Self Help

Machine, Platform, Crowd Harnessing Our Digital Future - Andrew McAfee & Erik Brynjolfsson

Author Photo

Matheus Puppe

· 66 min read

“If you liked the book, you can purchase it using the links in the description below. By buying through these links, you contribute to the blog without paying any extra, as we receive a small commission. This helps us bring more quality content to you!”



  • Andy thanks the McAfee family (David, Shannon, Amelia, Aurora, and Avery Mae) for letting him keep some of his money at the poker table sometimes.

  • Erik thanks his mother, Marguerite, whose love, smiles, and faith keep him going.

  • The essay summarized covers several chapters about the rise of artificial intelligence and machine learning. It discusses how Google’s AlphaGo system was able to master the game of Go by using deep neural networks and tree search to learn strategies from analyzing millions of past games, rather than having to explicitly program human expertise. This was a breakthrough as past Go programs struggled due to our inability to fully articulate our own tacit knowledge, known as Polanyi’s Paradox. AlphaGo achieved strong victories over both the European Go champion and Lee Sedol, the top-ranked human player in the world.

  • Goodwin described companies like Uber, Airbnb, Alibaba, and Facebook as an “indescribably thin layer” since they own minimal physical assets and infrastructure compared to traditional companies in their industries.

  • However, these companies were still able to experience rapid global growth and high valuations. For example, within a year of Goodwin’s article, Airbnb doubled the number of nights booked and faced challenges from cities worried about neighborhood impacts.

  • Uber and Facebook also continued expanding their services and popularity. Facebook in particular saw more traffic from its platform than Google searches.

  • Even large, established companies like General Electric saw opportunities in partnering with new tech companies. GE, known for innovation, product development and marketing capabilities, opted to crowdsource ideas for a new consumer product from strangers online through a site called FirstBuild.

  • FirstBuild was a “co-creation community” GE had launched with the University of Louisville to change how products reach the market by enabling online collaboration and rapid prototyping.

  • An entrepreneur at GE Appliances wanted to develop an easier way for people to access “nugget ice” - small, porous ice cubes that are easier to chew on than normal ice cubes.

  • Making nugget ice machines is complex and existing machines cost thousands, too expensive for most homes.

  • He used the FirstBuild platform to launch an online competition to design a cheaper home nugget ice maker. The winner was Ismael Ramos, who designed a cubical machine called “Stone Cold.”

  • GE worked with FirstBuild to prototype Ramos’ design. They engaged the online community for feedback on aspects like the ice bucket design.

  • GE then launched an Indiegogo crowdfunding campaign for the ice maker, now called the Opal Ice. Within a week it raised over $1.3 million, showing strong demand.

  • The campaign ended up raising over $2.7 million, becoming one of Indiegogo’s most popular campaigns. GE used it to test demand and get market intelligence, rather than needing the funds.

  • Over 5,000 preorders were shipped in late 2016 before a general release, showing success in using the crowd to develop and prove demand for a new product.

Here is a summary of the key points regarding technological advances and sound economic principles from the passage:

  • Technological changes like the shift from steam power to electric power in manufacturing factories in the early 20th century were highly disruptive to existing companies. While new technologies provided advantages like increased efficiency, their full potential was often not realized or anticipated by incumbents.

  • Factors like the “curse of knowledge”, where expertise in existing technologies blinded companies to new possibilities, and “status quo bias” made it difficult for even successful incumbents to adapt quickly to changes. Organizational and conceptual shifts were needed to fully unlock new technologies’ value.

  • During the electrification period, over 40% of large industrial trusts formed in the late 19th/early 20th centuries failed by the 1930s. While new technologies may provide efficiencies, sound economic principles like adaptability and anticipation of disruptions are vital for long-term success during periods of significant technological change. Flexible approaches to tasks, products, and business models are important to capitalize on innovations’ full potential value.

  • A study found that the average market share of dominant manufacturing firms declined significantly from 1905 to 1929, from 69% to 45%. This suggests the competitive environment in the US became more challenging in the 20th century.

  • Electrification is believed to have contributed to this change. Intelligently electrifying factories, by adding electric motors, assembly lines, cranes, etc., made production much more productive and efficient. This allowed companies to undercut competitors on price, flexibility, and production output.

  • Not all companies effectively electrified their factories. Some were slow to adopt new technologies like unit drive. As a result, early-adopters gained competitive advantages that contributed to the decline of older industrial trusts.

  • While multiple factors caused changes, the disruptions of electrification were a major reason why many top companies failed or struggled. Those that viewed electrification only as a new power source and did not modernize their factories ultimately became uncompetitive and went out of business.

  • Today another major technological transition is underway, driven by machines, platforms and crowdsourcing. Again, companies that do not understand and effectively leverage these new technologies risk becoming obsolete, just as those that failed to intelligently electrify in the early 20th century.

Here is a summary of the key points about explicit in the passage:

  • The passage discusses different modes of thinking - System 1 and System 2. System 1 thinking is fast, automatic, intuitive thinking while System 2 thinking is slow, conscious, effortful thinking.

  • Daniel Kahneman, who won the Nobel Prize, pioneered research showing that people have these two different thinking systems.

  • Both systems can be improved over time. System 2 is improved through courses, logic, math, etc. System 1 is improved more naturally through experience and exposure to examples.

  • Learning is often deeper when we understand principles (System 2) and see many examples to instantiate those principles (System 1).

  • Business school aims to sharpen both systems - System 2 through accounting, finance courses, and System 1 through case study discussions to build intuition.

So in summary, the passage explicitly discusses Kahneman’s dual process theory of thinking in which people have an intuitive, automatic System 1 and a deliberate, effortful System 2 mode of reasoning. It provides examples of how both systems can be developed.

  • Traditional models propose a “standard partnership” between human intuition/judgment and machines, with humans focusing on complex decision making and machines handling calculations.

  • However, research has shown human judgment and intuition are often poor predictors and more prone to biases compared to data-driven models.

  • Studies in fields like purchasing, wine quality predictions, real estate, academia (tenure decisions), criminal justice, education found statistical models outperformed human experts.

  • Meta-analyses of over 100 prediction comparisons in psychology and medicine found humans did worse than statistical models in 46% of cases and no different in 48% - only 6% did humans clearly outperform models.

  • While human intuition still has a role, companies are increasingly relying less on expert judgments and predictions and more on data-driven approaches due to their superior performance. The conclusion is we need to rely less on expert judgments alone.

  • The passage discusses the increasing adoption of data-driven decision making and its performance benefits as observed in a large survey of manufacturing plants. While algorithmic models have achieved success, they have limitations as well.

  • It argues that when a valid model can be created and tested on similar data, it tends to perform as well or better than human experts for similar decisions. However, models are not always possible due to limitations in available data.

  • The text then discusses the strengths but also cognitive biases and limitations of human judgment. While human intuition is powerful, it is also “buggy” and prone to numerous cognitive biases that can lead to faulty decisions. Models based only on numeric data may avoid some of these biases.

  • In summary, the passage examines both the performance gains seen from increased use of data-driven decision making, but also acknowledges the difficulties and limitations in replacing human judgment with algorithms in all cases. A balanced partnership of humans and machines may provide the best outcomes.

  • Dan Wagner saw how the Obama 2012 reelection campaign used machine learning models and extensive voter data to precisely target voters, moving beyond traditional demographic-based advertising targeting.

  • The campaign built models to predict individual voters’ likelihood to support Obama, vote, and be persuaded. They then matched these lists against viewership data to identify the best TV shows for reaching key voter groups.

  • This allowed them to find unexpected but highly effective shows, like late night programming on TV Land, that delivered a lot of “persuadable voters” at low cost compared to relying only on demographics.

  • After the election, Wagner started Civis Analytics to commercialize this data-driven approach to media buying, highlighting how companies also have detailed customer data that can optimize advertising placement.

  • While powerful, data-driven systems also risk perpetuating societal biases through algorithms if not designed carefully. Researchers found examples of racial bias in some online ad delivery systems. Fully replacing human judgment with algorithms is not ideal either, as defining and optimizing for the right goals is complex.

  • The challenge is to minimize biases and errors in decision systems, while allowing for quick correction. The best approaches may involve a combination of machine and human intelligence.

The story proposes that in the future, companies may adopt a hybrid model of human and AI collaboration instead of humans solely making decisions. It uses an exaggerated example of a factory where one human feeds a dog and the dog prevents the human from touching machines as a mock scenario of how jobs could look.

However, the article argues that model is unrealistic. While computers don’t have biases, humans have strengths like broad sensory perception and common sense that AI currently lacks. Humans can consider unexpected contextual factors like a broken leg affecting plans.

It’s useful to have humans oversee computer decisions to catch errors or issues like Uber’s surge pricing during crises. However, companies should also track decision accuracy to provide feedback and improve algorithms over time.

Some companies are now inverting the model by having human judgment serve as input to algorithms, like Google uses structured interviews scored on rubrics rather than unguided opinions. This limits biases while quantifying judgments for fair, consistent decisions.

In summary, the future likely involves collaborative human-AI models rather than either making solo decisions. Humans provide strengths like broad context and oversight, while algorithms aim to reduce biases if properly optimized with performance tracking over time. An inverted arrangement with human input to algorithms may be most effective.

  • Computers making decisions pushes people to the margins and diminishes their decision-making role. While uncomfortable, relying solely on human judgment can lead to poor outcomes.

  • Examples given include releasing the wrong prisoners, higher medical misdiagnosis rates, and hiring the wrong job candidates, all to preserve human roles. The goal should be good decisions, not preserving human roles.

  • Humans are also poor at predicting the future due to cognitive biases. Research shows experts are only slightly better than chance at predictions.

  • Some “superforecasters” can consistently outperform others by considering multiple perspectives rather than a single fixed view. Relying on them and their track records is better than common experts.

  • Companies should experiment more and predict less given increasing complexity. Small iterative experiments, not long-term forecasts, better allow adapting to feedback and “inventing” the future.

  • Online A/B testing enables rapid, data-driven experimentation. Multi-location companies can also experiment by varying policies across locations.

  • In summary, human judgment is flawed and algorithms often outperform even experts. Decisions should increasingly rely on data and testing over predictions and subjective judgment where possible.

  • Decision-making processes should focus on producing the best outcomes based on clear goals and metrics, rather than making decision-makers feel good.

  • Algorithms have biases if based on flawed data, but the standard is whether they outperform alternatives on relevant metrics and can continue improving, not whether they are perfect.

  • As technology has advanced, there are more opportunities to move beyond relying solely on subjective human judgments (“HiPPOs”) and incorporate data-driven analyses into decisions. Companies doing this tend to outperform peers.

  • Individuals and companies that can consider issues from multiple perspectives and iterate/experiment are higher performers.

The questions posed ask about tracking performance of human and algorithmic judgments over time; where HiPPOs are most common and potential opportunities to incorporate more data; whether algorithms or humans are generally more biased; preferences for generalists vs specialists; and characteristics of projects typically undertaken.

  • In the 1950s, researchers like Frank Rosenblatt created one of the first machine learning systems called the Perceptron, which could classify simple images. However, it had limitations and research in neural networks declined after a critique in 1969.

  • A few researchers persisted with neural networks and machine learning approaches. Advances like backpropagation (allowing information to flow backward in a network) led to better performance in the 1980s-90s, but progress was still slow.

  • In recent years, there has been a huge increase in the capabilities of neural network and deep learning systems. Examples like AlphaGo, which beat the world’s best Go player using a neural network trained via self-play and human data, show how effective these approaches have become compared to traditional symbolic/rule-based AI. Neural networks can now perform many complex tasks like image recognition, translation, and others at human or superhuman levels.

So in summary, early machine learning systems showed promise but had limitations, research declined, but persistence by some led to advances enabling the recent powerful applications of neural networks and deep learning for artificial intelligence tasks.

  • Deep learning and neural networks are now the dominant form of artificial intelligence, fulfilling some of the early promises of AI.

  • Key factors that enabled this include Moore’s law making more powerful neural networks affordable, cloud computing opening up resources, and the rise of big data from various digital sources like text, images, videos.

  • Specific developments like Hinton’s work on deep belief nets and unsupervised learning were important breakthroughs.

  • Companies like Google, Microsoft and IBM are now deploying deep learning extensively across their products and services with significant benefits. Deepmind’s use of deep learning to optimize energy efficiency in data centers reduced costs by up to 40%.

  • The technology is spreading rapidly both within large companies and in unexpected areas like a Japanese family farm, where a son automated cucumber sorting using Google’s TensorFlow and achieving 70% accuracy. The widespread availability of tools and cloud resources is enabling more widespread application of deep learning.

  • Makoto is excited to try new machine learning software that can perform image analysis and other tasks in the cloud. Kaz Sato from Google says the potential uses of machine learning are only limited by our imagination.

  • So far, most commercial successes in AI have used supervised learning, and some reinforcement learning. However, humans learn mainly through unsupervised learning. Developing better unsupervised learning algorithms will be key to achieving artificial general intelligence.

  • Current AI systems combine human oversight with machine autonomy, like having humans monitor self-driving cars. Google is working on fully autonomous self-driving cars that require no human involvement.

  • Machine learning is being applied to automate back-office office work that has resisted full automation. Systems can learn rules from past work rather than relying on human experts to explicitly define the rules.

  • Progress in areas like speech recognition using deep learning has surprised experts by achieving human-level performance without linguistic rules.

  • There are many opportunities to replace human decision making with more accurate data-driven AI systems. As machine learning becomes more advanced, it will make businesses smarter and more productive. Overall, machine learning and neural networks have become the dominant approach in AI over traditional rule-based and symbolic systems.

  • The Eatsa restaurant introduced a novel ordering and pickup process without any person-to-person interactions. Customers ordered meals using tablets, then received their orders via numbered cubby holes after their names appeared on a display screen.

  • This illustrated how many transactions that used to require human interaction can now be completed virtually through digital interfaces. Many business processes do not require physical transformations and can instead involve just moving and transforming digital information.

  • Ordering and receiving a meal at Eatsa were examples of such virtualized processes that did not truly require automation but simply eliminated person-to-person interactions by having the customer perform the steps themselves via technology.

  • Virtualization of processes is spreading to other areas like air travel, where passengers can now download boarding passes, pass through automated security lanes, and use kiosks for customs upon returning to the US from abroad. This reduces the need to interact with airline or immigration employees.

The key idea is that Eatsa demonstrated how transactions between businesses and customers can be virtualized by moving them entirely online and eliminating in-person interactions, thanks to technologies like touchscreen ordering, automated order preparation, and smart cubby pickup systems. This virtualization of processes is becoming more common across different industries.

  • Virtualization is accelerating as digital technologies become more ubiquitous. Banking has greatly virtualized through ATMs, online banking, mobile apps, etc. reducing the need for in-person tellers.

  • Some argue certain transactions will remain largely unvirtualized due to technical issues with self-checkout kiosks. However, the technology is improving over time and could lead to frictionless experiences like Amazon Go stores with no checkout process.

  • A human touch may still be desired for some sensitive transactions, but companies like Wealthfront have still been able to attract customers without any human interaction.

  • Virtualization is being shaped by customer self-selection, but it may become a secular trend as younger and more tech-savvy generations account for more of the population over time. Customers will be attracted by efficiency and convenience even if some prefer human interaction initially.

  • Automation is also expanding to food preparation as robot technology improves in handling irregular objects through advanced vision and touch. Early automated restaurants and robotic chefs show the potential for highly automated food production in the future through a “Cambrian explosion” of robotics capabilities.

  • The “Cambrian Explosion” was a period of rapid evolution in life on Earth where most major animal phyla appeared.

  • Andrew Pratt argues we are on the cusp of a similar explosion in robotic innovation driven by recent advances in data, algorithms, networks, cloud computing, and exponentially improving digital hardware (the “DANCE” factors).

  • One important enabler of the Cambrian Explosion was the evolution of vision in biological species. Similarly, machines are now developing machine vision capabilities through technologies like computer vision.

  • The summary then outlines recent developments in each of the DANCE factors that are fueling rapid innovation in robotics, drones, autonomous vehicles and other digital machines.

  • These include exponential growth of data from sensors, advances in machine learning algorithms, faster wireless networks, vast cloud computing resources, and continued exponential improvements/cost reductions in digital components per Moore’s Law.

  • The interaction of all these factors through a cycle of experimentation, data collection, algorithm refinement, and distribution via networks and cloud is driving what Pratt calls a “Cambrian explosion” in robotics.

  • Roles that are well-suited for robots include tasks that are dull, dirty, dangerous and/or expensive for humans, like construction site monitoring, precision agriculture and other applications discussed.

  • Drones are being used increasingly in agriculture to monitor crop growth and health over time, enabling more precise application of water, fertilizer and pesticides based on drone imaging data. This supports the trend toward precision agriculture.

  • Drones are also being used by insurance companies to assess storm damage, and to monitor wildlife and forests against poaching and illegal logging.

  • Drones inspect offshore oil rigs and other installations, taking over dangerous and difficult manual inspection tasks.

  • Automated systems are also taking over tasks like milking cows, crop spraying, and mining, with driverless trucks able to operate 24/7 with remote supervision.

  • As machines take on physical and dangerous jobs, human roles are shifting toward knowledge work using data and analytics to optimize agricultural, industrial and other processes. People also continue working alongside machines by leveraging human abilities in sensing, problem-solving and dexterous tasks.

  • Emerging technologies like 3D printing and additive manufacturing are enabling new possibilities for making complex plastic and metal parts without the constraints of molds, opening up new design freedoms. This represents a major shift from traditional manufacturing approaches.

Here are the key points from the passage:

  • Many believe creativity is a uniquely human ability, as machines cannot come up with truly new ideas. However, demonstrations in industrial design show machines are becoming quite capable at creativity.

  • heat exchangers are devices that transfer heat between fluids while preventing direct contact between the fluids. They are complex to design well as they must transfer the right amount of energy efficiently, safely, durably and at low cost.

  • An AI system called Anthropic designed novel heat exchanger concepts that experts said were creative and promising. The designs had complex internal channels and configurations not seen before.

  • This shows machines can be creative by combining knowledge in new ways, not just through brute force computation. The AI was able to synthesize its knowledge into original solutions.

  • While human creativity remains profound and hard to capture fully, this example demonstrates machines developing genuinely novel ideas in a challenging design domain through a form of computational creativity.

So in summary, the passage uses the example of an AI designing new types of heat exchangers to argue that machines are proving capable of creative ideation, challenging the view that creativity is uniquely and immutable human. More work remains but computational creativity shows promise.

  • Generative design software can automatically design complex objects like heat exchangers without human input by exploring design possibilities computationally.

  • A research project used sensors on a test race car to collect performance data, which was then fed into generative design software to design an optimized chassis from scratch. The resulting design looked very organic and asymmetrical, resembling natural structures.

  • Without human biases, generative design software can explore more design possibilities than human designers and come up with novel solutions. One example was an AI system that proposed novel hypotheses for cancer research based on analyzing scientific papers, several of which were later supported.

  • Digital art created by programs like The Painting Fool and Emily Howell also shows potential creativity, though some argue the works currently lack the depth of human artists. With continued advances, AI systems may one day match or exceed human creativity in diverse fields like engineering, science, music and visual art.

  • Our sense of beauty and taste is complex and hard to define, but researchers have discovered some underlying patterns and principles, like the golden ratio, that help explain human preferences.

  • This knowledge is being applied through technology to generate new designs, recipes, architecture, and other creations without direct human input. Computer programs can now design buildings, cookbooks, and customized websites.

  • While computers are getting better at generating novel ideas, creativity still requires an understanding of human experience and psychology that computers lack. Creativity may involve solving open-ended problems without a single right answer.

  • Future design tools will let people quickly test concept feasibility before committing to detailed work. They will also automate more of the routine and tedious tasks, freeing up human designers for higher-level thinking.

  • Areas like coaching, teaching, counseling, and other jobs requiring social and emotional skills will remain challenging for AI, as they rely on connecting with people through empathy, bonding, and motivation. Computers are still far from being able to tap into human social drives in the same nuanced way as people.

So in summary, computers are increasingly aiding and augmenting human creativity, but fully autonomous machine creativity, especially in social domains, remains an open challenge that depends on a deeper understanding of human experience and psychology.

  • In the mid-1990s, prior to widespread internet adoption, several major industries like newspapers, magazines, radio, landline phones, and recorded music were thriving under the existing business models and technologies.

  • Mobile phones were an expensive novelty, with only 13% of Americans owning one that cost around $1,000. Landline phones connected via copper wires were near-universal.

  • Long-distance phone calls between more distant locations cost more, and households received separate bills for local and long-distance calls.

  • Newspapers generated $46 billion annually from subscriptions and ad sales, with classified ads and non-classified ads being major revenue sources. Communities were well served by local newspapers.

  • Radio stations numbered over 10,000 and generated $20 billion in revenues, playing a mutually beneficial role with the recorded music industry by promoting album sales.

  • Recorded music was a $14.3 billion growing industry, with existing business models seen as robust enough to support creative financing approaches like Bowie Bonds.

  • In summary, these industries were thriving prior to the emerging disruption from new Internet technologies and the decline of existing business models they would bring about within one generation.

  • In the mid-1990s, music was mainly consumed through physical albums purchased at stores like Tower Records or through mail-order album clubs. Major artists like Michael Jackson saw huge crowds line up for new album releases.

  • Shopping malls were a dominant retail space in the US, with over 1,500 new malls built between 1956-2005 as suburban living grew. Malls were a place to buy music, drop off film for developing, or pick up photos.

  • Kodak was a dominant film manufacturer with a high stock valuation in 1997, but digital photography was emerging with early digital cameras like the Casio QV-10 in 1995.

  • In the following decades, many industries like newspapers, magazines, music, malls, and landline phones declined dramatically due to digital disruption. Newspaper revenues dropped 70% and 13,400 jobs were cut. Magazine circulation and revenues also fell sharply. Music sales declined 45% globally from 1999-2014. Many stores and radio stations struggled.

  • The combination of digital goods being free to copy/distribute, perfectly reproducible, and instantly transferable over networks like the internet made it very difficult for traditional industries to compete against digital/online rivals.

  • Platforms are digital environments that have near-zero marginal costs of access, reproduction and distribution due to their abilities to be free, perfect and instant. Examples include the internet and World Wide Web.

  • Platforms can often be built on top of each other through combinatorial innovation. The World Wide Web built upon the internet protocols.

  • New platforms like Craigslist and online advertising platforms disrupted print media by offering free or low-cost alternatives for classified ads and ad placement. This reduced revenues for newspapers.

  • WhatsApp gained popularity through network effects as more users joined, making the platform more valuable. Even SMS users switched to WhatsApp to communicate with other users on the platform.

  • Amazon created the Amazon Web Services platform by standardizing interfaces between its internal systems and offering excess computing resources to other businesses, helping to launch the cloud computing industry.

  • AWS grew dramatically for Amazon, contributing 9% of revenue and over half of operating income by 2016. It was considered the fastest growing enterprise tech company due to its success.

  • The recorded music industry provides an example of disruption by digital platforms. Global music revenues fell over 50% from 2000-2015, even as consumers listened to as much or more music.

  • Earlier platforms like Napster and iTunes facilitated piracy and unbundling of music purchases from full albums to individual songs. This benefited consumers but hurt industry revenues.

  • Later, streaming platforms like Spotify rebundled music into subscriptions. Consumers paid flat monthly fees for on-demand, unlimited streaming access. This became very popular, with streaming making up 47% of US music revenues by mid-2016.

  • Platforms keep transforming industries by unbundling traditional packaged goods and services, then often rebundling them in new subscription models. This disruption upends traditional business practices and benefits consumers through new low-cost or free options, even as it hurts industry revenues. The economics of digital goods favors platforms and bigger networks.

Here are the key points from the passage:

  • Combinatorial innovation and digitization continue to reshape industries like computer hardware and recorded music.

  • Digitization allows for the “free, perfect, and instant” sharing of information (bits), which is reshaping business models.


  1. Where could your organization apply the economics of digital bits (free, perfect sharing of information) next? What opportunities exist to digitize offerings?

  2. What are the most important current and emerging digital platforms in your industry? How might they change in the next 3 years?

  3. How many of your current offerings could be delivered via cloud computing? Are you moving to the cloud quickly enough?

  4. Consider your customers’ needs and wants. What product or service bundles, unbundles, or rebundles could better meet their needs compared to the status quo?

  5. What are realistic scenarios for how network effects (the value increasing with more users) could strengthen or become more common in your industry?

The key takeaways are that digitization allows for new business models based on free and perfect sharing of information. Organizations should look for opportunities to apply these digital economics internally and in their offerings. They should also consider how digital platforms and cloud computing can transform their industry and customer offerings. Bundling, unbundling or rebundling products/services may better meet evolving customer needs. Network effects also promise to reshape industries so considering these dynamics is important.

  • Initially, Steve Jobs did not want to allow third-party apps on the iPhone, wanting to maintain control over the user experience. Others at Apple argued external developers should be allowed.

  • Allowing third-party apps turned out to be the right decision. It’s hard to imagine a successful smartphone today without a large variety of apps from independent developers.

  • A large set of apps alone doesn’t make a phone successful - the apps need to be of different types at a variety of price points to appeal to different consumers.

  • iPhone apps and the iPhone itself are “complements” - when the price of one goes down, demand for the other increases. Having a range of free and low-cost apps increases demand for the iPhone.

  • Different people want different “killer apps.” External developers were better able to create diverse apps like Shazam and Angry Birds to satisfy varied consumer preferences.

  • Free apps like Angry Birds generate “consumer surplus” by allowing users to enjoy the app without having to pay. This helps boost demand for the complementary iPhone.

  • When the price of apps like Angry Birds goes to zero, consumer surplus is maximized as everyone who is willing to pay anything for the game gets it for free. This entire area under the demand curve is consumer surplus.

  • Free, perfect, and instant complements like Angry Birds have two effects: 1) They generate consumer surplus for users. 2) They shift the demand curve outward for complementary products like iPhones, meaning more people are willing to pay the iPhone price.

  • While one app provides a small demand shift, hundreds of thousands of free apps provide huge consumer surplus and shift demand substantially outward for iPhones. This is beneficial to Apple.

  • There are various motivations for developers to create free apps, including freemium models, ad revenue, customer service/branding, and pairing with other products. This allows Apple to benefit from external development.

  • Opening the iOS platform increased innovation, growth, data for Apple, and new revenue opportunities from paid apps and the app store. However, platforms also need curation to avoid issues like malware, fakes news, etc.

  • Apple retained control over the app approval process to ensure quality while allowing external development, balancing openness and control. Platform owners have discretion over curating their platforms.

  • Powerful platforms like Apple’s App Store prompted others to build their own platforms for mobile devices and apps.

  • Google’s Android was different in that it was open-source and available free to manufacturers, not a revenue source itself for Google but a way to spread its services and advertising. This strategy succeeded in making Android the most popular mobile OS.

  • Microsoft’s Windows Phone and BlackBerry’s platform failed to gain traction against iOS and Android. Developers did not widely support these platforms.

  • Successful platforms tend to be early to market, take advantage of complementarity to increase demand, are broadly open but also curated, and obsess over the user interface and experience.

  • User experience is more important than just the user interface. Platforms like Facebook succeeded where MySpace failed by restricting customization and prioritizing a consistent, usable experience.

The lesson is that powerful platforms can influence companies within their sphere, and being early as well as differentiating on user experience are keys to platform success.

  • Stripe saw an opportunity to build a payments platform that was easy for online and app-based merchants to use, handling all the complexity of accepting different payment methods in different currencies.

  • Existing payment offerings did not meet merchants’ needs of simply wanting to accept payments from customers in various ways. Stripe aimed to shield merchants from this complexity through a simple API.

  • Stripe’s approach allowed merchants, especially smaller ones, to easily try new things, iterate, and experiment without worrying about payment issues. This helped companies grow and find successful business models.

  • Stripe benefited from strong network effects as a two-sided platform connecting merchants and financial institutions. Both groups want to be on the platform with many of the other.

  • Stripe grew rapidly by processing payments for half of US internet users only 5 years after launching. It demonstrated there was demand for a payments platform that focused on great user experience over low fees.

  • Stripe continues adding valuable complementary services like fraud detection that increase overall demand for the platform. Patrick Collison’s goal is for Stripe to grow the economy of the internet.

  • ClassPass originally offered an “Unlimited” membership where users could take an unlimited number of classes each month for a flat fee. However, this proved unsustainable for the business.

  • As more users signed up for Unlimited and took classes frequently, it negatively impacted ClassPass’ finances. They had to pay studios each time an Unlimited member took a class.

  • Exercise studios using ClassPass have finite capacity in each class. ClassPass had to balance filling otherwise empty spots versus limiting reservations to leave room for regular studio members paying full price.

  • Revenue management techniques helped address this by optimizing how inventory (available spots) were allocated based on demand and willingness to pay. Data and algorithms aimed to sell as much inventory as possible to higher-paying customers first.

  • ClassPass convinced skeptical studios to try letting it manage reservations for a class or two. Studios saw benefits like fewer last-minute empty spots as ClassPass reserved spaces closer to class time. This gained studios’ trust in ClassPass’ revenue management approach.

So in summary, the Unlimited model proved unsustainable, and revenue management was key to ClassPass balancing the interests of users and studios in its marketplace for physical fitness classes with finite capacity.

  • ClassPass and Postmates are examples of online-to-offline (O2O) platforms that bridge the online and physical worlds. ClassPass provides access to in-person workout classes, while Postmates delivers goods from local stores and restaurants.

  • ClassPass ran into profitability issues with its unlimited subscription offering because its costs grew proportionally with members’ class attendance, while revenues only grew with total members.

  • Postmates addressed this by charging merchants a percentage of each delivery item’s cost. This generated revenue per transaction and allowed the unlimited offering to remain viable regardless of growth.

  • Rent the Runway also employs revenue management techniques like rotating inventory availability based on demand in different locations to preserve the value of its durable clothing assets over time.

  • Emerging examples show O2O platforms are also spreading from consumer-facing to business-to-business applications, connecting excess supply like warehouse space or freight hauling needs between companies online.

  • Upwork provides tools for project management and payment. By 2016 it was facilitating over 3 million projects annually worth over $1 billion.

  • Cvent provides a platform for finding and booking event venues. It handles activities like understanding facilities, availability, and pricing that traditionally required contacting venues directly. The company has expanded beyond venue booking to include mobile invitations, tickets, and surveys. By 2015 it was helping manage $9.2 billion in events annually for 15,000 customers.

  • Sociologist Robert Merton created the concept of the “focus group” when commissioned during WWII to study how Americans responded to mass communication. Focus groups and surveys are now commonly used market research tools handled by online platforms like UserTesting, Survata, dscout, and Google Consumer Surveys.

  • O2O (online to offline) platforms have proliferated globally to match local opportunities, environments, and transportation systems. Examples highlighted include BlaBlaCar for intercity ridesharing in France, Go-Jek for motorbike taxis and delivery in Indonesia, and various Chinese platforms like Edaixi for laundry and Guagua Xiche for mobile car washing. Large funding amounts for O2O businesses in China are also noted.

  • O2O (online-to-offline) platforms have expanded into industries like transportation, exercise, and lodging that deal with physical goods and services. These industries have unique constraints like perishable inventory and capacity limits.

  • Platform owners employ techniques like revenue management to better match supply and demand given these constraints. They also look to complement existing products to drive more demand.

  • China has been a major driver of O2O innovation, applying these models across many industries.

  • O2O platforms can scale rapidly by leveraging existing assets like cars/studios, controlling the customer experience, and using data/algorithms. This has attracted significant investor funding.

  • A key benefit is increasing utilization of underused assets like empty seats/vehicles. This can reduce waste and allow the same level of services with fewer total assets, reducing environmental impact.

  • Technologies like smartphones, location services, cloud computing were necessary precursors for most O2O models to become viable in recent years.

  • Uber was founded in 2008 in Paris and originally focused only on limos, but expanded to allowing standard cars and drivers on its platform through services like UberX and UberPool.

  • Uber’s two-sided network effects and user experience gave it rapid growth and big valuation, hurting incumbent taxi companies and the value of taxi medallions.

  • Regulation has been one way incumbents have tried to push back against Uber and other platforms, with laws and statutes targeting transportation network companies.

  • In finance and other industries, incumbents may remain ubiquitous through regulation but become more like “financial utilities” - heavily regulated, unprofitable businesses.

  • In wireless communications, platform providers like Apple (iOS) and Google (Android) captured the vast majority of profits, while manufacturers struggled despite huge sales, as platforms took over.

  • Examples given are Xiaomi, once valued very highly but saw sales declines, and Samsung, which has seen sales declines for years despite engineering expertise.

  • Platforms have significant advantages over product-focused companies in industries they enter, through network effects, access to users, and ability to capture bulk of profits from ecosystems.

Platforms are able to overcome significant information asymmetries that previously prevented certain transactions from occurring. By collecting data from transactions and allowing both sides to rate each other, platforms increase transparency and trust. This has enabled new markets like ridesharing and homesharing that were previously hampered by lack of information about strangers. While not perfect, ratings systems and reviews significantly reduce risks and uncertainty. This reduction of information asymmetries is a major reason platforms have been able to grow rapidly in various industries. However, platforms also threaten traditional companies by building their own brands and reducing pricing power over time. As platforms curate larger selections under their brand, individual sellers risk becoming commoditized and losing control over their reputation and pricing.

  • Platforms like ClassPass want to attract both prestigious brands and a large consumer base. But as they grow, they want to capture more of consumers’ spending.

  • Platforms have power through controlling the user interface and experience. They can feature lesser-known suppliers over more famous ones to build their membership.

  • Prestigious brands like SoulCycle have opted to stay off platforms to avoid losing control over pricing and the user experience.

  • Platforms usually prefer lower prices than suppliers because of network effects and the economics of two-sided markets. Lowering prices on one side increases demand on both sides.

  • In two-sided markets, users on one side benefit from more users on the other side. Platforms must manage both sides of the network.

  • Aggressive pricing strategies used by platforms like Uber make more sense when considering network effects across both sides of the market. Lower prices increase demand exponentially.

  • Giving away products or services for free, or even paying users, can be profitable for platforms due to increased demand across the network from cross-side effects. This is a sustainable strategy for two-sided markets like credit cards.

  • Platforms like credit card networks and Uber can drive adoption by initially offering low, zero, or negative prices to consumers on one side of the market. This increases use which attracts merchants/drivers who want more customers, even if costs are a bit higher.

  • Switching costs make it worthwhile for platforms to subsidize early adoption through incentives. Once users are “locked in”, the network becomes more valuable even without incentives.

  • Network effects create a tendency toward winner-take-all markets, giving larger platforms an advantage in attracting more users. This encourages platforms to lower prices initially to gain market share.

  • Balancing pricing and incentives on both sides of the market is key, as extracting too much value from one side can cause it to leave, hurting the other side as well. Successful platforms carefully manage value for all participants.

  • Platforms often become multisided, with different subgroups benefiting from each other’s participation through the platform, even if indirectly. This amplifies network effects.

  • While platforms threaten incumbent businesses used to traditional economics, some old guard sectors have endured even with platform competition, showing disruptive impacts have limits. Competition and commoditization tend to lower prices over time.

Here is a summary of the key points about Airbnb in the passage:

  • Airbnb introduced a second product in the urban lodging market, aimed at people wanting something different than traditional hotels, often cheaper short-term lodging in various residences that may include interactions with the host.

  • Research found Airbnb was responsible for a 10% decline in overall hotel revenues in Austin, Texas over five years, but the impact was not uniform - lower priced hotels and those not catering to business travelers were most affected.

  • Hotel stays have more durable differentiation than urban rides, as business travelers often want a specific location or rewards program, and there are meaningful differences in amenities and what families or extended stays need.

  • When offerings are differentiated and customers can be locked into a specific brand, the destructive potential of platforms like Airbnb is more limited compared to disrupting undifferentiated urban transportation.

So in summary, Airbnb has disrupted parts of the hotel industry but not completely replaced it, as hotel stays have more product diversity and differentiation than urban transportation, and some customers prefer or require specific hotel qualities.

  • In 1993, author Robert Wright correctly predicted that the internet would allow knowledge and answers to be gathered from a diverse, global crowd of people, rather than just traditional institutions.

  • This crowd-generated knowledge surpasses what traditional core institutions like libraries can provide. The internet contains over 45 billion web pages, far more than the estimated 130 million books published. It also includes diverse media like videos, images, etc.

  • The uncontrolled nature of the crowd creates challenges in organization and filtering content. Early search engines like Yahoo tried to manually organize content into categories, but couldn’t keep up as the web grew exponentially.

  • Google’s breakthrough was realizing web pages could be automatically organized through an algorithm (PageRank) that analyzes the link structure between pages, allowing smooth navigation of uncontrolled crowd-generated content.

  • The lack of hierarchy in the crowd also means some individuals misbehave or engage in harmful speech. While concerning, most crowd participants create and contribute content in good faith. Moderation techniques aim to address bad actors without compromising the freedom and benefits of the crowd.

  • The passage discusses how markets and decentralized crowd behavior can spontaneously generate valuable new knowledge through emergent processes like price signals. This is known as a “kind of magic.”

  • Friedrich Hayek was an early proponent of this idea. He argued centralized economic planning could never work because no single entity has access to all relevant private knowledge dispersed among individuals.

  • Free markets overcome this by allowing decentralized transactions and using prices to transmit knowledge widely in a simple way. Prices emerge from interactions and reflect information not apparent from observing individuals.

  • Prediction markets directly apply Hayek’s insights by creating event-based securities that people trade based on their private beliefs. Market prices aggregate dispersed knowledge into accurate probability assessments.

  • Other crowd-based solutions have since emerged to harness emergent knowledge, though organizing a crowd deliberately remains challenging due to coordination problems. The passage emphasizes Hayek’s foundational work on the surprising knowledge-generating powers of dispersed, decentralized systems.

  • Linus Torvalds started the Linux operating system as an open-source hobby project in 1991 and asked for community contributions and suggestions.

  • Key principles that helped Linux succeed with a crowd-developed model include openness, noncredentialism, verifiable and reversible contributions, clear outcomes, self-organization, and geeky leadership.

  • The open development model allowed contributions from a wide range of individuals and companies. Noncredentialism lowered barriers for anyone to contribute regardless of formal qualifications.

  • Contributions to software are verifiable through code review and testing, and easily reversed if problems emerge. This enables crowd collaboration in a way that isn’t feasible for more subjective creative works.

  • Contributors knew Linux would always remain free and open-source software under the GNU GPL license, providing clear expectations for how their work would be used.

  • Developers self-organized around components they found relevant rather than being assigned tasks, allowing the overall effort to remain decentralized and flexible.

  • Linus Torvalds exemplified “geeky leadership” through his technical skills, vision articulation, and engineering-focused mindset, gaining credibility within the community.

  • The chapter discusses how large crowds can be organized and motivated to collaborate on major projects through approaches like open source operating systems like Linux.

  • Critical principles that enabled Linux’s success include openness, non-credentialism, self-assignment of tasks, verifiability of contributions, and clear goals/outcomes. Strong “geeky leadership” from Torvalds also helped maintain the culture and momentum.

  • Wikipedia’s early predecessor Nupedia initially followed only some of these principles and saw limited success, adopting a closed, credential-focused model. This changed when Wikipedia switched to a fully open, non-credentialist wiki model.

  • Wikipedia successfully activated the crowd by adopting principles of openness, non-credentialism and self-organization rather than closed, standardized work processes and credential requirements.

  • These “geeky” approaches to collaboration are now gaining more acceptance beyond open source within mainstream business via tools like Slack that facilitate open, fluid collaboration across organizations.

  • Researchers conducted an experiment on Topcoder, an online platform, to solve the difficult problem of sequencing human white blood cell genes faster and more accurately.

  • The problem was opened up to a crowd of over 400,000 software developers, describing it in general algorithmic terms without specialized domain knowledge required.

  • 122 teams submitted over 650 algorithms, which were far more diverse and inexperienced than typical experts. However, 30 beat the accuracy of the best benchmark, and 16 beat another top benchmark. The best were over 30x faster.

  • The researcher, Karim Lakhani, found the crowd typically outperforms internal company/organization solutions in over 700 past challenges, exceeding rather than failing expectations.

  • Simply having biases or overconfidence does not fully explain why experts are outperformed. While real, organizational dysfunctions are also not the primary reason.

  • The core is often mismatched to today’s complex, fast-changing problems, while the crowd’s diversity allows it to consider novel solutions outside typical expertise and blind spots. When properly engaged, the crowd can beat experts by thinking differently.

  • The core (internal R&D teams) can become misaligned and out of date as new knowledge is constantly created in many disciplines. Advances like CRISPR gene editing have changed fields rapidly.

  • Having a diversity of perspectives is valuable for problem solving, but the core lacks this. Outsiders from different backgrounds may provide insights. While unlikely contributors seem unnecessary for the core to hire, they could help crack tough problems.

  • The crowd is massively marginal - it contains many smart, motivated people far away from any organization’s core in terms of geography, fields of expertise, etc. This diversity makes it a powerful problem-solving resource.

  • Smart companies are engaging the crowd for tasks like software development, data analysis, market research, and crowdsourcing ideas. Platforms like Amazon Mechanical Turk, Topcoder, Kaggle, Kickstarter connect organizations to the crowd.

  • Well-designed systems that divide work, verify quality, and provide incentives can harness the crowd’s collective abilities, similar to open source projects like Linux. The crowd is becoming a viable way to outsource work and gain insights.

  • Crowdfunding reverses the traditional model of bringing a product to market only after spending a lot of money developing it. With crowdfunding, people can pre-buy or pledge support for an idea, providing pre-funding and validation.

  • Crowdfunding allows creators to build a movement and social capital around an idea to attract financial capital in advance.

  • Crowdlending platforms matched individual lenders and borrowers but now also attract large institutional investors who analyze opportunities at scale.

  • Emerging platforms like Teespring allow YouTube/Instagram influencers to convert their social capital/followers into financial support by selling merchandise. This shows social capital can be converted into financial capital through crowd-powered businesses.

  • Established companies now watch the crowd for innovations, often buying promising startups rather than disrupting them. Quantopian aimed to open automated investing, traditionally done by hedge funds, to the crowd by building a platform for users to test algorithmic trading strategies.

Here is a summary of the key points about Quantopian from the passage:

  • Quantopian is a platform that enables crowd-sourced algorithmic investing by attracting traders from around the world to develop and share quantitative investment strategies/algorithms.

  • Over 100,000 traders from 180 countries have used the platform to develop over 400,000 algorithms. These traders tend to have technical backgrounds but be new to finance.

  • The platform handles backtesting algorithms, automatic trade execution, record keeping, compliance, etc. to enable the crowd-sourced strategies.

  • Having many diverse yet well-performing strategies is valuable since they can be combined into an optimal portfolio, improving returns. Crowd-sourcing aims to discover lots of such low-correlation strategies.

  • Contests have shown the crowd can outperform professionals, with outsiders winning 14 out of 19 contests so far. Future fund performance will show how it compares to quant hedge funds.

  • High-profile investor Steven Cohen has invested in Quantopian and given them $250 million to invest based on crowd strategies, showing confidence in their model.

  • MakerBot donated two 3D printers to help generate prototypes for prosthetic hands more quickly.

  • Owen and Van As came across plans for the “Corporal Coles hand”, a 19th century prosthetic with moving fingers built by surgeon Robert Norman.

  • Inspired by this, they built their own version called “Robohand” for a boy named Liam who was born without fingers on one hand.

  • Rather than patent Robohand, they shared the 3D printing plans online. Since then, over 1,800 plastic, printed hands have been created for people in 45+ countries through decentralized coordination online.

  • 3D printing enabled the cost of a basic prosthetic to plunge over 99% overnight. Designs continued improving through online collaboration between users and makers. Some were customized for specific activities.

  • Crowdsourcing and online sharing allowed for fast, widespread distribution of affordable prosthetics through a decentralized network.

Satoshi Nakamoto published a paper in 2008 proposing a new digital currency called Bitcoin. The key ideas were:

  • Bitcoin would operate independently of governments and financial institutions like banks, allowing for online payments without fees or requiring identity.

  • Transactions would be recorded in a shared public ledger called the blockchain to prevent “double spending” of coins.

  • The blockchain would be maintained in a decentralized way by a network of computers running Bitcoin software. These computers, called nodes, would verify and add new blocks of transactions to the chain.

  • The first node to solve a complex computational puzzle for each new block would be rewarded with newly created Bitcoins, incentivizing them to secure the network. This process is called mining.

  • Over time, the system would become more secure as the computational power needed to alter the blockchain grows exponentially with each new block added.

Nakamoto’s design created robust incentives for individual miners and users to participate without coordination, allowing the system to operate independently and survive changes over time. Many saw Bitcoin as a feasible new digital currency and payment system with economic and technical merits.

  • The global financial crisis of 2008 led some to doubt the stability and fairness of the existing financial system, making alternative currencies more appealing. Bitcoin emerged during this time as a new digital currency independent of governments.

  • In 2010, Laszlo Hanyecz made the first known exchange of Bitcoin for a physical product (pizza), establishing an early value for Bitcoin. However, Bitcoin’s value fluctuated wildly in its early years.

  • Exchanges like Mt. Gox facilitated Bitcoin trading but also experienced major hacks and losses. Mt. Gox filed for bankruptcy in 2014 after losing millions of Bitcoins.

  • As Bitcoin mining became more difficult, specialized computer hardware and locations with cheap electricity gave miners competitive advantages. One early miner sadly discarded a hard drive containing thousands of Bitcoins now worth millions.

  • While economists questioned Bitcoin’s viability as a currency, some saw potential in the blockchain technology underlying Bitcoin - a distributed public ledger that records transactions immutably. This sparked exploration of blockchain applications beyond digital currency.

  • Early examples of blockchain applications included recording student transcripts, tracking diamonds and goods to prevent fraud, and facilitating finance of international trades and stock offerings. Nasdaq launched a blockchain project for private stock markets.

  • In June 2016, the Republic of Georgia announced a project to design and pilot a blockchain-based system for land title registry. Moving elements of the land registry process to the blockchain was expected to reduce costs for homeowners and reduce possibilities for corruption by making the records unalterable on the blockchain.

  • Advocates argue that smart contracts on the blockchain could automate and decentralize contractual agreements in a transparent, immutable way without requiring trust in a centralized third party. This could potentially reduce costs and transaction times for activities like international trading.

  • As an example, in 2016 a Seychelles Trading transaction conducted on the blockchain reduced a typical seven-day process to just four hours by leveraging the blockchain and smart contracts to automatically execute elements of the international trade.

  • Bitcoin and blockchain technology sparked ideas about using decentralized networks and smart contracts to disrupt traditional companies and organizations. Some saw potential for direct peer-to-peer platforms that could undermine large tech companies.

  • Advocates believed the blockchain could become a universally available, transparent ledger for all kinds of digital transactions and records, not just cryptocurrency. This could include things like contracts, software, bank accounts, insurance policies, and investments.

  • The vision was that smart contracts could automatically execute transactions and enforce agreements on this global ledger without centralized oversight. This could open up new possibilities for streamlined and low-cost transactions.

  • Decentralization supporters argued that large companies and institutions like banks had become too powerful in both developed and developing nations. Blockchain technology potentially offered an alternative that moved important economic functions to a decentralized global network of computers.

  • There was significant interest and experimentation with new decentralized ledger technologies from the potential efficiencies and opportunities they could provide. However, it remained unclear how well massive decentralization would actually work in practice or what roles various blockchain-based systems might play.

  • While some saw potential for undermining large companies, others were more skeptical that blockchain alone could replace existing institutional structures and business models. Decentralization raised many open questions about practical implementation and governance.

The DAO and Bitcoin/blockchain efforts show that while crowd-based and decentralized models have potential advantages, they also face significant challenges that companies are better structured to address.

The DAO aimed to function as a venture capital fund without any central organization, relying entirely on software and crowd voting. However, it was hacked shortly after launching due to vulnerabilities in its code. This undermined the vision of rule by immutable code. Ethereum later underwent a “hard fork” to recover funds, angering some who felt this undermined decentralization.

Bitcoin also struggled with organizational challenges as it grew. Disagreements between programmers over technical issues led to infighting and paralysis. Performance suffered as mining power concentrated in China. This showed that avoiding “too big to fail” institutions is difficult without some centralized coordination and decision-making.

Both examples demonstrate that while crowds and decentralized models offer innovations, companies remain better equipped to navigate complex technical and strategic issues over time through centralized leadership and governance structures. Fully replacing companies may not be viable or desirable given the ongoing need for coordinated adaptation and problem-solving.

  • Centralization of mining power poses risks to Bitcoin’s decentralization. If any entity controlled over 50% of the total mining power, they could unilaterally decide which transactions are validated, taking control away from others.

  • The concentration of mining in China was especially troubling. The Chinese government closely oversees financial institutions and sometimes intervenes directly. Having control over Bitcoin behind China’s “Great Firewall” could turn the cryptocurrency dream of freedom from government into a nightmare.

  • New technologies like cryptocurrencies and smart contracts are often hoped or believed to decentralize economic activity by reducing the roles of large companies. But in reality, large companies continue to dominate many industries and the economy overall.

  • Ronald Coase’s seminal 1937 paper “The Nature of the Firm” asked why so much economic activity takes place within hierarchical companies rather than purely through markets. He argued it is often more cost-effective for companies to organize production than relying entirely on atomistic market exchanges.

  • Later work in transaction cost economics, building on Coase, helped explain why companies persist as the dominant form of organization even as technologies reduce some transaction and information costs. Additional factors like asset specificity, uncertainty, and frequency of transactions continue to give companies advantages over pure markets.

  • According to transaction cost economics, markets often have lower production costs while hierarchies typically have lower coordination costs.

  • Digital technologies are great at reducing coordination costs by lowering search costs, communication costs, and costs associated with information goods.

  • As coordination costs go down, markets become more attractive relative to hierarchies because their comparative disadvantage shrinks.

  • We should see more use of markets and less use of hierarchies as digital technologies improve and diffuse. This is happening to some degree with more outsourcing, offshoring, and freelancing.

  • However, firms are still prevalent, so the basic TCE framework needs to be updated to account for things like incomplete contracts and residual rights of control.

  • Contracts can never fully anticipate all future contingencies, so they will always be incomplete to some degree.

  • Ownership confers residual rights of control over assets not fully covered by contracts. This affects incentives and is an important reason why firm boundaries and ownership structures matter.

  • Highly decentralized structures like Bitcoin, blockchains, and The DAO struggle because they can’t address problems of incomplete contracts and lack of oversight that firms are designed to handle through ownership and management.

  • Smart contracts and decentralization don’t fully address why companies exist - contracts can never be truly complete due to unforeseen contingencies.

  • While new technologies like sensors and AI could help monitor outcomes and simulate decisions, they also enable counterparties to consider more possibilities, so contracts remain incomplete.

  • Companies provide stability through their long-term nature and established legal framework that individuals/contracts cannot.

  • Even disruptive companies like Uber, Airbnb still use the traditional corporate structure like other established companies.

  • Management remains important for coordination in a complex world. Managers solve problems, facilitate communication, and use social skills to persuade and motivate people.

  • Successful companies emphasize egalitarian ideas where ideas from all levels are considered through experimentation rather than manager bias. This opens the company to new opportunities despite risks of bad ideas.

So in summary, while technologies enable new models, companies will continue serving important economic and coordination functions that decentralized models cannot fully replace due to the inherent incompleteness of contracts. Management also remains critical for navigating complexity through social skills.

  • Oliver Cameron wanted to test whether outsiders could evaluate student projects for Udacity as well as employees, and potentially faster.

  • He had external people evaluate projects that were also evaluated internally, and found the evaluations to be quite similar. This showed tasks didn’t need to be done by Mountain View employees.

  • They started experimenting with paying external evaluators less, and found they could do it for 30% less cost.

  • Cameron launched the external evaluation program in just 6 weeks through manual testing and experimentation.

  • The COO Vish Makhijani had not formally approved the plan but supported Cameron’s experiments, seeing potential benefits of tapping broader talent pools.

  • Makhijani described Udacity’s culture as not requiring approval for new ideas and giving employees flexibility to experiment and quickly launch new initiatives.

So in summary, Cameron was able to show external evaluators could do the same quality work as employees, and developed the external program quickly through experimentation and hands-on testing to reduce Udacity’s evaluation costs. The culture supported these kinds of agile innovation experiments.

Here are the key points about the principal-agent problem and how it relates to the economics of incentives:

  • The principal-agent problem arises in situations where one party (the principal) delegates a task to another party (the agent) but cannot directly monitor or control the agent’s actions. This creates an incentive issue.

  • Bengt Holmström’s 1979 paper provided the foundation for analyzing these types of incentive contracting problems. It showed how providing proper incentives to agents is important to get them to act in the principal’s interests.

  • Holmström and Milgrom noted in their 1994 paper that firms themselves can be usefully thought of as incentive systems - they establish rules, norms, and incentive structures to motivate employee behavior.

  • This framework of analyzing incentive problems in delegation situations, and thinking of firms as incentive systems, spawned a large body of subsequent economic research on topics like incomplete contracts theory and the design of optimal incentive contracts.

  • In summary, Holmström’s initial paper established the principal-agent framework, while his later work with Milgrom emphasized how firms function as incentive systems - providing the foundation for lots of follow-on research on incentives in economic organizations.

The passage discusses how new technologies are enabling more sustainable, inclusive and higher-paying jobs. It provides two examples:

  1. 99Degrees Custom is an apparel maker in Massachusetts that embraces automation to partially automate their production line. This has allowed them to create more varied, skilled and better-paid jobs compared to traditional factory work. It has increased value for more people.

  2. Iora Health employs health coaches who work with patients to help them stick to treatment plans. Studies show this can reduce costs by 15-20% while leaving patients and workers better off, as the coaches add value through compassion and motivation even though they don’t have medical degrees.

The key point is that depending on how new technologies are used, they can either concentrate power and wealth or distribute decision-making and prosperity in a way that creates more inspiring and purpose-driven workplaces. The potential for technologies to improve lives increases the importance of clarity on goals and deeper thinking about values.

  • The passage discusses several companies and projects that involved using data and predictive modeling/algorithms to make predictions or decisions. This includes projects on predicting housing prices, research impact, judicial decisions, gifted student identification, and clinical vs mechanical diagnosis.

  • One study found predictive algorithms could make better procurement decisions than humans. Israeli judges were found to grant parole at a higher rate just before breaks.

  • Predictive modeling has been used to forecast Supreme Court decisions and identify more low-income/minority gifted students. Clinical psychologists have also been shown to be outperformed by mechanical prediction methods.

  • The level of data and information available to judges/clinicians can impact their decisions compared to purely data-driven algorithms. Overall the passage argues data and predictive methods can potentially make more objective decisions than humans in some cases. Several examples of projects applying these techniques in different domains are discussed.

  • Early work in artificial intelligence in the 1950s-60s showed promise but faced limitations in computational power and understanding of human cognition. One early program demonstrated basic skills in answering questions but had no common sense reasoning abilities.

  • Neural networks reemerged in the 1980s with the development of backpropagation, a method for training multi-layer neural networks developed separately by Paul Werbos and Geoff Hinton. This enabled networks to learn complex patterns from large datasets.

  • While neural networks have achieved human-level performance in narrow tasks like image recognition, they still lack general commonsense reasoning, semantic knowledge, explanatory capabilities and other attributes of human-level intelligence. Giving machines these broader cognitive abilities remains a challenge.

  • The goal of early AI research with systems like the Perceptron was to simulate human learning and reasoning at the level of neurons and connections between neurons. This inspired newer neural network approaches but our understanding of human cognition is still limited.

  • Most commercial AI systems today focus on narrow applications rather than general human-level intelligence. Broader capabilities in reasoning, knowledge and explanation remain targets of ongoing research. Historical limitations in computing power and understanding of the brain have held back progress but newer techniques continue to push the boundaries.

Here is a summary of the key points from the specified section:

  • The section discusses the development and increasing capabilities of artificial intelligence and robotics technologies. It notes that advances like Moore’s law and improvements in machine learning algorithms have led to accelerated progress.

  • Examples are given of robots and AI automating various jobs like agriculture, supply chain logistics, customer service, and cooking. Driverless vehicles and drones are being used for tasks like cargo transportation and oil pipeline inspection.

  • new technologies like 5G networks and advanced sensors are enabling even more robotic applications. Some experts predict a “Cambrian explosion” of new robot forms and uses as these enabling technologies mature.

  • The decreasing cost and size of components is allowing for more distributed robotic systems. Examples mentioned include automation of tasks like milking cows and oil extraction operations using robotics.

  • It’s suggested 90% of all digital data created is unlabeled and much of this data, if organized and utilized, could enable new AI and robotic applications. Overall the section outlines the growing capabilities and uses of robotics and AI technologies across various industries.

Here is a summary of the article:

The article discusses how milking automation is gaining popularity on dairy farms. Some key points:

  • Milking robots allow cows to be milked voluntarily 24/7 without human assistance. This improves cow comfort and welfare.

  • Automatic milking systems are more expensive than traditional milking parlors initially but provide a good return on investment over time due to higher milk yields and labor savings.

  • Farmers can remotely monitor cows using the robots to catch health issues early. This improves herd management and milk quality.

  • Younger farmers seem more open to adopting the technology. Automation allows them to scale up their herds without needing to hire as much labor.

  • Overall, milking robots are helping dairy farmers operate more efficiently and profitably while providing cows with a better environment. As the technology advances and costs come down further, milking automation looks set to become standard on more and more dairy operations globally.

  • The Rocky Mountain News, a major newspaper in Denver, published its final edition in February 2009 after declining advertising revenues forced it to shut down.

  • Many newspaper companies saw their stock values decline by over 90% during the 2000s, including McClatchy which owns several major papers.

  • In 2013, Jeff Bezos purchased the Washington Post. Several magazines also faced bankruptcy or closure including Penthouse, National Enquirer, and Men’s Fitness.

  • Newsweek had over 3 million in circulation in the 1990s but ceased print publication in 2012, moving fully online. The New Republic was also sold in 2016 as it struggled financially.

  • Playboy announced in 2015 that it would cease publishing nude photos as part of a rebranding effort. Hugh Hefner remained influential but stepped down as chief editor in 2017.

  • The recorded music industry saw worldwide sales decline 45% between 1999-2015, from $27 billion to around $15 billion currently, as digital downloading and streaming grew.

  • Major music retailers like Tower Records and HMV went bankrupt in the 2000s as CD sales declined. Musicians also sought financing through “Bowie bonds” and other securitization deals using their music royalties.

  • Between 2005-2015, about 20% of U.S. malls closed as retailers like Sears and Macy’s cut stores. The real estate crash led to bankruptcies like General Growth Properties.

  • Telephone companies like AT&T saw landline revenues drop from $77 billion in 2000 to $16 billion in 2010 as cell phone use increased dramatically. Radio ad revenues also declined from $20 billion to $14 billion over the same period.

Here is a summary of the key points from the provided sources:

  • In 2014, Google decided to shut down its Google News platform in Spain in response to demands from Spanish publishers that Google pay fees to link to snippets of their news articles. This came amid ongoing debates around online platforms paying for news content. (Source 1)

  • By 2016, WhatsApp had grown to over 1 billion active users, sending more than 40 billion messages per day. At the time of its 2014 acquisition by Facebook, WhatsApp had around 600 million monthly active users while employing just 70 people. (Sources 2, 3, 4)

  • Jeff Bezos assigned Rick Dalzell to develop Amazon Web Services, which launched services like S3 for cloud storage and EC2 for cloud computing resources in 2006. By 2007 over 290,000 developers were using AWS, and by 2016 it accounted for 9-12% of Amazon’s total revenue and over half its operating income, making it the fastest growing enterprise technology company. (Sources 5, 6, 7, 8, 9, 10, 11)

  • The music industry saw declines from $37 billion in 2000 to $15 billion in 2016, amid the rise of peer-to-peer file sharing platforms like Napster, LimeWire and increasing unbundling of tracks from albums. Streaming and downloaded tracks now generate over half of US music revenues. (Sources 12, 13, 14, 15)

Here is a summary of the key points about mergers and acquisitions involving Nokia, Microsoft, AOL, and Time Warner:

  • Microsoft acquired Nokia’s mobile phone business in 2013 for $7.2 billion, hoping to boost its presence in the smartphone market. However, the deal was a failure and resulted in almost $8 billion in write-downs for Microsoft, the largest in the company’s history.

  • AOL was acquired by Time Warner in 2000 in a historic $160 billion merger, the largest merger in business history at the time. The merger turned out to be a major failure due to the rapid growth of the internet and changes in the media business.

  • BlackBerry’s market share peaked at 20% of the global smartphone market in 2009 but steadily declined over the following years as iOS and Android gained dominance. By 2016, BlackBerry announced it would no longer produce its own phones, marking the decline of the once-dominant BlackBerry platform.

  • The failures of the Nokia-Microsoft and AOL-Time Warner mergers show how technology markets and business models can change rapidly, making large acquisitions highly risky endeavors if the acquiring companies cannot adequately adapt to new conditions in the market.

Here is a summary of the article “Baidu to Invest $3.2 Billion in Online-to-Offline Services,” Reuters, June 30, 2015:

  • The article discusses Chinese internet company Baidu’s plans to invest $3.2 billion over three years in its online-to-offline services.

  • Baidu is the dominant search engine in China. It has been pushing to develop services that connect online users with offline services like restaurants, movies, and other local businesses.

  • The investment aims to expand Baidu’s model of using its search and maps data to provide recommendations and reservations for local services. It wants to become an important platform for offline transactions.

  • Baidu sees huge potential in China for combining the massive online audience with opportunities in industries like local services, food delivery, and travel. Close to 80% of China’s internet users are now accessing the web via mobile.

  • The investment signals Baidu’s ambition to develop from a search company into a major online platform integrating online and offline activities in China. It aims to compete with Alibaba and Tencent in e-commerce and digital payment areas.

That covers the key details from the Reuters article about Baidu’s $3.2 billion investment in expanding its online-to-offline service platform in China. Let me know if you need anything else summarized.

  • A 2015 study found that prize-based contests were able to solve computational biology problems, outperforming existing algorithms like MegaBLAST. The winning algorithms were developed by non-academics. (Lakhani et al.)

  • In over 700 challenges on InnoCentive, non-experts solved problems academic and industrial experts could not solve. (Lakhani interview)

  • When problems are broadcast to large crowds, “all bugs are shallow” as more eyes find solutions. Nonexperts outperformed experts in solving some problems. (Jeppesen and Lakhani)

  • Amazon’s Mechanical Turk and other platforms crowd-source microtasks from transcription to design tasks. Nonexperts perform many information processing jobs. (Wilson, Bernstein et al.)

  • Kickstarter successfully crowdfunded the Veronica Mars movie, reaching its goal in the first 12 hours. It premiered in 2014. (Thomas, Rappaport, Business Wire)

  • Indiegogo and others now allow large companies to crowdfund and get early customer feedback. Hedge funds invest in platforms like LendingClub. (Kastrenakes, Demos, Banjo)

  • Platforms like Teespring allow anyone to be an entrepreneur by designing and selling custom products. (Andreessen interview)

  • Schumpeter argued economic change often comes from new players outside the existing industry or expertise, the ” heterogeneity of men.” (Schumpeter)

Here are the key points from the summary:

  • Chapter 12 discusses the dream of decentralizing all things using technologies like cryptocurrencies and blockchain.

  • Bitcoin was created in 2008 by Satoshi Nakamoto as a peer-to-peer electronic cash system not relying on central authorities. It introduced the concept of mining and a blockchain to verify transactions.

  • Early bitcoin exchanges like Mt. Gox suffered hacks and failures, demonstrating issues in scaling the technology. Over time, mining became more centralized in locations with cheap electricity.

  • The value of bitcoin rose dramatically but also crashed, demonstrating volatility. However, some companies and institutions have experimented with using blockchain for applications like academic certificates, diamond tracing, and preventing counterfeiting.

  • Companies like Overstock became early corporate adapters of accepting bitcoin as payment. Some have also explored issuing bonds or dividends using blockchain/cryptocurrency technologies.

  • The chapter discusses the vision of using technologies like cryptocurrencies and blockchain to decentralize systems and cut out centralized intermediaries across many applications and industries. But challenges around security, scalability and volatility remain.

Here is a summary of the key points from the passages:

  • In June 2015, announced plans to issue the world’s first cryptosecurity, using bitcoin technology. They later conducted the first blockchain-based public stock offering in March 2016.

  • Nasdaq reported that its Linq platform, which used blockchain for securities issuance, reduced settlement risk exposure by over 90%.

  • When Ornua, an Irish food company, used blockchain for trade finance with Barclays, it was one of the first real-world applications of the technology beyond currencies.

  • Nick Szabo’s 1996 paper described smart contracts as computerized transaction protocols that execute terms of a contract. This was a precursor to modern smart contract platforms like Ethereum.

  • Ethereum, launched in 2015, created a decentralized platform for distributed applications and smart contracts using blockchain and cryptocurrency.

  • The DAO attempted to organize governance as a decentralized autonomous organization but suffered a hack in 2016 which led to a hard fork splitting it into Ethereum and Ethereum Classic.

  • The failure of bitcoin technology to scale led to departures from the community by some, while China’s influence over mining raised centralization concerns.

  • Blockchain and distributed ledgers may impact firm boundaries by facilitating multi-party transactions without traditional firms, but key economic factors like transaction costs and incomplete contracts still favor firms for now. Overall, while technology changes firm organization, companies are not made obsolete.

Here is a summary of the key points from the article:

  • The article discusses how harnessing automation and AI can help create a future where technology augments human work and improves people’s quality of life, rather than replacing jobs.

  • New technologies like robotics, artificial intelligence and advanced analytics are automating tasks and activities across many industries and functions. This disruption creates risks of job loss and increased inequality if not managed well.

  • To maximize the benefits of automation, companies, governments and workforce need to take proactive steps like retraining workers, implementing job rotations, and investing in communities. Workers also need to actively reskill and learn new skills.

  • For automation to benefit all, governments need to modernize education systems to emphasize computing, science and engineering skills. They also need to make continued reskilling and lifelong learning accessible and affordable for workers.

  • Companies need to automate ethically and have a dialogue with stakeholders to ease concerns. They should also create new jobs and roles that integrate humans and machines, and redeploy displaced workers internally.

  • If the transition is managed well with a long-term vision, automation can augment human capabilities and improve lives by freeing up time for more creative and meaningful work. But active management of change is required from all stakeholders to realize this future.

  • Summarize discussed summarizing various topics such as Mechanical Turk, Amazon services, smartphones, Angry Birds game, anonymity and digital currencies, artificial intelligence techniques, blockchain technology, Bitcoin, crowd leveraging, creativity, crowdsourcing, and platform businesses.

  • Topics included Amazon’s Mechanical Turk platform for work, AWS cloud computing platform, Amazon Go cashier-less stores, Amazon S3 storage service. Also discussed smartphone companies’ profits, the Angry Birds mobile game, how digital currencies provide anonymity, various AI techniques from rule-based to pattern recognition.

  • Blockchain and its use for currencies like Bitcoin and potential applications were covered. Crowdsourcing discussed leveraging the crowd for tasks, funding, and research. Creativity discussed limits of computers and role of human connections. Platform businesses and how they create value through networks and complements was summarized.

Here is a summary of the key points from the provided text snippets:

  • En platforms refers to online-to-offline (O2O) platforms that integrate online and offline services, like food delivery apps. They help address information asymmetries between customers and brick-and-mortar businesses.

  • Bitcoin is considered a digital currency that operates on blockchain technology without a central authority.

  • Current accounts refers to bank accounts used for regular expenditures rather than long-term savings.

  • The “curse of knowledge” refers to the difficulty of imagining what it’s like to not know something that is perfectly obvious to you.

  • Companies can use crowdsourcing to acquire new customers by encouraging existing customers to refer others through incentives or discounts.

  • Free apps rely on customer service to retain users and gather valuable data, since they don’t generate direct revenue.

  • Cvent is a company that provides event management software.

  • The Cyclopaedia published in 1728 was an influential early encyclopedia.

Here is a summary of key points about perishing/perishable inventory:

  • Perishing/perishable inventory refers to goods that have a limited shelf life, such as food, flowers, etc. It poses unique challenges for retailers and inventory management.

  • The main risk is that perishable goods will expire before they can be sold, resulting in lost revenue and sunk costs. Careful management is needed to minimize expired inventory.

  • Marginal costs are higher for perishable goods compared to non-perishable items, since more waste is likely. Factors like transportation and storage costs also factor in.

  • Revenue management strategies aim to optimize sales and pricing of perishable goods over their lifetime. This may involve dynamic pricing that lowers prices as expiration nears.

  • Online-to-offline (O2O) platforms help manage perishable inventory risks by enabling just-in-time replenishment and reducing overstock. They also provide sales data insights to better forecast demand.

  • The ability to move perishable goods quickly and optimize pricing is key to success. Platforms that facilitate this, like food delivery apps, help minimize losses from expired inventory.

  • Platforms create economic value by facilitating exchanges between different user groups (producers and consumers). They are characterized by network effects where each additional user increases the value for other users.

  • Successful platforms have strong user experiences, are open to extensions/complements, and benefit from economies of scale. Incumbents tend to struggle competing against disruptive platforms.

  • Music and print media industries were disrupted by digital platforms like Napster, Spotify, Facebook that unbundled products and monetized through other means like ads or subscriptions.

  • Non-digital industries are also adopting platforms to manage perishable inventory, pricing, and connect online/offline customers like Airbnb, Uber, Postmates.

  • Robotics and automation are expanding to restaurants, warehouses, factories for tasks that are dull, dirty, dangerous or dear. Conditions like falling prices of sensors/actuators and machine learning are enabling this.

  • Blockchain enables new platforms like cryptocurrencies but challenges remain around scalability, governance, and addressing social problems like financial inclusion. Smart contracts provide programmable transactions.

  • AI is evolving from rule/symbol-based systems to data-driven machine learning. Techniques like deep learning are achieving human-level performance inconstrained domains but open challenges remain around general intelligence, bias, and control.

Here’s a high-level summary of some of the key concepts, companies and terms discussed in the excerpt:

  • Transportation network companies (TNCs) like Uber that provide ride-hailing services digitally connecting drivers and riders.

  • Two-sided platforms that connect two distinct user groups, like riders and drivers, and must address pricing and costs for both sides. Uber is an example.

  • Network effects which increase a platform’s value as more users join on either side of the network.

  • Information asymmetry challenges for platforms around verifying users/drivers and building trust. Uber uses reputation systems for drivers.

  • Rapid growth of digital platforms like Uber that leverage existing physical assets (cars) without owning most of the assets themselves.

  • Regulation issues for new digital services like temporary bans of Uber’s low-cost UberPop service in some cities.

  • Mentions several other companies/terms - Airbnb, Postmates, Twitter, Wikipedia, Amazon, etc. to provide context around the sharing/platform economy.

Let me know if any part of the summary needs more clarification or expansion. I aimed to synthesize the key takeaways at a relatively high level.

Author Photo

About Matheus Puppe