affect the overall timing of innovation diffusion in social systems (see Session 1 lecture notes), e.g. by making S-curves more gradual, more steep, or leave them unchanged? Explain.

Tuesday, January 24, 2017

Stanford CSP. BUS 152 - Innovation Timing. Session 2, Quiz 1


The public's interest in Bitcoin rose sharply in 2013-14.

For example, on January 21, 2014, in a NYT article titled "Why Bitcoin Matters", Marc Andressen wrote:
Bitcoin gives us, for the first time, a way for one Internet user to transfer a unique piece of digital property to another Internet user, such that the transfer is guaranteed to be safe and secure, everyone knows that the transfer has taken place, and nobody can challenge the legitimacy of the transfer. The consequences of this breakthrough are hard to overstate.
Despite its great promise, this major breakthrough has not materialized yet. Nevertheless, the Google Trends chart above shows a noticeable uptick in Bitcoin-related interest in 2017. For example. a recent post on CloudTech by James Bourne titled "Blockchain beyond Bitcoin: Assessing the enterprise use cases" states that the technology "has serious potential to disrupt a multitude of industries." Also, Cade Metz in a January 6, 2017, Wired article titled "Bitcoin Will Never Be a Currency—It’s Something Way Weirder" reports on the general sentiment about Bitcoin, "Bitcoin is not something the average person will ever use to buy and sell stuff... It’s not something that will improve what the world has, such as money or stock. It’s something that will give the world stuff it has never had."

1. In your opinion, does Bitcoin follow the process generally described as Hype Cycle? Explain briefly.
1a. If yes, what is the current stage of the technology relative to the cycle?
1b. If no, how do you explain the 2014 peak and the significant investments VC funds put into Bitcoin-related startups?

2. Is the 2017 Bitcoin (and related technology) comeback for real? Where would you place the technology on the S-curve and Adopter Distribution chart as of today? Explain briefly.

Thursday, January 19, 2017

Stanford CSP. Business 152. Innovation Timing. Session 1, Quiz 3

Background: Sequoia Capital, one of the leading Silicon Valley venture capital firms, typically asks its prospective portfolio startups "Why Now?"

1. List at least 10 major innovations that are either happening now or about to happen within the next 3-5 years.
2. Assume that you are going to participate in one or two of those innovations.
3. Pick your role, e.g. startup founder, employee, corporate CEO/CTO, investor, scientist, student, journalist, president, non-profit, etc.
4. Given your role, select two innovation opportunities that you want to start working on now.
5. Explain "Why now?"

tags: stanford, quiz, innovation

Sunday, January 08, 2017

The Structure of Technology Revolutions

Since last summer, I've been working on a book project tentatively (and modestly!) titled "The Structure of Technology Revolutions." The purpose of the book is to show how technology enables completely new possibilities, by breaking trade-offs that are considered unbreakable.

To demonstrate the underlying structure of the innovation process, I'm using Category Theory tools (OLOGs) originally created by D.I. Spivak from MIT.

Here's a series of draft figures with an example of how the logic of innovation had worked in the technology revolution initiated by the automobile with the internal combustion engine (see below).

 Note, that the same logic can be applied to the modern autonomous vehicle. The technology is going to be successful because it creates incredible maneuverability at the "traffic" level of abstraction.

Now, back to the horses example:

Fig. 1 introduces the trade-off between Power and Maneuverability. An eight-horse carriage has a lot of power, but it's difficult to maneuver. Adding more horses will create a huge maneuverability problem. On the other hand, a horse rider is highly maneuverable but he lacks the carrying capacity of the horse carriage.

Fig. 2 introduces a logical representation of a horse carriage and maps it onto a "Conflicting Desires Diagram." That is, we show that any "designer" of a horse carriage faces a trade-off between Power and Maneuverability.

Fig. 3 sheds horse pictures and shows a logical generalization: a horse carriage is a kind of power-driven vehicle. 

Fig. 4 indicates the desired situation (the green dot on the right): We want a vehicle that has the best of both worlds, it's highly powerful and highly maneuverable.

Fig. 5 shows that the Automobile breaks the trade-off and creates a vehicle with the potential to hit the green dot. That is, we create a technology that disentangles human ability to control horses from the power. Thus, we achieve a new state that was considered impossible before.

To model the Autonomous Vehicle technology revolution we need to abstract from "a vehicle" to "traffic" and show how the new technology breaks the traffic congestion trade-off. In general, congestion trade-offs are ubiquitous in economic systems and technology revolutions break through them quite often.

Fig. 6 is a generalized diagram of how technological innovations make the impossible possible.

tags: innovation, trade-off, logic, technology, revolution

Saturday, November 19, 2016

Stanford CSP. Business 152. Innovation Timing. Session 1, Quiz 2.

On November 14, 2016, the New York Times wrote a story about a Spark Capital, a 11-year old technology venture firm.

The article opened with a description of one of Spark's recent deals:
Betting on an automated driving start-up in 2015 may not have been the most intuitive gamble at a time when Google and Uber had already declared that self-driving vehicles were among their top research priorities. 
But in the fall of 2015, Spark Capital was one of a few established venture capital firms to wade into the industry, helping lead a $12.5 million investment in Cruise Automation, a start-up based in San Francisco whose software helps cars pilot themselves. One of Spark’s partners became the only outside board member of the firm. 
It was a bet that paid off quickly: Within six months, Cruise sold itself to General Motors for about $1 billion.

1. In your opinion, why the timing of the deal turned out to be so good? Was it pure luck? If you were the analyst who "discovered" Cruise Automation back in 2015, how would you justify the $12.5M investment to your VC partners?
2. (optional) Consider the generic innovation diffusion S-curve, as described in Everett Rogers' "Diffusion of Innovations." What stage of the curve the self-driving car technology is now? Why?

tags: course, stanford, innovation, s-curve

Thursday, November 17, 2016

From junk food to junk news

Remarkably, surrounded by an abundance of choices people chose what feels good, not what is good. With junk food, it's a combination of fat, sugar and salt that fools taste buds into craving for more. With junk news, it's the confirmation bias that fools brains into craving for more news that conform to their world view.

According to Buzzfeed, during the 2016 election cycle fake news outperformed real news.

Due the difference in feedback mechanisms, the situation with junk news is worse than with junk food. That is, after having a junk food diet for an extended period of time, people can at least use scales to discover that their weight has gone up. By contrast, after having a junk news brain diet, people can only get stronger in their opinions because their social network keeps rewarding them for consuming and sharing the junk.

Can we solve the problem without resorting to censorship? One way to look at it would be to consider the situation from a point of view widely adopted in another domain - money and finance. That is, today's fake money is easily detected and discarded, so that the society doesn't fall into the trap of the Gresham's Law. Similarly, fake news can be detected by a variety of technologies, including a BitCoin-like approach that verifies authenticity of the news and news sources. Fake news, like fake coins should be taken out of circulation. Otherwise, our brains get stupid by consuming junk news, just like our bodies can get fat, by consuming junk food.

Lunch Talk: Counterintuitive approach to building startups (Stanford University)

This is Lecture 3 from a Stanford University course "How to start a startup". The speaker is Paul Graham; his transcript is here:

tags: startup, stanford, entrepreneurship, innovation, lunchtalk,

Wednesday, November 16, 2016

Stanford CSP. Business 152. Innovation Timing. Session 1, Quiz 1.


Timing is critical for innovation success. Sometimes, companies introduce new products and services when it’s too late. For example, neither Google+ nor Microsoft smartphone succeeded, despite their respective companies putting major resources behind them, both in money and development efforts.

On the other extreme, some innovations fail because they seem to appear too early. Social networks Friendster and Livejournal started gaining traction in the early 2000s, but never got to the scale of Facebook. Similarly, WebTV started in 1995, with the intent to provide users with a broad range of content over the Internet. In almost two decades that followed, the company burned through hundreds of millions of dollars, went through a major acquisition, but failed in the very same marketplace where NetFlix and other video streaming services managed to succeed a few years later.

Finally, certain products and services had perfect timing. For example, Gmail and Youtube spread like a California wildfire. The Apple iPhone succeeded where Apple Newton failed.

In preparation for the course, please answer the following questions:

1. List 2-3 novel products or services in each of the timing categories:
a) too late;
b) too early;
c) just perfect.

2. (Optional). Pick one example from the list and explain your reasoning with regard to innovation timing. Mention at least 3 factors that played a role in the success or failure of the innovation.

Wednesday, July 20, 2016

Stanford CSP BUS 74 [Principles of Invention and Innovation], Session 4 Quiz 1

On July 19, 2016, Bloomberg Technology News reported that Google used its DeepMind AI technology to reduce power consumption in the company's data centers:
In recent months, the Alphabet Inc. unit put a DeepMind AI system to reduce power consumption by manipulating computer servers and related equipment like cooling systems. It uses a similar technique to DeepMind software that taught itself to play Atari video games, Hassabis said in an interview at a recent AI conference in New York.

The system cut power usage in the data centers by several percentage points, "which is a huge saving in terms of cost but, also, great for the environment," he said.

The savings translate into a 15 percent improvement in power usage efficiency, or PUE, Google said in a statement. PUE measures how much electricity Google uses for its computers, versus the supporting infrastructure like cooling systems.

Question 1. Using the system model, name the functional element that DeepMind technology helps to improve directly.
Question 2. Based on what you know about the improved element, describe other functional elements within the same system.

A new way to map brains

Neuroscientists at Washington University Medical School created a method to build maps for individual brains:

(MIT Tech Review) Researcher Matthew Glasser says that unlike many previous studies, this map considers several features of the brain simultaneously to mark its boundaries. Some neuroscientists still define brain regions based on a historical map called Brodmann’s areas that was published in 1909. That map divided each half of the brain into 52 regions. Each hemisphere on the new map has 180 regions.

Glasser defined these regions by looking for places where multiple traits—such as the thickness of the cortex, its function, or its connectivity to other regions—were changing together. After drawing the map onto one set of brains, the researchers developed an algorithm to recognize the regions in a new set of brains where the size and boundaries vary from person to person. “It’s not just a map that people can make reference to,” Glasser says. “You can actually find the areas in the individuals that somebody is studying.”

From an innovation perspective, mapping methods create opportunities to systematically explore and coordinate knowledge about a broad class of objects. This particular approach enables scientists and engineers to move back and forth from generalized information about human brain to specific aspects in a particular brain. For example, we might be able to understand why 3D VR can replace painkillers in some medical applications.

Wednesday, July 13, 2016

Stanford CSP, BUS 74. Session 3, Quiz 1.

In a recent NYT article titled "Ads Evolve Into New Forms as Media Landscape Shifts", Sydney Ember mentioned an emerging trend in the advertisement industry:
Consumption habits have become increasingly fragmented, with more people watching programming, including television shows and live sports, on different online platforms. As a result, traditional television, with its 30-second commercials, is losing its commanding share of advertising dollars. Digital media is expected to pass TV as the biggest advertising category in the United States this year, with roughly $68 billion in ad sales compared with $66 billion for TV, according to the Interpublic Group’s Magna Global.

With online ad spending growing, finding ways to stand out among the onslaught of other online ads has become more important for advertisers. And therein lies a possible conundrum: Advertisers want their ads to look less like ads even as they are fighting harder for attention.

Question 1.
Based on our brief class discussion (see slide 33 Lecture Notes from July 11, 2016) and an earlier post on this blog, use the 10X Change diagram to map ad-related business models mentioned in class. Briefly explain parameters for each model.
(a ppt version of the 10X Change diagram is available for download here).

Question 2 (bonus).
What major technology developments enabled key ("disruptive") business model transitions?

Question 3 (bonus). Use the 10X Change diagram to map potential ad-related business models that are now available with augmented reality games like Nintendo Go. What technologies (existing or new) can further improve such models?

Thursday, June 30, 2016

Lunch Talk: Moral Tribes (Joshua Greene gives a talk at Google)

Note how his experiments show the relationship b/w physical distance and psychological distance. A similar effect happens when inventors are trying to explain their ideas to investors. I also like his analogy between Kahneman's System 1 vs System 2 on one the hand, and point-and-shoot and SLR cameras on the other: the former is set on automatic, while the latter on manual.

Tuesday, June 28, 2016

Stanford CSP 74 Principles of Invention and Innovation (BUS 74). Session 2 Quiz 1

In a recent MIT Technology Review article, Antonio Regaldo describes a new genetic engineering approach that promises to eliminate malaria:
Malaria kills half a million people each year, mostly children in tropical Africa. The price tag for eradicating the disease is estimated at more than $100 billion over 15 years. To do it, you’d need bed nets for everyone, tens of thousands of crates of antimalaria drugs, and millions of gallons of insecticides.
A gene drive is an artificial “selfish” gene capable of forcing itself into 99 percent of an organism’s offspring instead of the usual half. And because this particular gene causes female mosquitoes to become sterile, within about 11 generations—or in about one year—its spread would doom any population of mosquitoes. If released into the field, the technology could bring about the extinction of malaria mosquitoes and, possibly, cease transmission of the disease.

Question 1: Using the "Divergeng-Exploratory-Convergent" thinking technique,
a) list lots of benefits and problems that the new approach creates;
b) create an explicit criteria for selecting top benefits and problems;
b) according to your criteria, what are the most important short- and long-term benefits/problems (at least one each)?

Question 2 (optional): What dilemma did the researchers solve, while trying to create their genetically modified mosquito?

Question 3 (optional): What's the difference between system levels that the existing and the new malaria solutions target?

Thursday, June 23, 2016

Stanford CSP 74 Principles of Invention and Innovation (BUS 74). Session 1 Quiz 1

Research shows that online privacy remains a controversial topic. For example, a review article from the Science Magazine states*:

If this is the age of information, then privacy is the issue of our times. Activities that were once private or shared with the few now leave trails of data that expose our interests, traits, beliefs, and intentions.
Both firms and individuals can benefit from the sharing of once hidden data and from the application of increasingly sophisticated analytics to larger and more interconnected databases (3). So too can society as a whole—for instance, when electronic medical records are combined to observe novel drug interactions (4). On the other hand, the potential for personal data to be abused—for economic and social discrimination, hidden influence and manipulation, coercion, or censorship—is alarming. The erosion of privacy can threaten our autonomy, not merely as consumers but as citizens (5). Sharing more personal data does not necessarily always translate into more progress, efficiency, or equality (6).

Question: How would an IDEAL privacy system would change the situation.

*Science 30 Jan 2015:
Vol. 347, Issue 6221, pp. 509-514
DOI: 10.1126/science.aaa1465

Direct link to the article (pdf) on