Machine, Platform, Crowd - Andrew McAfee

“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!”

BOOK LINK:

CLICK HERE

Here’s a summary of the key points:

• Go is an ancient Chinese strategy game that is extremely complex, with more possible board configurations than atoms in the universe. Top human players rely on intuitive skills they can’t fully explain.

• Programming computers to play Go well has been very difficult, because the huge number of possibilities makes brute force simulation ineffective. Key strategies can’t be explicitly programmed due to “Polanyi’s Paradox” - humans can’t articulate their tacit knowledge.

• In 2016, Google DeepMind’s AlphaGo program beat the world’s top Go player through machine learning techniques, analyzing millions of game positions to develop its own strategies. This was an AI breakthrough, overcoming the obstacles of the game’s complexity and Polanyi’s Paradox.

• Some tech companies like Uber and Airbnb have grown huge without owning key assets traditionally associated with their industries, illustrating how digital platforms can disrupt traditional business models.

• The successes of AlphaGo and these tech platforms exemplify the potential of digital technologies like machine learning and platforms to transform industries in novel ways.

• Companies like Uber, Airbnb, Alibaba, and Facebook have achieved massive success and growth despite owning few long-lived physical assets. They rely on software, algorithms, and network effects.

• These “asset-light” companies can expand extremely quickly compared to traditional companies that require large investments in infrastructure and equipment. Their thin ownership layer spreads rapidly.

• Uber grew to operate in 300 cities globally in just a few years. Airbnb doubled booked nights within a year. Alibaba enabled over $500 billion in sales with no physical storefronts.

• Facebook attracted nearly 1 billion daily users sharing content and grew profits massively, without generating any content itself. It offers highly targeted ads based on user data.

• Despite its long history and assets, GE partnered with an online community to crowdsource ideas for a new consumer product. Rather than relying solely on its internal R&D and marketing, it reached out to tap external creativity.

• This highlights how even giants see value in using networked communities and assets like software and data, not just physical assets, to innovate in the modern economy. Asset-light models are spreading.

• Recent advances in AI, platforms, and crowdsourcing are disrupting businesses and industries. Three examples:

1. AlphaGo defeated the world champion Go player Lee Sedol in 2016, demonstrating major advances in AI.

2. Platform companies like Facebook and Airbnb have grown rapidly and disrupted established industries without owning traditional assets.

3. GE used crowdsourcing to help design and market a new home ice maker called Opal.

• These examples illustrate three major trends reshaping business: the rise of intelligent machines, the power of platforms, and the availability of the crowd.

• Companies need to rethink the balance between minds and machines, products and platforms, and their internal core and external crowds. Machines, platforms, and crowds have become more powerful, so organizations need to understand how to best leverage them.

• We are now in the second phase of the second machine age. In the 1990s, digital technologies boosted productivity by automating routine work. Now, previously sci-fi technologies like autonomous vehicles and AI are becoming real and fundamentally transforming the economy.

• The goal of the book is to provide lenses to make sense of these changes and guide business success. Rebalancing mind and machine, product and platform, and core and crowd is essential.

• The second phase of the second machine age is characterized by AI systems that can do non-routine work like creative tasks and natural language processing. Powerful mobile devices have also spread rapidly, connecting billions of people digitally.

• Around 1900, electricity started replacing steam power in factories. Many successful incumbent companies did not survive this transition. Existing expertise and processes made it hard for them to see the potential of new technologies like electricity.

• Electrification enabled major conceptual changes like assembly lines and unit drive (motors for each machine rather than one central power source). But these were hard for incumbents to envision.

• By the early 1930s, over 40% of the major industrial trusts formed between 1888-1905 had failed, largely due to electrification. Surviving trusts often struggled in later years.

• Incumbents can be blinded by their existing knowledge and success with the status quo, making it hard to adapt to new technologies, even if initially only marginally better. This “curse of knowledge” helps explain why technological disruptions often catch experts by surprise.

• In the mid-1990s, IBM’s Deep Blue chess computer defeated world champion Garry Kasparov, demonstrating the growing capability of AI systems. More recently, Google’s AlphaGo AI beat the world’s top Go player Lee Sedol. These victories illustrate how machines are becoming increasingly intelligent and able to perform complex tasks that were previously only done by humans.

• However, AI systems like AlphaGo still rely heavily on human knowledge and heuristics encoded by programmers and engineers. They do not truly “think” independently like humans. There is still progress to be made before AI becomes fully creative and autonomous.

• We are entering a second machine age where digital technologies like AI, robotics, and digitization are advancing rapidly. This is ushering in societal and economic changes on par with the first machine age of electrification and mass production.

• Companies today face competitive threats if they fail to take advantage of new technologies. The book aims to provide guidance on rebalancing the relationships between minds and machines, products and platforms, and the core and the crowd in light of technological advances.

• Insights will be drawn from economics and other disciplines to understand the business implications. The goal is practical but not prescriptive, as much uncertainty remains during this time of transition.

• Each chapter will summarize key takeaways and questions for applying the ideas. The overall aim is to help businesses capitalize on the opportunities of the second machine age.

• Twenty years ago, businesses adopted a division of labor where computers handled routine tasks like math, record-keeping, and data transmission, freeing up people to exercise judgment, creativity, and problem-solving.

• This approach was spurred by business process reengineering in the 1990s, which aimed to streamline cross-functional business processes rather than siloed departmental tasks. It was accelerated by enterprise software systems and the rise of the World Wide Web and e-commerce.

• People were seen as uniquely capable of judgment, in contrast to routine digital tasks. This aligns with the idea from behavioral economics of two modes of thinking - fast, intuitive System 1 and slow, effortful System 2.

• Business training has focused on honing both systems, teaching principles like accounting and finance as well as judgment via case studies. The belief has been that judgment is beyond computation.

• However, recent AI advances are challenging this division of labor, as machines become capable of some tasks requiring judgment, not just routine calculation. This raises questions about the human role in an AI-capable world.

• Many believe people should focus on using intuition and judgment (System 1 thinking), while machines handle logical and analytical tasks (System 2 thinking). This is called the “standard partnership.”

• However, research shows human judgment is often inferior to statistical and data-driven approaches:

• A model beat managers at predicting outcomes of equipment purchases. Veteran managers did no better than new ones.

• A model using just 4 variables beat wine experts at predicting wine quality and prices.

• Models beat housing experts at predicting home sales and Google search data improved predictions.

• A model beat tenure committees at predicting impactful academic hires.

• Judges made harsher rulings after their alma mater lost a football game.

• Objective tests identified more minority gifted students than subjective nominations.

• A simple model beat legal experts at predicting Supreme Court rulings.

• Reviews show algorithms match or beat human experts in ~50% of studies, while humans win in only ~6%.

• More companies are relying on data over intuitions and judgments. The “standard partnership” does not always work well. Human judgment still has a role but should be relied on less.

• Algorithms and data-driven models are increasingly outperforming human experts in making decisions and predictions in a wide variety of domains. Examples include diagnosing medical conditions, making business forecasts, and assessing creditworthiness.

• However, models can only replace human judgment where sufficient data exists to create them in the first place. Many important decisions lack the volumes of structured data required.

• The intuitive thinking system (System 1) that humans rely on is amazing but also buggy, with many biases and flaws. The rational system (System 2) has limitations too.

• We often lack insight into when our intuitive thinking is sound versus when it is biased. The rational system can simply provide justifications for the intuitive system’s flawed judgments.

• To make the best decisions, we need a new partnership between human and machine capabilities. Pure algorithmic decisions should be used where possible, with human oversight and input reserved for situations that lack sufficient data.

• Many highly automated decisions are already common, like credit scoring, e-commerce recommendations, and dynamic pricing. The “second economy” of fully automated decisions is expanding.

• Dan Wagner and his colleagues on the 2012 Obama campaign used machine learning to score each voter on their likelihood to support Obama, turn out to vote, and be persuaded by campaign messaging. This allowed more targeted and effective TV ad buying compared to traditional demographic-based approaches.

• After the election, Wagner founded Civis Analytics to bring this data-driven ad targeting to companies, by matching their customer lists to TV viewership data. This makes ad buying more of an optimization exercise and less reliant on intuition.

• However, even data-driven systems have risks of perpetuating societal biases, like Latanya Sweeney found with racially biased Google ad results. Careful design and testing is required to minimize bias.

• Machine systems can be tested and improved more systematically than human decision makers. The goal should be choosing the approach that minimizes errors and biases, not perfect decisions.

• Some balance of human and machine intelligence is likely best. Humans are needed to define problems, objectives, and evaluation criteria. Machines are good for optimization and pattern recognition once goals are properly defined.

• Humans have strengths like gathering diverse sensory data and using common sense that computers currently lack. The “broken leg role” involves humans noticing exceptions that trip up computer models.

• However, algorithms often make better decisions than humans. Humans tend to be overconfident in their intuition and judgment. Keeping score of human vs algorithm accuracy over time is important.

• Uber learned this lesson when its surge pricing algorithm kicked in during a hostage crisis, prompting severe criticism. Human override of algorithms is sometimes essential.

• Structured interviews that quantify human judgments as inputs to hiring algorithms have proven very effective at Google.

• Overall, data-driven algorithmic decisions are superior to human intuition and judgment in most situations where both exist. Some find this uncomfortable, but it leads to better outcomes. Humans and algorithms working together, with humans providing oversight and exceptions, is often an optimal arrangement.

Here are a few key points from the summary:

• The “standard partnership” of humans and machines often relies too heavily on human intuition and judgment, which are prone to bias. Algorithms and data often lead to better decisions.

• Human reasoning has two systems - System 1 which is fast, intuitive, and prone to bias; and System 2 which is slower, analytical, and more rational. Relying too much on System 1 leads to poor judgments.

• Evidence shows algorithms alone usually outperform even expert human judgment for decisions and forecasts. Humans should be taken out of the loop for many decisions and judgments.

• In some cases, humans should provide oversight on algorithmic decisions. In other cases, algorithms should make decisions entirely on their own with no human input.

• Subjective human judgment still has a role for some types of decisions, but overall we need to rely less on intuition and more on data/algorithms.

• Organizations should move away from long-term forecasts and toward constant experimentation and testing.

• Since the advent of digital computers, researchers have tried to get them to mimic human intelligence and reasoning. This field became known as artificial intelligence (AI).

• Two main approaches to AI emerged early on:

1. Symbolic/rule-based AI - Tries to encode human knowledge and reasoning as explicit rules that computers follow step-by-step. Focused on replicating logic and high-level reasoning.

2. Statistical/pattern recognition AI - Inspired by how young children learn languages, it uses statistical learning principles to discern patterns in data that can then be used to make predictions and decisions. More focused on perceptual abilities.

• Initially, symbolic AI was dominant and had some impressive successes like proving mathematical theorems.

• However, symbolic AI struggled to deal with the real world’s messiness and complexity. It proved difficult to codify all the murky rules needed for commonsense reasoning.

• Meanwhile, statistical AI made steady progress on narrower tasks like character recognition. Computing power also expanded rapidly, enabling more ambitious statistical learning.

• By the 1980s, statistical AI was ascendant while symbolic AI entered a long period of reduced funding and interest known as the “AI winter.”

• Recently, statistical AI has seen major advances thanks to bigger data sets, neural networks, and increased computing power. This has enabled impressive accomplishments in areas like computer vision and natural language processing.

• Early attempts at symbolic AI, based on codifying rules, failed to achieve human-level performance in complex real-world tasks like speech recognition and translation. These systems performed poorly, generating unintelligible or nonsensical outputs.

• Two main obstacles explain the failures of symbolic AI: 1) the vast number of intricate rules needed to perform well in most domains, and 2) Polanyi’s Paradox - humans can’t identify or articulate many of the tacit rules they use to accomplish tasks.

• An alternative machine learning approach tries to have systems learn from experience and repetition, like humans, rather than following explicitly programmed rules. Early neural networks like the Perceptron showed promise but were limited.

• Recent breakthroughs in deep learning and neural networks, combined with increased computing power, have enabled major advances in machine learning systems for applications like Go, image recognition, and language translation.

• Machine learning offers a potential path to overcoming Polanyi’s Paradox by developing AI that can learn tacit rules automatically through experience rather than having them explicitly programmed.

Here is a summary of the key points about the recent progress in neural networks and artificial intelligence:

• Neural networks have become the dominant type of AI, fueling major advances recently. This is largely due to more powerful hardware (Moore’s Law), big data, and key algorithmic innovations.

• Deep learning allows neural nets to learn in an unsupervised manner by identifying patterns in large datasets. This has enabled breakthroughs in image recognition, speech transcription, translation, and more.

• Tech giants like Google and others have achieved remarkable results by applying deep learning to products and challenges like energy efficiency in data centers. The techniques are generalizable across domains.

• Cloud computing and APIs enable these AI capabilities to spread quickly to other companies. Machine learning is already being deployed in diverse sectors like agriculture.

• Overall, the field of AI seems to be finally fulfilling some of its long-standing promise, thanks to neural nets and deep learning. These techniques are likely to continue spreading rapidly and impacting industries and economies unevenly in the coming years.

• The symbolic, rule-based approach to AI has declined in favor of statistical, machine learning techniques like deep learning. This is because machine learning has proven much more effective at tackling difficult problems like speech recognition and natural language processing.

• Machine learning systems like deep neural networks can overcome Polanyi’s Paradox - they can learn to do tasks even if people cannot explicitly state the rules for how to do them. This makes them well-suited to automating white-collar and back-office work.

• Companies are now using deep learning systems for tasks like processing insurance claims and providing customer service. Microsoft has built a speech recognition system that achieves human-level performance.

• AI experts believe we are moving into an “AI-first world” where many decisions will be automated and augmented by AI rather than relying on human intuition alone. The capabilities of AI systems are advancing rapidly.

• Unsupervised learning, where systems learn patterns from raw data rather than being explicitly trained, is seen as key to achieving artificial general intelligence. But current successes in AI rely primarily on supervised learning.

• Eatsa is a restaurant that offers nutritious, tasty, and affordable vegetarian meals with quinoa as the main ingredient. Customers order, pay, and receive their food without interacting with any employees.

• This illustrates the phenomenon of virtualization, where transactions and interactions that used to occur between people now take place via digital interfaces. Many business processes don’t actually require physical transformation or movement, just the transfer of information.

• Virtualization is spreading to many areas, like airline check-ins and customs processes. Interactions between people are being removed from these transactions.

• This doesn’t necessarily mean full automation, as there are still people involved (the customers). But it does allow for streamlining and removal of labor from certain steps.

• The key enabler is information technology, which allows easy capture, storage, retrieval, and transmission of information. Physical presence is no longer required.

• This virtualization through IT allows businesses to reduce costs and improve consistency in transactions. But it also reduces opportunities for human interaction and judgment.

• As more processes become virtualized, companies and customers need to consider the tradeoffs involved and the effects on jobs, customer service, and human relationships.

• Virtualization of services is accelerating as networks and digital devices become more ubiquitous. For example, the number of bank tellers has declined as online and mobile banking has grown.

• Some argue certain services require human interaction, but examples like Wealthfront show some are willing to virtualize even traditional high-touch processes like investment advisory.

• Virtualization appears to be a secular trend that will gain momentum over time as digital natives make up more of the population.

• Automation is also advancing quickly, with examples like robot chefs, though most robots still lag behind human capabilities like dexterity and perception.

• Improvements in robotics and AI are leading to a “Cambrian Explosion” in automation across many domains.

• Increased virtualization and automation will transform many jobs and industries, potentially requiring adjustments like retraining programs. But history suggests the long-term outcome is broadly positive for prosperity and standards of living.

The Cambrian Explosion was a brief period when most major forms of life on Earth appeared. Today, we may be on the cusp of a similar explosion in robotic innovation driven by advances in data, algorithms, networks, the cloud, and hardware.

• Data: The amount of digital data is exploding, enabling more machine learning.

• Algorithms: New techniques like deep learning improve with more data.

• Networks: Faster wireless networks allow robots to coordinate and share learning.

• The Cloud: Provides computing resources to lower barriers to robotics innovation.

• Hardware: Exponential improvement continues per Moore’s Law, enabling cheaper, better robot parts.

Together these advancements are fueling a robotic Cambrian Explosion as innovators rapidly experiment, generate data, and improve algorithms. Robots are well suited for dull, dirty, dangerous and expensive tasks. Examples include construction site monitoring by drones and crop monitoring in agriculture. The explosion in robotics will likely lead to major economic disruption as many jobs are automated.

• Drones and other machines are taking over dull, dirty, dangerous, and expensive (dull, dirty, dear) tasks previously done by humans, like crop monitoring, equipment inspection, and spraying. This frees up humans to use our minds in more valuable ways.

• Humans are still better at some physical tasks requiring dexterity, vision, and locomotion. So in some cases people work alongside robots, like in Amazon warehouses where robots bring shelves to human pickers.

• Humans have advantages over robots for now due to our evolved senses, hands, and mobility. But robot capabilities are advancing rapidly.

• New digital production techniques like 3D printing allow more complexity in part shapes and remove the need for molds. This enables faster, more flexible, and more distributed manufacturing.

• Overall, progress in digital technologies is enabling machines to do more tasks interacting with the physical world, freeing up humans for knowledge work. And it’s enabling more complexity and flexibility in how we arrange atoms to build physical things.

Here are a few key points summarizing the section on creativity and heat exchangers:

• Many people believe creativity is an irreducible human ability, but recent advances in AI suggest machines can be quite creative.

• Heat exchangers are an important but often overlooked technology that transfers heat between fluids. Designing good heat exchangers requires deep technical knowledge and creativity.

• An AI system developed by researchers was able to autonomously design innovative new heat exchanger designs that beat human experts. This demonstrates AI’s potential for highly creative work.

• The AI leveraged evolutionary algorithms, running thousands of simulations to refine designs. This trial-and-error process reflects how human creativity often works.

• While the heat exchanger example shows AI can be creative, human creativity still has unique qualities. AI may complement and enhance human creativity more than replace it outright.

• Creativity relies on intuition and “out of the box” thinking, which remain human strengths. But the boundaries of what AI can do are rapidly expanding.

Here are a few key points about natural and artificial digital creativity:

• Generative design software can automatically design novel products like heat exchangers that meet specified requirements, without human designers being involved. The software is not constrained by human biases or experience.

• Adding sensors and real-world data to generative design can produce even more optimized and unusual designs, like an asymmetric race car chassis. The software mimics natural evolution to create efficient, resilient structures.

• In science, AI like IBM’s Watson can generate promising new hypotheses that humans may not think of, by analyzing large volumes of research papers. An experiment showed Watson successfully predicted cancer research directions.

• In the arts, AI programs can create original paintings, music, and more without human input. Some find the art emotionally moving, even if it lacks a distinct human style.

• As digital creativity advances, computers may be capable of novel innovations and “Eureka!” moments in science and engineering. But some argue AI art lacks depth compared to human creations. The boundaries between natural human creativity and artificial creativity continue to blur.

Aesthetics - our sense of beauty and appeal - are complex but can be understood through principles like the golden ratio. This knowledge is being used in technology, like The Grid’s customized websites and IBM’s AI-generated recipes. Architects employed computers to design the energy-efficient Shanghai Tower.

However, there are limits to computer creativity. Humans uniquely understand the human experience needed for many creative endeavors. Computers lack the consciousness and deep comprehension of humanity required to reliably generate great art, lyrics, fiction, etc.

Our social nature also gives us an edge over computers. Jobs tapping into human drives like compassion and solidarity are least threatened by automation. A computer coach couldn’t truly motivate and unite a girls’ soccer team. Computers can detect emotions but are far from deeply understanding social relationships.

The key is combining computers’ strengths with human strengths like creativity and emotional intelligence. This allows more total creativity and progress in the world.

• In the mid-1990s, many traditional industries like newspapers, magazines, radio, and recorded music were thriving and generating billions in revenue annually.

• Landline phones were ubiquitous, with most households having one, and mobile phones were still an expensive novelty. Long-distance phone calls were more expensive than local calls.

• The Internet and digital technologies profoundly disrupted these industries over the course of about one generation. Entire business models were overturned.

• For example, classified ads and music sales plummeted as free alternatives emerged online. News and entertainment content became available for free on sites like Craigslist, Google, Facebook, and YouTube.

• The speed and scale of the disruption across so many industries was unprecedented in business history. The Internet fundamentally transformed the economy.

• Many industries were disrupted in the 1990s and 2000s by digital technologies, including music, newspapers, magazines, malls, telecommunications, and radio.

• Digitization made information goods essentially free to reproduce and share. Digital copies are perfect replicas of the original. Networks like the internet allow instant distribution.

• The music industry was transformed as CD sales dropped and digital piracy emerged. Major labels consolidated, stores like Tower Records went bankrupt. Total industry revenues declined dramatically.

• Newspapers and magazines saw print ad revenue collapse as readership moved online. Many iconic publications went bankrupt or ceased print editions. Newsrooms had huge job cuts.

• Shopping malls declined as online retail expanded. Many malls closed and mall operators went bankrupt.

• Landline telephones declined as mobile phones took over. Long distance revenue dropped precipitously.

• Radio station revenue declined with competition from digital music. Stations consolidated into large operators like Clear Channel.

• Kodak went from a $31 billion valuation in 1997 to bankruptcy in 2012, failing to adapt to digital photography.

• Digital technologies allowed disruption by making information goods free, perfect in quality, and instantly distributable. This challenged traditional business models.

• Platforms like the internet leverage the economics of being free, perfect, and instant to rapidly grow and disrupt incumbent businesses. The marginal cost of distribution and reproduction is nearly zero.

• Platforms build on each other through combinatorial innovation. The World Wide Web was a platform built on top of the internet protocols.

• Platforms like Craigslist and Google News disrupted newspapers by providing free classifieds and content aggregation.

• Mobile messaging platforms like WhatsApp disrupted SMS messaging by providing free messaging over data networks. They benefitted from strong network effects - more users made the platform more valuable.

• Facebook acquired WhatsApp for $22 billion due to its massive user base and network effects.

• Amazon Web Services emerged from Amazon’s own efforts to standardize interfaces between its systems. It provided modular, recombinable computing resources accessible over the internet.

• Platforms can proliferate and disrupt rapidly due to their ability to leverage the economics of being free, perfect, and instant, build on other platforms, and benefit from network effects.

• The internet and digital technologies have disrupted many industries over the past 20 years, leading to falling revenues but more options for consumers. This happened due to the “free, perfect, and instant” economics of digital information goods on networks.

• Digital goods have a marginal cost of nearly zero to reproduce, are perfect copies of the original, and can be transmitted instantly around the world on networks. This gives them an advantage over traditional physical goods and services.

• Network effects mean that networked digital goods become more valuable as more people use them, favoring bigger networks.

• Platforms are digital environments with near-zero marginal costs of access, reproduction, and distribution. They utilize network effects and economies of scale.

• Platform economics have disrupted industries like retail, journalism, photography, music, etc. Incumbents have struggled while consumers gained more choice.

• Music industry revenues dropped as platforms like Napster, iTunes, and streaming services emerged. These unbundled and rebundled music in new ways that were very attractive to consumers but challenging for rights holders.

• Platforms are transforming business models across industries. Traditional practices face upheaval due to the economics of platforms and digitization.

• In 2007, Steve Jobs and Apple launched the revolutionary iPhone, which was an instant success. However, Jobs initially refused to allow third-party apps on the iPhone, wanting Apple to control the platform.

• This was a mistake, as economists have shown that allowing complementors to build on your platform generates network effects and increases the value of the platform. Jobs was ignoring basic economic principles.

• Apple board member Art Levinson and head of Internet services Eddy Cue convinced Jobs to change course in 2008. This opened up the App Store and unleashed a flood of apps that made the iPhone more useful and popular.

• Jobs benefited from changing his mind and embracing complements, leading to the iPhone’s dominance. But his initial failure to appreciate basic economics nearly stalled Apple’s momentum in 2007.

• The story illustrates the value of economic thinking about platforms, ecosystems, and complements. Even visionary leaders like Jobs can benefit from economic insights about how to maximize value.

Here is a summary of the key points about the iPhone’s existence on the wrong side of an important debate:

• The iPhone was intended to be a full-fledged computer, with a processor, memory, storage, operating system, user interface, and applications (apps).

• Steve Jobs initially wanted Apple to develop all the apps for the iPhone in order to maintain tight control over the user experience. He didn’t want outsiders creating apps that could potentially “mess up” the iPhone.

• Highly placed people at Apple argued in favor of allowing external developers to make iPhone apps, including senior executives and board members.

• Jobs eventually changed his mind and allowed outside developers to create iPhone apps after the product launched. This was the right decision.

• Having a large variety of apps at range of prices, created by independent developers, helped make the iPhone hugely popular.

• The apps and the iPhone are complementary goods - the demand for one increases demand for the other. Different consumers have different “killer apps” they want.

• Turning the App Store into an open marketplace enabled discovery of these varying preferences and allowed satisfaction of them, in a way Apple alone could not.

• Many developers were willing to make free apps, creating tremendous consumer surplus, which greatly increased demand for the iPhone.

• When apps like Angry Birds and Shazam are free, it generates consumer surplus for iPhone users who feel like they are getting a bargain. It also shifts the iPhone demand curve outward since more people are willing to pay for the iPhone due to the availability of desirable free apps.

• Many different entities create free apps, including companies using a “freemium” model, apps with advertising revenue, customer service apps, public service apps, and apps that pair with physical products.

• Opening up the iOS platform brought tremendous innovation and growth by allowing outside developers to create apps. This increased the volume, variety, and quality of apps beyond what Apple alone could have done.

• Open platforms provide the benefits of greater consumer surplus, shifting demand outward, collecting data on usage, and creating new revenue opportunities. However, too much openness can also enable harmful content and behavior.

• Curation through app approval processes and reputation systems can help platform owners maintain quality and control. Platform owners have significant discretion over how they configure and curate their creations.

• Platforms like Apple’s App Store and Google’s Android can force hard choices on companies in their sphere of influence, as they rely on these platforms for distribution and revenue.

• Successful platforms tend to be early to the space, take advantage of complementary goods economics, open themselves up to a broad range of contributors, curate their offerings, and obsess over user experience.

• Android provided a robust alternative to Apple’s iOS app platform by being free and open source. This allowed it to spread quickly and become the world’s most popular mobile operating system.

• Attempts by Microsoft, BlackBerry, and others to build competing mobile platforms largely failed, showing there is often only room for a handful of platforms in a given domain.

• Facebook succeeded where social networks like Friendster and MySpace failed in large part due to delivering a better user experience and interface. Stripe also succeeded by deeply understanding user experience needs.

• The lessons are to be early, leverage complementary goods, curate offerings, and obsess over user experience in order to build a successful platform. Powerful platforms force hard choices on companies in their domains.

• Platforms like Stripe are powerful drivers of many successful companies today because they aggregate both supply and demand.

• Platform owners benefit when they open up their platforms to outside contributors, as this brings in complementary products that increase demand for the platform owner’s core product.

• After opening up a platform, owners typically have to curate contributions to maintain standards. Low quality contributions can diminish the platform’s value.

• Platform owners compete based on their ability to attract contributors and effectively curate their contributions. It’s harder to build a platform if competitors already exist, especially if consumers are unwilling to use multiple platforms.

• Successful platforms pay close attention to user interface and experience. Ease of use helps attract both contributors and users.

• Many platforms are two-sided, with different types of users on each side. Stripe has merchants on one side and financial institutions on the other. The presence of each side attracts the other.

In summary, platforms leverage network effects and complements to grow demand. Good curation and user experience help them succeed. Many have two-sided network effects.

• Xercise seems an unlikely activity for digital disruption, but ClassPass shows how platforms can transform atom-based industries like fitness studios.

• ClassPass offered an unlimited membership plan in 2014 to attract users. But it proved unsustainable as people worked out too frequently.

• Unlike digital goods, physical services like classes have limited, perishable inventory. Studios worried ClassPass would take too many spots.

• ClassPass used data and revenue management software to fill empty class spots at the last minute. This benefited both parties.

• Short experiments convinced studios that ClassPass optimized class spots for all. Complements like the revenue software expanded ClassPass’s value.

• Platforms like ClassPass face unique challenges bringing the economics of digital goods to physical services with finite capacity. But complements like software can align incentives.

• ClassPass ran into issues with its unlimited offering because costs grew faster than revenues as usage increased. This was because revenues were tied to number of members while costs were tied to number of classes taken.

• Postmates found a way to better balance revenues and costs with its unlimited offering by charging merchants a percentage of each delivery. This ensured revenues scaled with usage.

• There is a rise of “online-to-offline” (O2O) platforms that bridge digital and physical worlds, like Uber, Airbnb, GrubHub, etc. They bring network effects and digital economics to physical goods/services.

• O2O platforms are spreading beyond just consumer offerings into business services too. Examples include platforms for trucking, warehousing, freelancing, etc.

• Core economic fundamentals that made digital platforms successful also apply to these O2O platforms. We can expect them to continue spreading throughout the economy.

• O2O (online-to-offline) platforms like Airbnb, Uber, and Instacart are using technology to make traditional services and businesses more convenient and efficient.

• These platforms provide tools like pricing optimization, yield management, and demand mapping that were previously only available to large companies. This levels the playing field for small businesses.

• O2O platforms combine the unlimited scalability of digital information (the economics of bits) with physical goods and services (the economics of atoms). This creates powerful network effects.

• As they grow, O2O platforms provide tremendous liquidity by matching large numbers of buyers and sellers. This reduces uncertainty and makes transactions easier.

• Insights from economics, operations research, machine learning, and user interface design have all contributed to the success of these platforms.

• Successful O2O platforms are emerging around the world, tailored to local environments and opportunities. Examples include BlaBlaCar, Go-Jek, Edaixi, and others.

• Digital platforms like Uber, Airbnb, and ClassPass are spreading into industries that deal with physical goods and services (online-to-offline or O2O platforms).

• These industries have perishable inventory and capacity constraints, unlike pure digital platforms, so O2O platforms use techniques like dynamic pricing to balance supply and demand.

• O2O platforms benefit from complementary products that increase overall demand, just like digital platforms.

• O2O platforms are emerging globally, including in China which has been an innovation hotbed.

• O2O platforms can scale rapidly due to quick onboarding, leveraging existing assets, controlling the experience, and using data/algorithms. Investors see the potential and are funding aggressive expansion.

• By increasing asset utilization, O2O platforms provide benefits like variety and convenience while also using resources more efficiently to tread lightly on the planet.

Here are the key points about platforms vs. product makers:

• Platforms like Uber have disrupted many incumbents through powerful network effects. Regulations have sometimes been the best defense for incumbents.

• In finance and other regulated industries, incumbent firms may become “utilities” - ubiquitous but unprofitable.

• The same dynamic has played out in the smartphone industry. Platforms like iOS and Android capture most of the profits, while product makers like Xiaomi and Samsung struggle despite making popular devices.

• Platforms have many inherent advantages over product makers:

  • Network effects
  • Control over the user experience and ecosystem
  • Access to user data
  • Ability to leverage common components across products
  • Economies of scale

• As platforms spread, product makers across many industries may see shrinking margins and insecure positions, even if they make excellent products. Platforms are likely to capture more profits and value.

The key takeaway is that platforms benefit from strong network effects and ecosystem control, which allows them to capture the lion’s share of profits in many industries, relegating product makers to low/no margin utility roles. This trend will likely continue as platforms become more pervasive.

• Platforms like Stripe, ClassPass, Postmates, and Transfix are part of a larger trend of platforms diffusing into many industries. This will continue due to the significant advantages platforms have over competitors.

• Not all business activity will happen on platforms, but they will spread widely. Network effects and self-reinforcing growth through iteration and experimentation make platforms powerful.

• Larger platform networks bring greater liquidity, which participants value. They also generate more data to understand and shape member behavior.

• Platforms control the user experience and interface. They can reduce barriers to transactions and steer benefits to themselves.

• Platforms like Uber overcame information asymmetries like drivers being strangers by having riders and drivers rate each other. This “design for trust” facilitates transactions.

• Rating systems help overcome biases like not wanting to rent to strangers. High reputation can beat high similarity.

• Platform brands can become more salient than the brands of companies that participate on them. This gives platforms pricing power and control.

• Overall, platforms will continue spreading due to their advantages, though not all industries will be subsumed. Both platform owners and ecosystem participants can profit, but platforms are gaining power.

• Platforms like ClassPass want to attract prestigious brands but also want to keep consumer mindshare and wallet share. Platforms have an advantage in controlling the user interface and experience.

• Platforms can use revenue management techniques to favor lesser known suppliers over famous brands. Many brands choose to stay off platforms rather than cede control.

• Platforms prefer lower prices than sellers for two reasons:

1. Demand curves for platform services are very elastic - small price drops bring large increases in demand. Maximizing revenue means pricing low.

2. Platforms are two-sided markets. Lowering prices on one side brings more users, which makes the platform more attractive for the other side. This added cross-side network effect amplifies the value of price cuts.

• Two-sided platforms often subsidize one side to bring in more users and make money from the other side. Credit cards charge merchants fees but give rewards to consumers to bring in more transactions.

• The side to subsidize is typically the one with more elastic demand and greater cross-side network effects. For credit cards that is consumers.

In summary, platforms subsidize the side of the market that is most price sensitive, in order to harness cross-side network effects and maximize overall platform participation. This often leads platforms to prefer lower prices than sellers.

• Platforms like Uber rely on network effects and low prices to grow rapidly. Uber offers low prices to attract more riders, which in turn attracts more drivers.

• Uber has invested billions of dollars to subsidize growth and activate these network effects. Investors bet this will allow Uber to eventually profitably charge low prices due to its minimal marginal costs.

• Platform economics put pressure on incumbents used to higher prices. Suppliers face commoditization and more competition on platforms.

• However, platform disruption is not universal. Many incumbent companies in sectors like hotels continue to thrive despite platforms like Airbnb.

• Platform advantages like network effects are formidable but not insurmountable. Incumbents can sometimes slow disruption through regulation, differentiated products, controlling distribution, or aligning with platforms.

In summary, platforms aim to rapidly scale networks through low pricing. This disrupts incumbents used to higher prices but with higher marginal costs. However, legacy companies in some sectors endure despite platforms, slowing disruption through various strategies. The spread of platforms is powerful but not unlimited.

Airbnb rentals represented 12% of all lodging, indicating the company has made significant inroads into the hotel industry. However, Airbnb is not likely to completely displace hotels because lodging is a differentiated market, unlike ride-hailing which offers a more commoditized service. Factors that limit Airbnb’s disruption of hotels include:

• Business travelers often want specific locations or hotel chains with rewards programs.

• Large differences exist in room furnishings and amenities that appeal to different customers.

• Hotels cater to families, extended stays, and conferences in ways Airbnb does not.

• When offerings are differentiated and lock in customers, platforms have less destructive potential.

The differentiation in lodging keeps it from becoming a single-product market vulnerable to a platform like Airbnb. While Airbnb has expanded the market for lodging, it has cannibalized hotel revenues primarily at the lower end, affecting budget hotels and those not serving business travelers. The hotel industry has durable sources of differentiation that limit Airbnb’s ability to completely displace incumbent hotels.

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

• The author contrasts the “core” (dominant pre-Internet institutions like libraries) with the “crowd” (decentralized, uncontrolled participants enabled by the internet).

• The web is like a huge, ever-growing library created by the crowd, unconstrained by traditional hierarchies. It contains vastly more information of all kinds than any physical library.

• The crowd’s lack of control brings benefits like free expression, but also problems like disorganization and misbehavior.

• Early web directories like Yahoo tried to organize the web like a library catalog, but couldn’t keep up.

• Google solved the search problem by using the links between web pages to measure importance, creating order from the crowd’s uncontrolled output.

• The crowd does contain bad actors, but most participants contribute positively. Tools like reputation systems and content moderation help mitigate this issue.

• Overall, the author sees the crowd as a powerful new way to gather diverse knowledge, enabled by the economics of free digital information. It represents a major change from traditional hierarchical institutions.

• The crowd can produce high-quality content when people contribute in good faith. Powerful search tools and enlightened platform policies that assume good faith help minimize bad content.

• Crowd-created collections like Wikipedia spontaneously generate new knowledge, an idea articulated by Friedrich Hayek. Prices in markets emerge from decentralized interactions and transmit information.

• Prediction markets are an innovative way to harvest crowd knowledge. By letting people trade securities corresponding to future events, prediction markets generate probability estimates that are often quite accurate.

• It remains to be seen how platforms like Facebook and Twitter will respond to challenges like fake news. Possible solutions include allowing people to flag content and training AI to spot it.

• The crowd can create things through deliberate organization, not just spontaneously like markets. But organizing crowds online seems very difficult due to lack of incentives, expertise, and authority. Wikipedia shows it is possible with the right principles, norms, institutions and technologies.

Here are a few key points about how Linux was developed as an open source operating system:

• Linus Torvalds started Linux as a hobby project in 1991 and made an open call for contributions and suggestions. This demonstrated the principle of openness.

• Torvalds didn’t require contributors to have specific credentials or work experience. This embodied the principle of noncredentialism.

• Contributions to Linux code can be objectively evaluated and tested to see if they improve the software. If not, they can be easily reverted. This allows for verifiable and reversible contributions.

• The GNU GPL license that Linux uses makes it clear that the software will remain free and open. This provides clarity on the outcome.

• Contributors self-organize and choose what aspects of Linux to work on, rather than being centrally directed. This enables self-organization.

• Torvalds provides geeky, technically proficient leadership, setting an overall vision but not micromanaging the project.

So in summary, principles like openness, noncredentialism, verifiable contributions, clear outcomes, self-organization, and geeky leadership helped Linux become collaboratively developed by a crowd of volunteers. The software could tap into diverse motivations and skills without traditional limitations.

• The crowd is large, diverse, and unpredictable, the opposite of the core. But crowd-based models like stock markets, prediction markets, and search engines can extract valuable information from the crowd’s emergent structure.

• Overcentralization fails due to limitations on knowledge (Hayek’s insights) and articulation (Polanyi’s Paradox). People can’t always articulate what they know and want.

• Products like Linux show how large crowds can collaborate by following principles of openness, noncredentialism, self-selection, verifiability, clarity of goals, and geeky leadership.

• Wikipedia succeeded by adopting these principles, unlike its predecessor Nupedia which failed by being closed, credentialist, and requiring expert editors.

• Getting the right balance between openness and control is tricky - it requires trial and error. But crowd-based models are gaining acceptance even within traditional organizations.

• To leverage the crowd, organizations need to decentralize, incorporate market-like mechanisms, and empower people through open collaboration platforms. But the core may resist giving up power and control.

• Karim Lakhani and colleagues conducted a study where they posed a complex biology problem (annotating genomes of human white blood cells) to the online community Topcoder rather than biology experts.

• The Topcoder community, comprised of software developers with little biology expertise, produced solutions that were much faster and more accurate than existing algorithms developed by biology experts.

• This result is typical - in over 700 similar studies, Lakhani found that crowdsourcing online communities consistently meet or exceed solutions produced by internal experts.

• Organizations often underperform at innovation and problem-solving, not because their experts are unqualified, but because they are mismatched to the problem. Experts tend to have deep but narrow expertise, whereas outsiders bring diverse perspectives.

• Problems today often require “three-dimensional” thinkers who combine expertise across multiple domains. Organizations typically lack these, but online crowds contain many.

• Core members also suffer from organizational dysfunctions like bureaucracy that stifle creativity. Online crowds are free from these constraints.

• To leverage online crowds, organizations must properly structure challenges and incentives to attract participation and effort. But doing so reliably produces superior solutions.

• Organizations often set up teams or departments to tackle challenges, but these “cores” are frequently misaligned or mismatched to the problems they face. Meanwhile, the “crowd” outside the organization, because it is so large and diverse, is often better equipped to solve these problems.

• One reason for this is that critical new knowledge is constantly being created, and core team members can become outdated if they don’t work hard to stay current. The crowd, however, is more likely to contain some people familiar with the latest advances.

• More importantly, most problems benefit from different perspectives and backgrounds. The crowd contains much more diversity than any core team realistically could. Solutions often come from “marginal” people with unfamiliar approaches.

• Smart organizations are starting to use the crowd in various ways: to get work done via platforms like Amazon Mechanical Turk, to find specialized talent through sites like TopCoder, to conduct market research via crowdfunding campaigns, and more.

• The core and the crowd can complement each other. The core sets the strategy and oversees quality, while the crowd provides skills, ideas, and perspectives.

• Crowdfunding reverses the traditional model of developing and marketing a product before assessing market demand. Instead, it allows companies to gauge interest and secure funding from the crowd before investing heavily in manufacturing. This provides valuable customer feedback and pre-sales.

• Crowdlending platforms were initially intended for peer-to-peer lending but are now dominated by institutional investors who use software to identify loan opportunities. This allows established lenders to access new customers identified in innovative ways.

• Incumbent companies are threatened by crowd-based rivals and often acquire promising startups rather than compete against their innovations. Examples include Facebook acquiring Instagram and WhatsApp.

• In investing, quantitative algorithms have long been used by specialized firms to automate trading decisions. John Fawcett founded Quantopian in 2011 to open up quantitative investing to the broader crowd by building a robust technology platform for testing trading algorithms. This aims to democratize algorithmic trading beyond the few thousand specialized quants in the industry.

• The “maker movement” consists of tinkerers, do-it-yourselfers, engineers, and scientists who help each other online by sharing instructions, recipes, blueprints, etc for making things ranging from autonomous go-carts to synthetic biology experiments.

• The biohacking/DIY biology movement creates and shares recipes for useful genetic sequences. It got a huge boost from the discovery of the CRISPR gene-editing tool, which allows modifying DNA with precision. Former NASA scientist Josiah Zayner sold inexpensive CRISPR kits so the technology could be widely available.

• MIT’s Caleb Harper developed “food computers” - enclosed environments where growers can precisely control conditions like temperature and nutrients to experiment with different “climate recipes” for crops. They share recipes and collaborate to improve them.

• When South African carpenter Richard Van As lost fingers in an accident, he collaborated online with Ivan Owen in the U.S. to build affordable 3D-printed prosthetic hands. They shared the design openly so others could replicate and improve it.

• This highlights the “permissionless innovation” of maker crowds - by collaborating and sharing designs openly they can create affordable solutions without needing approval from established institutions.

• Economist John Maynard Keynes observed that influential deceased economists can shape the world through their ideas long after they are gone.

• Similarly, the anonymous founder of Bitcoin, known only as Satoshi Nakamoto, created a potentially revolutionary technology and economic system through Bitcoin.

• Bitcoin and the blockchain technology behind it enable decentralized peer-to-peer transactions without central authorities like banks. This has major implications.

• Blockchain technology could decentralize many industries by eliminating middlemen. It allows new models like decentralized autonomous organizations (DAOs).

• Complete decentralization has risks, like the 2016 hack of the DAO on Ethereum. But the concept is powerful if implemented carefully.

• Decentralization appeals both to individualists who want freedom and to collectivists seeking communal empowerment. It enables new governance and economic models.

• Powerful technologies like blockchain are ideologically neutral - their implications depend on how society chooses to use them. But they unlock new possibilities.

• Truly decentralized models have yet to change the world but contain immense potential to distribute power away from centralized hierarchies in radical new ways.

• Satoshi Nakamoto published a paper in 2008 proposing Bitcoin, a new digital currency that aimed to enable online payments without banks or other financial intermediaries.

• Bitcoin was designed to have properties similar to physical cash - no fees, anonymity, durability, and reuse.

• It relies on cryptography and a decentralized network to solve the “double spend problem” that arises with digital assets that can be freely copied.

• A timestamped, immutable ledger records all Bitcoin transactions to prevent double spending. This ledger is called the blockchain.

• The blockchain is maintained by a decentralized network of “miners” who compete to verify transactions and create new blocks. Miners are rewarded with newly created Bitcoins.

• The proof-of-work system and decentralized nature of Bitcoin make it very difficult for any single entity to take control of the network.

• Nakamoto’s design aligned incentives such that the system could function in a decentralized way, with participants acting in their own self-interest.

• The paper proposed a novel technical solution to enable a decentralized digital currency, at a time when many people were losing faith in banks and governments after the 2008 financial crisis.

• Bitcoin emerged after the Great Recession as a decentralized digital currency independent of governments and institutions. It appealed to those disillusioned with the status quo.

• In 2010, the first known Bitcoin transaction occurred when someone traded 10,000 bitcoins for two pizzas. This gave Bitcoin a nominal value.

• Exchanges emerged to facilitate Bitcoin trading, with Mt. Gox becoming the largest before it collapsed in 2014 due to a major hack and theft of bitcoins.

• Mining bitcoins became increasingly difficult and resource-intensive over time, requiring specialized hardware and cheap electricity globally.

• While Bitcoin was volatile as a currency, the blockchain technology enabling it emerged as the real innovation - an immutable, transparent, decentralized ledger with many potential use cases.

• Applications of blockchain emerged in areas like academic records, diamond certification, preventing counterfeits, financial transactions, stock offerings and trade financing.

• The blockchain enables trusted recording of events and transactions without central authorities. Its potential is still being explored across industries.

• In June 2016, the Republic of Georgia announced a project with economist Hernando de Soto to design a blockchain-based land title registry system. This could reduce costs and corruption by creating immutable records.

• Smart contracts are similar to computer programs in that they involve definitions and specify actions under different conditions. Nick Szabo proposed using the blockchain’s immutability to secure smart contracts.

• For example, a book contract could access sales data and bank accounts to automatically pay authors based on sales, without needing to trust the publisher’s sales reports.

• The blockchain allows smart contracts to exist securely without needing trusted third parties like courts to verify and enforce them. Projects like Ethereum aim to create platforms for smart contracts.

• Some blockchain efforts aim to decentralize concentrated power and information held by large tech companies (“the stacks”). The blockchain offers a way to decentralize and “beat the core at its own game.”

• The idea is to apply blockchain’s decentralizing philosophies more broadly to restructure economics and industries like finance. This represents a solutionist belief that blockchain can solve tough problems involving centralized power.

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

• Bitcoin and blockchain technologies have inspired visions of a decentralized world where companies and other centralized institutions are less necessary. However, companies are likely to remain critical.

• Companies provide key benefits like scale, specialization, and coordination. No blockchain-based system has yet shown it can replace these.

• Companies also handle tasks like ideation, aggregation, and accountability that crowdsourcing often struggles with. Companies will likely use blockchains as tools, not be replaced by them.

• Open, decentralized systems like Bitcoin have advantages but also major limitations. They lack incentives, authority, and agility. Hybrid systems that mix open and hierarchal elements may perform better.

• Private blockchains controlled by companies avoid some limitations of fully open blockchains like Bitcoin’s. However, they also lose key benefits like transparency and lack of central control.

• Early blockchain experiments have shown promise in areas like accounting, identity management, and supply chains. But replacing entire companies will be extremely difficult.

• Companies will likely adopt blockchain technologies selectively where it improves efficiency and enables new business models. But a wholesale shift to decentralized organizations powered by public blockchains is improbable.

In summary, while Bitcoin and blockchains are innovative, companies will remain essential elements of a prosperous economic system for the foreseeable future.

• The DAO was created as a decentralized autonomous organization that would function purely through open-source software and smart contracts, with no human oversight. It raised $162 million in crowdfunding, aiming to democratize business and funding decisions.

• However, The DAO was quickly hacked, with one third of its money stolen by exploiting vulnerabilities in the code. This revealed flaws in the concepts behind The DAO and Ethereum.

• In response, Ethereum executed a “hard fork” to undo the hack. This outraged some who felt it violated the principle of an autonomous system ruled by immutable code.

• Bitcoin has also faced challenges, with arguments between lead developers causing stagnation. Mining power has become concentrated in China, threatening decentralization.

• Mike Hearn, a key Bitcoin developer, denounced the project as a failure due to lack of community agreement and impending technical collapse.

• These cases illustrate the difficulties of replacing traditional corporate structures with autonomous, decentralized software systems at this point. Technical and governance vulnerabilities can arise. More work is needed to achieve the ideal of a fully decentralized organization.

• Coase’s 1937 paper “The Nature of the Firm” posed a fundamental question: If markets are so efficient, why do firms exist? His answer was that firms can minimize certain transaction costs compared to pure markets.

• Digital technologies like the internet and smartphones lower many of the costs Coase identified, suggesting a shift away from firms toward markets. Thinkers like Malone predicted the rise of e-commerce for this reason.

• However, the opposite has happened - firms have gotten more dominant in many industries, with more concentration of sales and profits.

• To understand why, we need to look at subsequent transaction cost economics. This highlights other sources of advantage for firms beyond minimizing Coase’s costs.

• These include unique capabilities, control of intellectual property, ability to make coordinated adaptations, and advantages of scope and scale. Digital tech often strengthens these.

• So firms are likely to remain important and dominant, even as disruptive innovations like blockchain decentralize certain activities. The economics suggest a more nuanced rather than all-or-nothing shift.

• Transaction costs are important in determining whether markets or hierarchies (firms) are more efficient. Markets tend to have lower production costs, while firms have lower coordination costs.

• Digital technologies like search engines and global communication networks reduce coordination costs. This implies markets should be used more as their disadvantage shrinks.

• However, firms are still thriving despite the rise of digital technology. Transaction cost economics needs to be modernized.

• Contracts are inherently incomplete - it is impossible to specify all contingencies. This gives owners residual rights of control.

• Ownership affects incentives and innovation. Employees have different incentives than independent contractors who own assets. This is why firm boundaries matter.

• Firms exist as a solution to incomplete contracts. They determine who gets residual control and who gets rewards.

• Decentralized systems like Bitcoin and The DAO struggle with incomplete contracts and lack of residual control. This can lead to failure modes when the systems evolve in unintended ways.

• Fully decentralized, crowd-based entities will likely never dominate as they cannot solve the problems of incomplete contracts and lack of residual control.

• Companies exist in large part because complete contracts are impossible to write. Technologies like sensors and AI may help anticipate more outcomes, but contracts will remain incomplete.

• Companies serve important economic and legal functions like longevity and predictable governance that freelancers cannot replicate. Even companies pursuing decentralization mostly use traditional corporate structures.

• Management remains important for coordination, persuasion, and solving problems, even with new technologies. Demand for social skills is increasing.

• Managers should adopt an egalitarian approach to ideas, using iteration and experimentation rather than just their own judgment. This helps surface and test new ideas.

• Leading through the standard partnership of minds and machines involves steering and coordinating collective intelligence while empowering individuals. It requires a balance of human judgment and machine analysis.

Here is a summary of the business model and cost structure for the company discussed:

• Udacity offers online programming courses that are project-based rather than exam-based. Students write and submit code to demonstrate their learning.

• Originally, Udacity employees evaluated student code submissions, which took an average of two weeks per submission.

• Udacity experimented with having outsiders evaluate code submissions instead of employees. They found outsiders could provide comparable quality feedback, often faster than employees.

• Outsourcers were paid a fraction (e.g. 30%) of what Udacity paid employees to evaluate submissions. This significantly reduced grading costs.

• By leveraging crowdsourced evaluators, Udacity was able to scale grading and provide rapid feedback to students at a lower cost than using employees.

• The business model is based on students paying to access project-based programming courses. The cost structure was reduced by crowdsourcing the grading of student project submissions.

In summary, Udacity uses a project-based course model and reduced grading costs by shifting from employees to crowdsourced evaluators. This enabled rapid scaling while lowering costs.

• Technological progress tests companies. The average lifespan of S&P 500 companies has fallen from 60 years in 1960 to less than 20 years today due to creative destruction in the digital era.

• But technology itself does not determine outcomes. Success depends on how people use technology and the values they embed in organizations. There are often multiple viable strategies and equilibria, not just one optimal balance.

• Individual decisions shape the path of history. No single mind has all the knowledge needed to make economic decisions. The free market channels dispersed knowledge through prices and property rights.

• Digitization raises challenges as automation threatens jobs. But it also creates opportunities for more people to contribute and share prosperity through new combinations of technology, skills, and assets.

• Shaping society’s use of technology emerges from all parts of society, not just government. Entrepreneurs invent new jobs combining technologies and skills to meet needs. Humans are better at this creativity than machines.

• More power through technology means values matter more than ever. We should ask not “what will technology do to us?” but “what do we want to do with technology?”

• New technologies like artificial intelligence are allowing machines to master complex tasks like the board game Go, which was long considered too intuitive for computers. This demonstrates the rapid advances being made in AI.

• Many major tech companies like Uber, Airbnb, Facebook, and Alibaba are disrupting traditional industries by owning the customer interface and leveraging data, algorithms, and AI. They are expanding quickly across the globe.

• Traditional companies are struggling to keep up with the pace of technological change. Even large companies like GE that invest heavily in R&D are having difficulty internally innovating at the rate of Silicon Valley tech firms.

• There is a “triple revolution” underway with digital, cheap computation, and the power of data/algorithms that is allowing tech firms to rapidly disrupt established industries.

• This revolution presents challenges for traditional companies and workers, but also great opportunities if the technology is steered toward inclusive growth rather than concentration of power. The choices societies make about how to use these technologies will shape the future.

• In 1993, business process reengineering (BPR) emerged as a hot new management trend, promising dramatic performance improvements by radically rethinking business processes. However, BPR failed to live up to the hype, with over 60% of projects failing to achieve the desired results.

• BPR relied too much on intuition and “gut feel” instead of data and evidence. Many executives thought they could intuitively redesign processes better than by analyzing data.

• Research in psychology and other fields shows that algorithms and statistical models consistently outperform human judgment and intuition across a variety of domains. Humans are prone to cognitive biases that algorithms don’t have.

• Computers have now demonstrated superior performance in areas thought to require human discretion and intuition, such as predicting wine quality, housing prices, academic impact, court rulings, and clinical outcomes.

• By combining machine learning with organizational change management, companies today can take a more data-driven approach to process redesign and achieve the performance improvements promised by BPR. Humans and machines together can reengineer business processes better than either one alone.

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

• Artificial intelligence (AI) is the field focused on creating intelligent machines that can simulate human cognition. Early AI systems in the 1950s-60s generated excitement but were limited in their capabilities.

• A major challenge for AI is equipping systems with common sense, something humans acquire through lifetime experience. Without common sense, AI systems can make absurd errors in reasoning.

• AI and machine learning have advanced significantly in recent decades. Key developments include neural networks, backpropagation algorithms for training neural nets, and deep learning architectures.

• Neural networks are modeled after the human brain, with interconnected nodes similar to neurons. They can recognize patterns when trained on large datasets.

• Backpropagation allows neural nets to adjust internal parameters during training to improve performance on a task. This enables learning.

• Deep learning uses neural nets with many layers, which can learn abstract representations of data at higher levels. Deep learning has driven breakthroughs in image recognition, speech recognition, and other areas.

• Advances in compute power, data, and algorithms have enabled AI systems to achieve human-level or superhuman performance on narrow tasks. But general intelligence across many domains remains elusive. AI still struggles with common sense and flexibly applying knowledge to novel situations.

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

• In 1986, a seminal paper by Rumelhart, Hinton, and Williams introduced the backpropagation algorithm, allowing neural networks to improve by propagating errors backward through the network layers. This enabled deep learning.

• Yann LeCun developed convolutional neural networks for image recognition in the 1990s. These form the basis for many deep learning image classifiers today.

• Deep learning began taking off around 2006 when Hinton published a fast deep belief net algorithm. Major tech companies like Google and Microsoft now invest heavily in deep learning.

• Deep learning is being applied in many industries, like agriculture, finance, retail, and robotics. For example, deep learning helps cucumber farmers in Japan, powers automated teller machines, enables cashier-less stores like Amazon Go, provides AI support chatbots, and allows robots to do things like cook food.

• Advances like faster computer chips, big data, and better algorithms have enabled recent breakthroughs in deep learning and AI. For instance, Microsoft researchers recently achieved human parity in conversational speech recognition.

• Experts predict a “Cambrian explosion” in robotics thanks to deep learning and other AI advances, with major impacts on jobs and the economy. Driverless cars and trucks are one example already on the roads.

Here is a summary of the key points about the toll of new technology on several industries in the mid-1990s to early 2000s:

• Cell phone adoption grew rapidly, from 13% of Americans in 1995 to over 90% by 2015. This disrupted industries like landline phones, pagers, pay phones, and more.

• Newspapers faced declining print ad revenue as readers moved online. From 2000-2013, print ad revenue fell 70%. Thousands of newsroom jobs were cut. Major papers like the Rocky Mountain News and Tucson Citizen closed.

• The music industry was disrupted by digital piracy and then digital downloads. Recorded music revenue dropped from $14 billion in 2000 to $7 billion in 2013.

• Shopping malls were hurt by the rise of online retail like Amazon. Mall construction peaked in the 1990s and then declined.

• Kodak, once dominant in film photography, struggled to adapt to the rise of digital cameras. Casio released an early consumer digital camera in 1995 for $900.

So in summary, many major industries faced significant disruption in a short period due to new digital technologies that changed how consumers accessed information, entertainment, shopping, and more. This led to the decline or collapse of many established companies and business models.

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

• The Rocky Mountain News, Denver’s oldest newspaper, published its last edition on February 27, 2009 after nearly 150 years in business. The newspaper was unable to survive the decline of the newspaper industry caused by the internet and lost circulation and advertising revenue.

• Many other print publications have also struggled or gone out of business in the digital age, including Penthouse, National Enquirer, Newsweek, The New Republic, and Playboy. Even publications once considered indispensible have declined.

• The music industry has also been disrupted by the internet and shift to digital distribution. Global sales of recorded music have dropped dramatically since 1999. Major music retailers like Tower Records have gone bankrupt.

• Shopping malls have declined as online shopping has grown. Many malls have closed and major mall owners like General Growth Properties have gone bankrupt.

• The long distance telecommunications industry has shrunk from $77 billion in revenue in 2000 to $16 billion in 2013 due to cell phones and internet calling.

• Radio advertising revenue has dropped from $20 billion in 2000 to $14 billion in 2010.

• Technological improvements like cheaper data storage and the “death of distance” enabled by the internet have fueled massive disruption of traditional industries. New internet companies like Craigslist and AppNexus have taken business from traditional media.

• Steve Ballmer and Microsoft initially dismissed the iPhone’s high price tag, not realizing the power of Apple’s platform business model.

• Apple subsidized the iPhone’s cost to attract app developers. The App Store proved hugely successful, generating over $6 billion in revenue for Apple by 2016.

• Complementors like app developers can add tremendous value to a platform. Smart platforms actively court complementors by making it easy and attractive for them to join.

• Facebook offered to pay publishers to post content, helping drive engagement on its platform.

• Google acquired Android in 2005 for just $50 million, one of its best deals ever. Android has grown to dominate global smartphone market share.

• In contrast, Microsoft’s failed Nokia acquisition cost it billions and put its Windows mobile OS efforts behind.

• Successful platforms pay complements, subsidize access, acquire key technology, and find other creative strategies to attract users and complementors.

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

• Some of the worst tech mergers and acquisitions include Nokia’s acquisition of Microsoft’s mobile business, AOL’s merger with Time Warner, and Verizon’s acquisition of Yahoo. These deals resulted in huge write-downs and loss of value.

• Nokia acquired Microsoft’s mobile business for $7.2 billion in 2013, hoping to catch up in smartphones. But Nokia ended up writing down almost $8 billion on the deal and Microsoft cut over 7,800 jobs related to the acquisition. It was the largest write-down in Microsoft’s history.

• AOL and Time Warner merged in 2000 in a $165 billion deal that was touted as a transformative combination. But the merged company struggled to capitalize on synergies between the businesses and the deal was seen as a colossal failure.

• Verizon acquired Yahoo in 2017 for $4.48 billion, but the deal ended up discounting Yahoo’s value after data breaches affecting billions of accounts. Verizon wrote down nearly the entire value of the deal.

• The article highlights how poorly strategic some of these large tech mergers and acquisitions have turned out, failing to create value and synergies as intended. The deals ended up being costly write-downs and disasters.

• Baidu, a major Chinese search engine, announced plans to invest $3.2 billion in online-to-offline services over 3 years. This is part of Baidu’s strategy to expand beyond search into newer areas like food delivery, travel booking, etc.

• The investments highlight how tech companies are trying to integrate online and offline experiences and bridge the gap between the digital and physical worlds.

• Baidu aims to leverage its huge user base and data to provide customized services connecting online users to offline options. It sees opportunities in local O2O services given China’s rising middle class.

• The move comes amid slowing growth in search advertising revenue, Baidu’s core business, as more Chinese users shift to mobile. Baidu hopes O2O services will provide new revenue streams.

• Critics argue it’s unclear if Baidu can successfully compete in these new areas against more established specialists. But Baidu believes its user data gives it an edge in targeting services.

• The investments in online-to-offline mark a major strategic shift for Baidu beyond its search engine roots into e-commerce and local services.

Here is a summary of the key points from the excerpts you provided:

• In a 2013 Nature Biotechnology paper, researchers found that solutions developed through open prize-based contests were comparable or superior to those developed by either academic or industrial experts. As an example, a solution for a computational biology problem called MegaBLAST, developed through a contest, outperformed a solution designed by academic experts.

• Karim Lakhani, a professor at Harvard Business School, notes that over 700 challenge contests on the Topcoder platform, the winning solutions have come from people without domain-specific expertise over half the time. Marginal participants, not domain experts, are often the most successful problem solvers.

• Drawing on many diverse perspectives and knowledge sets allows groups to be collectively intelligent and outperform individual experts on many tasks. Eric Raymond coined the phrase “given enough eyeballs, all bugs are shallow” to capture this idea in open source software development.

• New platforms like Amazon Mechanical Turk, Topcoder, and TaskRabbit allow organizations to access talent and expertise on demand for specific tasks. Kickstarter and other crowdfunding platforms let creators access funding from a broad base of supporters.

• The key insight is that for many problems and opportunities, leveraging collective intelligence and marginal perspectives is more effective than relying on centralized expertise. As Marc Andreessen notes, “One could argue that crowdsourcing is the most fundamentally new economic idea since the advent of socialism and capitalism.”

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

• On October 31, 2008, Satoshi Nakamoto published a paper proposing Bitcoin, a peer-to-peer electronic cash system that allows online payments to be sent directly between parties without going through a financial institution.

• Bitcoin operates by having a public ledger called the blockchain that records all transactions. New Bitcoins are generated through a process called mining, where computers compete to solve complex math problems in order to add new blocks to the blockchain. This both secures the network and releases new Bitcoins as a reward.

• In 2010, the first known Bitcoin commercial transaction occurred when a programmer paid 10,000 Bitcoins for two pizzas. As Bitcoin has grown, the value of a single Bitcoin has increased dramatically.

• Bitcoin faced early setbacks, such as the Mt. Gox exchange hack in 2014 that resulted in approximately $470 million worth of Bitcoins being stolen. However, Bitcoin has continued to gain more mainstream recognition.

• The blockchain technology behind Bitcoin has inspired new innovations utilizing decentralized public ledgers for purposes like tracking credentials and preventing fraud. Overall, Bitcoin represents a groundbreaking experiment in decentralized digital currency and peer-to-peer transaction networks.

Here is a summary of the key points about e bonds from the article by Cade Metz:

• In June 2015, Overstock.com announced it was issuing “cryptosecurities” called e bonds using blockchain technology. This was the first time a blockchain was used for securities trading.

• The e bonds were privately issued and only available to accredited investors. Overstock raised $5 million through the initial offering.

• In March 2016, Overstock issued public blockchain bonds, marking the first public securities issue documented on a blockchain.

• Using blockchain for securities issuance reduces settlement risk exposure by over 90% compared to traditional methods, according to Nasdaq.

• In 2016, Barclays conducted the first blockchain-based trade finance deal with an Irish company called Ornua. This demonstrated blockchain’s potential for trade finance.

• Overall, the examples show blockchain technology allowing for more efficient and lower risk securities issuance and trading compared to traditional methods. Companies like Overstock and Barclays are pioneering the use of blockchain for finance.

Here is a brief summary of key points about Amazon:

• Amazon is a major e-commerce company that sells products directly to consumers through its website and mobile apps. It has pioneered data-driven approaches in retail.

• Amazon uses algorithms and large amounts of data to provide personalized product recommendations to customers based on their browsing and purchase history. This helps drive additional sales.

• The company offers a mobile app with a barcode scanner that allows people to check prices and purchase items from Amazon while shopping in physical stores. This leverages Amazon’s vast product selection and low prices.

• Amazon adjusts product prices frequently based on factors like supply, demand and competitor pricing. Automated algorithms determine when and how much to adjust prices rather than humans making those decisions.

• Amazon Web Services provides cloud computing services to businesses. It pioneered the public cloud model and is the market leader in this space. AWS enables companies to rent computing power and data storage instead of owning their own servers.

In summary, Amazon has been at the forefront of using data, algorithms, and technology to transform major industries like retail and cloud computing. Its innovations have helped make it one of the most valuable and influential technology companies.

Here are a few key points about the development of the crowd and how it relates to the core:

• The crowd is a large, decentralized network of people that can be harnessed to perform tasks, generate ideas, or solve problems. It is often contrasted with the core, which is a centralized group that oversees and directs the crowd’s efforts.

• Early examples of crowds include markets and cities, which brought together large numbers of buyers, sellers, and other actors. The web enabled the formation of even larger online crowds.

• Companies and organizations can leverage the crowd in various ways, such as crowdsourcing innovation through open calls, funding projects via crowdfunding platforms, outsourcing microtasks through systems like Amazon Mechanical Turk, or tapping collective intelligence through prediction markets.

• The core faces challenges in effectively organizing the crowd, evaluating contributions, aggregating information, and handling issues like bias and bad actors. Principles like openness, partitioning tasks, incentives, and transparency can aid in productively working with the crowd.

• While the core retains a coordinating role, the crowd is capable of impressive feats of decentralized creation, such as open source software, Wikipedia, and blockchain systems. The boundaries between the crowd and core are blurring as technology enables greater decentralized organization.

• Properly harnessing the crowd can help the core access more ideas and resources than it could alone. But finding the right balance of crowd freedom and core coordination remains an ongoing challenge.

Here is a summary of the key points about successful platforms and open platforms from the passage:

• Successful platforms leverage network effects to grow rapidly. They bring together groups that benefit from interacting, like buyers and sellers. The more users a platform has, the more valuable it becomes.

• Open platforms allow outside developers to build complementary products and services. This expands the platform’s capabilities and makes it more useful to users.

• Open platforms can gain data that helps improve the platform. They also benefit from shared innovation as outside developers add their own ideas.

• Well-designed platforms make it easy for developers to build on top of them. Open APIs and clear guidelines support third-party involvement.

• Successful open platforms strike a balance between openness and control. They need to maintain consistency of user experience but also give developers freedom.

The key ideas are leveraging network effects for growth, enabling complements through openness, gaining data and innovation from third parties, providing good developer tools and APIs, and balancing openness with control over the platform.

Here are the key points about soning:

• Soning is a form of automated inventory management that uses AI and algorithms to track perishable inventory in real time. It helps businesses avoid overstocking or understocking perishables.

• Perishing/perishable inventory refers to products with a short shelf life, like food or flowers. Managing this inventory is challenging as it can expire or spoil quickly.

• Soning uses sensors, predictive analytics, and replenishment algorithms to optimize perishable inventory levels. It accounts for factors like historical sales, seasonality, promotions, etc.

• By providing real-time visibility into perishable stock levels, soning helps reduce food waste and improves profitability. Businesses can avoid lost revenue from spoilage and out of stocks.

• Overall, soning leverages AI and automation to solve the complexities of perishable inventory management. It gives businesses enhanced visibility, control, and efficiency over short-lived products.

Summary of key points about Pindyck, Pinker, and piracy:

• Robert Pindyck is an economist known for his research on uncertainty and risk in financial markets. The 196n reference likely indicates a footnote citing his work.

• Steven Pinker is a psychologist and linguist known for his books on language, cognition, and human nature. The 68n reference likely indicates a footnote citing his work.

• Recorded music piracy refers to the unauthorized copying and sharing of music, often enabled by peer-to-peer file sharing services like Napster in the early 2000s. The book discusses how this disrupted the music industry from 1999-2015, causing sharp declines in music sales.

• Together, these references indicate the book touches on economics, cognitive science, and the disruption caused by digital piracy in the music industry as examples relating to its key themes. The footnotes point to sources supporting claims made in the text about these topics.

Keya Li

Associate Professor: Andrew McAfee

I enjoyed and appreciated your thoughtful summary. A few additional points:

• The book highlights how platforms and other digital technologies are transforming business, the economy, and society in fundamental ways. Key themes include the shift from physical capital to intangible assets like data, algorithms, platforms, and business processes.

• Network effects and winner-take-all dynamics are accelerating the pace of change and concentration of power for leading platforms like Google, Facebook, Amazon etc. This raises important policy questions around privacy, security, competition and more.

• New business models like the sharing economy (Uber, AirBnB, etc) are disrupting traditional industries by leveraging underutilized assets. However, there are ongoing debates about appropriate regulations.

• AI and automation are transforming work, eliminating some jobs but also creating new ones. Managing this transition in a way that benefits all of society is a critical challenge.

• Overall, the book aims to provide a balanced perspective on the opportunities and challenges created by digital technologies, and encourage thoughtful debates on the best way forward. The future remains uncertain, but we must try to shape it in a responsible and ethical manner.

Let me know if you would like me to expand on any of these points! The book covers a wide range of important topics at the intersection of technology, business and society.