Start free trial
Trending New Genres Authors
EnglishEnglish
EspañolSpanish
简体中文Chinese
繁體中文Chinese (Traditional)
FrançaisFrench
DeutschGerman
日本語Japanese
PortuguêsPortuguese
ItalianoItalian
한국어Korean
РусскийRussian
NederlandsDutch
العربيةArabic
PolskiPolish
हिन्दीHindi
Tiếng ViệtVietnamese
SvenskaSwedish
ΕλληνικάGreek
TürkçeTurkish
ไทยThai
ČeštinaCzech
RomânăRomanian
MagyarHungarian
УкраїнськаUkrainian
IndonesiaIndonesian
DanskDanish
SuomiFinnish
БългарскиBulgarian
עבריתHebrew
NorskNorwegian
HrvatskiCroatian
CatalàCatalan
SlovenčinaSlovak
LietuviųLithuanian
SlovenščinaSlovenian
СрпскиSerbian
EestiEstonian
LatviešuLatvian
فارسیPersian
മലയാളംMalayalam
தமிழ்Tamil
اردوUrdu
Searching...
SoBrief
Genius Makers

Genius Makers

The Mavericks Who Brought AI to Google, Facebook, and the World
An AI breakthrough triggers a $44 million bidding war. What had the machines actually learned?
by Cade Metz 2021 384 pages
4.26
3k+ ratings
Listen
Immersive
V2.3
Amazon Kindle Audible
Try Full Access for 3 Days
Unlock listening & more!
Continue

Key Takeaways

Keep promising ideas alive long enough for conditions to change.

A rejected idea can wait. In 1958, psychologist Frank Rosenblatt demonstrated a Perceptron, an early learning machine that could identify simple patterns. Its limits were real, but publicity inflated expectations, and critics Marvin Minsky and Seymour Papert showed that the system could not solve certain basic problems. Funding and attention shifted elsewhere. Geoff Hinton kept working on neural networks through years when colleagues treated them as a dead end. In 2012, after his group showed that a neural network could recognize images with exceptional accuracy, Google, Microsoft, Baidu, and DeepMind competed to acquire his three-person start-up. Google paid $44 million.

The lesson is timing, too. The idea had not suddenly become new. Better algorithms, more data, and faster computers had finally made it useful. Hinton chose Google over a higher possible auction price because he valued a home for the research more than extracting every dollar.

Analysis

What's striking is how scientific rejection can be both corrective and distorting. Minsky and Papert identified genuine limits, but institutions treated those limits as a verdict on an entire approach. This resembles the difference between falsifying one model and ruling out a whole research program. Persistence matters, but it is not automatically wisdom: the useful habit is to preserve a testable possibility, then keep asking what changed in the evidence or tools. Hinton's story makes intellectual patience look less like stubbornness and more like disciplined attention to conditions.

Deep learning needed algorithms, data, and specialized computing together.

Three ingredients converged. Deep learning refers to neural networks with multiple processing layers that learn patterns from examples. In the 1980s, Geoff Hinton, David Rumelhart, and others helped make multilayer networks trainable through backpropagation, a method that adjusts internal connections using errors from prior attempts. For years, the approach remained limited by the size of available datasets and the speed of computers.

The hardware changed the pace. Graphics processing units, or GPUs, were built to render video-game images, but their ability to perform many calculations at once made them useful for training neural networks. In 2012, Alex Krizhevsky used GPUs and ImageNet's labeled photographs to build AlexNet, which dramatically outperformed competing image-recognition systems. In speech recognition, a neural network trained on more data also beat systems that had taken Microsoft years to refine. Compute did not replace ingenuity. It amplified it.

Analysis

This history is a useful corrective to invention stories centered on a lone genius or a single breakthrough. The advance depended on complementary resources: a learning method, usable data, affordable parallel computing, and engineers able to make them work together. Innovation research often describes such moments as convergence rather than creation from nothing. The caveat is that more compute can make experimentation faster without making the resulting system more reliable or fair. Capacity is an enabling condition, not a measure of social value.

Protect long-horizon research from institutions optimized for familiar wins.

A research refuge mattered. Neural-network researchers remained scattered across universities while symbolic AI, which encoded explicit rules, dominated much of the field. Hinton, Yann LeCun, and Yoshua Bengio helped sustain a small research community through workshops and shared work, especially in Canada. They pursued ideas that could take years to prove, even when prestigious labs and journals showed little interest.

Communities preserve options. Hinton's workshops brought together computer scientists, engineers, psychologists, and neuroscientists. Researchers could compare results rather than pursue the same problems in isolation. The group also popularized the label deep learning for networks with many layers, giving an old line of work a name that attracted attention. The story is not simply that outsiders were right and insiders were wrong. Skepticism and scarce funding can expose weak claims, but they can also make institutions overly dependent on the methods already in favor.

Analysis

The book shows that research capacity is a social structure, not just a budget line. A network of peers can retain specialized knowledge, recruit students, and help an unfashionable idea survive long enough to meet better tools. This is close to the logic of intellectual diversity: competing approaches reduce the risk that an entire field shares one blind spot. The challenge is to distinguish fruitful dissent from unfalsifiable loyalty. Protected research needs room to explore, alongside clear standards for evidence and a willingness to abandon claims that repeatedly fail.

Demand evidence on hard benchmarks, not persuasive demonstrations.

A benchmark can settle a dispute. For years, researchers questioned whether neural networks could recognize complex images. Hinton's group trained AlexNet on ImageNet, a large collection of labeled photographs used in an annual image-recognition contest. The system performed far better than the competing entries, turning a theoretical argument into a measurable result. Its success persuaded skeptics who had asked for evidence on a demanding shared test.

Tests still have boundaries. ImageNet helped establish that the approach worked for image classification, but it did not prove that a system understood the world as people do. Researchers debated whether tidy, labeled images reflected ordinary visual experience. The distinction matters: a benchmark can compare systems fairly while still measuring only a slice of the real task. Strong claims should specify what a system was tested on, how it performed, and what the test leaves out.

Analysis

Benchmarks are powerful because they make disagreement inspectable. Instead of arguing from intuition, researchers can compare performance under common conditions. Yet measurement creates its own incentives: teams optimize for the test, and success on a curated dataset can be mistaken for broad competence. This is a familiar problem in education and public policy, where a proxy becomes a target. The right response is not to reject benchmarks, but to use several, include unfamiliar cases, and keep the measured capability narrower than the marketing language.

Do not mistake a game champion for general intelligence.

Games offer a clear scoreboard. DeepMind trained systems to play Atari games through reinforcement learning, a method in which an agent improves by trying actions and receiving rewards or penalties. AlphaGo used neural networks and extensive self-play to defeat elite Go players. Its unusual move in the match against Lee Sedol revealed that training could produce strategies human experts had not considered.

A game is not the world. Games define the rules, goals, and feedback. A robot, doctor, or online assistant faces situations without such clean boundaries. A system that masters Go does not thereby learn medicine or common sense. DeepMind treated games as stepping stones toward more general capabilities; critics questioned how much those skills would transfer. The distinction is practical: impressive performance in a controlled environment proves real competence there, not universal intelligence.

Analysis

Games are unusually good scientific laboratories because they make progress legible. A score tells researchers whether a system improved, and self-play can generate experience at a scale unavailable to human competitors. But the same clarity hides a central difficulty: real environments rarely supply a single agreed reward. In economics and public administration, optimizing a target can distort the broader purpose it was meant to represent. Game mastery is therefore meaningful evidence about learning and planning, while transfer to open-ended life remains a separate claim requiring separate tests.

Treat training data as a design decision with human consequences.

Machines inherit their examples. Deep-learning systems learn statistical patterns from their training data, so gaps in those examples can become gaps in performance. Google Photos once labeled a Black user's photos of his friend as gorillas. At Clarifai, researcher Deborah Raji found that a face-recognition training set was overwhelmingly composed of white people, mostly men. A content filter trained on imbalanced examples also risked associating Black people with pornography.

Measurement exposed unequal error. Joy Buolamwini tested commercial face-analysis tools and found error rates far higher for darker-skinned women than for lighter-skinned men. In one study, Microsoft's error rate for darker-skinned women was about 21 percent, and IBM's was 35 percent. These systems were not neutral simply because they used mathematics. Who is represented, who labels examples, and which errors count as acceptable all shape the product.

Analysis

The key insight is that bias can enter through ordinary engineering decisions, not only through openly discriminatory intent. Dataset composition, labeling rules, and evaluation groups determine whose experience becomes legible to a model. This connects to measurement theory: categories and instruments do not merely record reality, they structure what can be seen. Broader representation can expose failures, though diversity alone cannot guarantee justice. High-stakes systems also need independent evaluation, a route to contest errors, and limits on uses where mistakes impose severe costs.

Build safeguards around misuse before deployment, not after a scandal.

Capabilities have double uses. Ian Goodfellow's generative adversarial networks, or GANs, use two neural networks in competition: one generates examples, while the other tries to distinguish them from real data. The approach produced increasingly convincing images, including fabricated faces. Related techniques made it possible to alter videos and voices, raising the risk that realistic media could be used to deceive. Researchers also showed that small, carefully chosen changes could fool image-recognition systems. A few stickers on a stop sign could confuse a model used by a self-driving car.

Intent is not a safeguard. Project Maven, a U.S. Defense Department effort, sought machine-learning tools to analyze drone footage. At Google, employee protests forced a debate over whether the company should contribute. Some researchers feared the technology could support targeting; others argued AI might save lives. The controversy revealed that deployment choices, contracts, and oversight shape consequences as much as the algorithms do.

Analysis

Dual-use technology makes ethics an engineering and governance problem at once. A tool can improve accessibility or creative work while also enabling fraud, surveillance, or military targeting. Waiting for a dramatic misuse event means the incentives and infrastructure may already be entrenched. At the same time, broad prohibitions can obscure differences among applications and prevent beneficial research. A more precise approach asks who can use a system, what harms are foreseeable, what controls are enforceable, and whether affected communities have power to challenge deployment.

Keep accountable humans in systems that shape health, war, and speech.

Automation does not settle responsibility. Google's system for detecting diabetic retinopathy analyzed retinal scans and, in a study, identified signs of disease more than 90 percent of the time. That could help reach patients in places with few eye doctors. But clinical use still depended on validation, regulation, and trust. DeepMind's health project in Britain showed a different risk: access to records for 1.6 million patients triggered privacy concerns, and a regulator ruled the data-sharing arrangement unlawful.

Judgment remains necessary. Social platforms used automated systems to flag terrorist propaganda and other harmful material, but people still reviewed many decisions. Machines struggled with context, such as distinguishing a violent video from a superficially similar clip. Human oversight is not a magic fix, especially when reviewers are overburdened or lack authority. The point is to make responsibility visible rather than treating an algorithmic output as an unanswerable verdict.

Analysis

High-stakes automation should be judged not only by accuracy but by the whole decision process: who supplied the data, who can inspect outcomes, and who bears the cost of mistakes. Medicine illustrates the potential for expanded access; privacy disputes show that a useful goal does not automatically justify every means. Human review can provide context, but it can also become a thin layer of liability protection. Genuine accountability requires empowered reviewers, appeal mechanisms, and institutions prepared to act when evidence shows harm.

Open tools can democratize research while strengthening platform power.

Sharing sped adoption. Google released TensorFlow, software for building and training machine-learning systems, as an open-source tool. Researchers and companies could use it without starting from scratch, which widened participation and encouraged shared improvements. The company also benefited: a larger TensorFlow community could generate talent, research, and demand for Google's cloud services and specialized chips.

Openness is not powerlessness. The deep-learning boom was also a race for scarce researchers, computing capacity, and data. Google, Facebook, Microsoft, and Baidu built labs, bought start-ups, and recruited leading scientists. China developed its own research capacity and national AI ambitions, while Google sought to promote its tools there. Open code spread capability, but control of infrastructure, cloud services, and large datasets still gave powerful firms strategic advantages.

Analysis

Open infrastructure can function as both a public good and a competitive strategy. Shared tools lower the cost of entry, support reproducibility, and let researchers build on one another's work. Yet the firm that maintains a popular platform can shape defaults, attract users, and convert ecosystem dependence into commercial leverage. The tension resembles standards-setting in telecommunications: interoperability helps everyone, but control over the standard carries influence. Openness should therefore be assessed not only by whether code is available, but by who controls compute, data, and governance.

Treat AGI as a contested wager, not an established destination.

The pioneers disagreed about the finish line. Demis Hassabis and Shane Legg built DeepMind around the aim of artificial general intelligence, or AGI: systems able to perform a wide range of intellectual tasks. Elon Musk warned that such systems could become dangerous, while Sam Altman and Ilya Sutskever treated their creation as a major goal. Gary Marcus argued that deep learning struggled with common sense, reasoning, and learning from few examples. Hinton, despite helping launch the field's modern rise, favored solving specific problems over building general-purpose machines.

Progress is real; predictions are uncertain. Deep learning transformed speech and image recognition, translation, and other tasks. But claims about when machines might match or exceed human intelligence remained speculative. The book presents both the ambition and the doubt, showing how belief can mobilize money and talent while blurring the line between demonstrated ability and projected future.

Analysis

The AGI debate combines technical uncertainty with a powerful institutional incentive: dramatic futures attract investment, talent, and attention. That does not make the ambition empty, but it does mean forecasts should be separated from demonstrated results. A useful standard is to ask what capabilities have been measured, how robustly they generalize, and what assumptions connect today's systems to tomorrow's claims. The debate also concerns values, not just timelines: whether general-purpose intelligence is desirable differs from whether it is technically possible.

Analysis

Genius Makers is both a history of an algorithmic revival and a study of how institutions decide what counts as progress. Its central narrative runs from Rosenblatt's early Perceptron, through decades of skepticism, to the convergence of multilayer learning, abundant data, and GPU computing. The dramatic protagonists are scientists, but the decisive actors also include students, research communities, executives, hardware teams, and institutions that supplied money and infrastructure.

The book's strongest structural insight is that technical success does not arrive in isolation. Neural networks became influential when neglected ideas met enabling conditions and were translated into products. Yet the same movement also changed who controlled research. Corporate labs offered resources and routes to deployment, while concentrating talent and computational power. Open publication widened access, but platforms and cloud infrastructure gave dominant firms new forms of leverage.

The story also complicates the usual contrast between optimism and fear. Narrow systems produced measurable advances in speech, vision, translation, games, and medical screening. Those successes did not establish general intelligence, but they created real consequences in privacy, labor, misinformation, surveillance, and military applications. The most immediate risks were often less cinematic than superintelligence: biased classifications, opaque decisions, and systems deployed without adequate consent or accountability.

As a historical account, the book is especially good at showing that scientific consensus can lag behind evidence, then swing toward overconfidence once the evidence becomes commercially valuable. Its caution is not that progress should stop, but that competence in one environment is easily mistaken for competence everywhere. The enduring question is therefore not simply whether machines can learn, but who defines the task, controls the data and hardware, benefits from deployment, and carries the cost when the system fails.

Summary in 30 Seconds
Deep learning took off when improved algorithms met large datasets and powerful GPUs. Strong benchmark results can establish a system's skill at a defined task, but do not prove broader understanding. Training data shapes performance and can encode unequal errors. Deployment choices affect privacy, safety, and accountability, so human oversight and safeguards matter. Open tools widen access, while control of data and computing still advantages major platforms. Claims about general intelligence remain uncertain.
Contains spoilers

Last updated:

Report Issue

Review Summary

4.26 out of 5
Average of 3k+ ratings from Goodreads and Amazon.

Genius Makers chronicles the development of artificial intelligence, focusing on deep learning and neural networks. The book profiles key figures like Geoffrey Hinton, Yann LeCun, and Demis Hassabis, detailing their struggles and breakthroughs. Readers appreciate the accessible writing, engaging personal stories, and insights into the AI industry's competitive nature. While some critics desire more technical depth or ethical exploration, most find it an informative and entertaining history of modern AI, suitable for both newcomers and those familiar with the field.

Your rating:
4.55
396 ratings
Want to read the full book?

Glossary

Perceptron

early learning network

An early neural network developed by Frank Rosenblatt in the 1950s. It learned to classify simple patterns by adjusting the importance of its inputs after errors. The Perceptron demonstrated that a machine could learn from examples, but its single-layer design could not solve some basic pattern problems, limiting its abilities.

connectionism

learning through connections

An approach to artificial intelligence that builds learning systems from many interconnected computational units, loosely inspired by networks of neurons in the brain. Instead of specifying every rule by hand, connectionist systems adjust their internal connections while processing examples. Neural networks are the best-known expression of this approach.

deep learning

multilayer neural learning

A form of machine learning using neural networks with multiple layers of calculations. The layers can learn increasingly complex patterns from data, such as combining image details into recognizable objects. In the book, the term helped revive interest in neural networks and became associated with their advances in speech, vision, translation, and other tasks.

backpropagation

error-based weight adjustment

A method for training a multilayer neural network. After the network produces an answer, backpropagation sends information about its error backward through the layers so the system can adjust the connections, or weights, that contributed to the result. The method helped make deeper networks trainable, although useful performance also depended on data and computing power.

convolutional neural network

network for image patterns

A neural-network design particularly effective for analyzing images. It processes local sections of an image to detect small patterns, then combines those patterns into larger features. Yann LeCun developed a prominent version inspired by the visual system, and related networks later powered image-recognition breakthroughs such as AlexNet.

reinforcement learning

learning by rewarded actions

A training approach in which an agent tries actions in an environment and adjusts its behavior according to rewards or penalties. DeepMind used it to teach systems to play Atari games and to improve AlphaGo through self-play. Games make the method convenient because success can be defined with clear scores, unlike many real-world tasks.

generative adversarial networks

dueling image generators

A system in which two neural networks train against one another. One network generates examples, such as images, while the other tries to distinguish generated examples from real ones. The generator improves by trying to fool its opponent. Ian Goodfellow proposed this approach, which helped produce increasingly realistic synthetic images.

adversarial attacks

inputs that fool models

Carefully chosen changes to an input that cause a machine-learning system to make an incorrect prediction, sometimes while leaving the input looking unchanged to a person. The book describes examples involving altered pixels and stickers placed on a stop sign. Such attacks matter when models guide consequential tasks, including vehicle perception.

artificial general intelligence

broad intellectual capability

A proposed form of artificial intelligence able to perform a wide range of intellectual tasks, rather than being limited to a particular function such as image recognition or playing Go. DeepMind and OpenAI made AGI an explicit ambition. The book emphasizes that its feasibility, timeline, desirability, and risks remain contested.

FAQ

What's Genius Makers about?

  • Focus on AI pioneers: Genius Makers by Cade Metz chronicles the journey of key figures in artificial intelligence, such as Geoff Hinton, Yann LeCun, and Demis Hassabis, highlighting their groundbreaking work in deep learning.
  • Historical context: The book provides a historical overview of AI, tracing its evolution from early concepts in the 1950s to modern applications, and detailing the challenges and skepticism faced by researchers.
  • Impact on society: It emphasizes the societal implications of AI advancements, discussing both the potential benefits and risks, and raising questions about the future of intelligence and technology.

Why should I read Genius Makers?

  • Insight into AI development: The book offers a deep dive into the minds of the innovators who shaped AI, providing an understanding of the scientific and personal struggles behind major breakthroughs.
  • Engaging storytelling: Cade Metz combines rigorous research with compelling storytelling, making complex topics accessible and engaging, with anecdotes and personal stories adding a human element.
  • Relevance to current issues: As AI continues to evolve, the book provides context for ongoing debates about technology's role in society, encouraging readers to think critically about AI's implications.

What are the key takeaways of Genius Makers?

  • Importance of collaboration: The book illustrates how collaboration among researchers, companies, and institutions was crucial for AI's success, highlighting the synergy between academia and industry.
  • Resilience in research: Many pioneers faced rejection and skepticism but persisted, leading to significant breakthroughs, underscoring the importance of resilience and belief in one’s ideas.
  • Ethical considerations: It emphasizes the need for ethical frameworks in AI development, raising awareness about the potential risks associated with unchecked AI advancements.

What are the best quotes from Genius Makers and what do they mean?

  • “It’s the best possible time to be alive, when almost everything you thought you knew is wrong.”: Reflects the transformative nature of AI and the rapid pace of change, encouraging readers to embrace new ideas.
  • “Success is guaranteed.”: Ilya Sutskever’s bold claim highlights the confidence researchers had in deep learning's potential, signifying the optimism that drove the AI revolution.
  • “You probably heard Elon Musk and his comment about AI causing WW3.”: Underscores serious concerns surrounding AI and its potential consequences, reminding of the ethical dilemmas accompanying technological advancements.

Who are the main figures discussed in Genius Makers?

  • Geoff Hinton: Known as the "godfather of deep learning," his work on neural networks laid the foundation for modern AI, pivotal in reviving interest in neural networks.
  • Yann LeCun: A key figure in developing convolutional neural networks, his contributions significantly advanced image recognition technology and AI research at Facebook.
  • Demis Hassabis: Founder of DeepMind, he aimed to create artificial general intelligence (AGI), with his leadership in the AlphaGo project showcasing deep learning's potential.

What is deep learning, as described in Genius Makers?

  • Neural networks: Deep learning is a subset of machine learning using neural networks with multiple layers to analyze data, learning from vast amounts of information to identify patterns.
  • Applications: The book discusses applications like image recognition, speech recognition, and natural language processing, transforming industries from healthcare to entertainment.
  • Revolutionary impact: Deep learning has enabled machines to perform tasks previously thought exclusive to humans, leading to significant breakthroughs and raising ethical questions.

How did the auction for DNNresearch unfold in Genius Makers?

  • Competitive bidding: Geoff Hinton auctioned his start-up, DNNresearch, to tech giants like Google, Microsoft, and Baidu, with bids escalating rapidly via email.
  • Strategic decisions: Hinton’s auction strategy demonstrated his understanding of market dynamics, allowing him to maximize the value of his research and talent.
  • Outcome: Google won the auction for $44 million, marking a significant moment in the tech industry and underscoring the growing importance of deep learning and AI research.

What challenges did AI researchers face, according to Genius Makers?

  • Skepticism and rejection: Pioneers like Hinton and LeCun faced skepticism from the scientific community, leading to periods of isolation and doubt.
  • Funding and resources: Securing funding was challenging, especially during "AI winters," requiring navigation through fluctuating interest in their work.
  • Ethical concerns: As AI technologies advanced, ethical considerations became crucial, with researchers grappling with potential consequences like bias and societal impact.

How did the rise of deep learning change the tech industry?

  • Shift in focus: Success in deep learning led companies to prioritize AI research, reshaping business strategies and evident in investments by tech giants like Google and Facebook.
  • New applications: Enabled breakthroughs in fields like healthcare and finance, with companies leveraging AI for predictive analytics and automation.
  • Talent acquisition: Demand for AI researchers skyrocketed, leading to intense competition among tech companies to attract top talent, creating a new landscape for researchers.

How does Genius Makers address the ethical implications of AI?

  • Bias in algorithms: Discusses instances of AI bias, highlighting the importance of diverse training data and ethical considerations in AI development.
  • Surveillance and privacy concerns: Explores AI's use in surveillance technologies and potential misuse, encouraging critical thinking about innovation and individual rights.
  • Call for responsible AI: Advocates for proactive AI ethics, urging researchers and companies to consider societal impacts, emphasizing transparency and accountability.

What is the significance of AlphaGo in the context of Genius Makers?

  • Milestone achievement: AlphaGo's victory over Lee Sedol marked a breakthrough, demonstrating deep learning and reinforcement learning's capabilities.
  • Public awareness: Brought widespread attention to AI, sparking interest and investment, serving as a catalyst for further research and development.
  • Implications for future AI: Techniques from AlphaGo can apply to other domains like healthcare, illustrating broader implications beyond games.

What are the future prospects for AI, according to Genius Makers?

  • Path to AGI: Explores the quest for artificial general intelligence, a machine capable of human-like intellectual tasks, with experts believing AGI is still years away.
  • Technological advancements: Discusses potential for new technologies like advanced microchips to accelerate AI development, leading to more capable systems.
  • Societal impact: AI's future will likely impact sectors like healthcare and education, encouraging consideration of both opportunities and challenges as AI evolves.

About the Author

Cade Metz is a technology reporter for The New York Times, covering artificial intelligence, driverless cars, robotics, virtual reality, and other emerging technologies. Previously, he worked for Wired magazine for eight years. Metz's extensive experience in technology journalism is evident in Genius Makers, his first book, which draws from hundreds of interviews conducted over nearly a decade. His writing style is praised for its accessibility and ability to make complex subjects engaging for a general audience. Metz's work provides valuable insights into the world of AI research and development, balancing technical information with human interest stories.

Download PDF

To save this Genius Makers summary for later, download the free PDF. You can print it out, or read offline at your convenience.
Download PDF
File size: 0.46 MB     Pages: 21

Download EPUB

To read this Genius Makers summary on your e-reader device or app, download the free EPUB. The .epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.
Download EPUB
File size: 1.51 MB     Pages: 28
Want to read the full book?
Follow
Listen
Home
Swipe
Library
Get App
Try Full Access for 3 Days
Listen, bookmark, and more
Compare Features Free Pro
📖 Read Summaries
Read unlimited summaries. Free users get 3 per month
🎧 Listen to Summaries
Listen to unlimited summaries in 40 languages
—
❤️ Unlimited Bookmarks
Free users are limited to 4
—
📜 Unlimited History
Free users are limited to 4
—
📥 Unlimited Downloads
Free users are limited to 1
—
Risk-Free Timeline
Today: Get Instant Access
Listen to full summaries of 26,000+ books. That's 12,000+ hours of audio!
Day 2: Trial Reminder
We'll send you a notification that your trial is ending soon.
Day 3: Your subscription begins
You'll be charged on Nov 24,
cancel anytime before.
Consume 2.8× More Books
2.8× more books Listening Reading
Our users love us
1M+ readers
Trustpilot Rating
TrustPilot
4.6 Excellent
This site is a total game-changer. I've been flying through book summaries like never before. Highly, highly recommend.
— Dave G
Worth my money and time, and really well made. I've never seen this quality of summaries on other websites. Very helpful!
— Em
Highly recommended!! Fantastic service. Perfect for those that want a little more than a teaser but not all the intricate details of a full audio book.
— Greg M
Save 62%
Yearly
$119.88 $44.99/year/yr
$3.75/mo
Monthly
$9.99/mo
Start a 3-Day Free Trial
3 days free, then $44.99/year. Cancel anytime.
Unlock a world of fiction & nonfiction books
26,000+ books for the price of 2 books
Read any book in 10 minutes
Discover new books like Tinder
Request any book if it's not summarized
Read more books than anyone you know
#1 app for book lovers
Lifelike & immersive summaries
30-day money-back guarantee
Download summaries in EPUBs or PDFs
Cancel anytime in a few clicks

Why one summary?

Scanner
Find a barcode to scan

⭐
⭐
⭐
We have a special gift for you
Open
38% off
DISCOUNT FOR YOU
⭐
⭐
⭐
$79.99
$49.99/year
Only $4.16 per month
Continue
2 taps to start, super easy to cancel
Settings
General
Widget
Loading...