​These are my short reviews of books I've read recently which I think are important to data science, machine learning, and the future of humanity in general.
​

If you like my reviews and are planning to purchase one of these books as a result, please consider supporting my work by following the links to Amazon in the pictures - I get about 5%.

---

The future

rationality

The Past

fiction

---

The Future

---

​Superintelligence
Nick Bostrom

This influential book is responsible for bringing the topic of existential AI risk a little closer to the mainstream. Praised by Elon Musk and Bill Gates, it lays out many of the possible dangers of AI and what we can start doing now to reduce the risk.

Even if you believe there's only a 0.1% chance that AI could be dangerous to life on earth, so many future beings could suffer for so long that the topic is of great importance. If we care about suffering no matter where it happens, we should similarly care about suffering no matter when ​it happens. 

Though the writing sometimes becomes obscure and mathematical, the main points are salient and terrifying. How can we program morality into a computer if most people and ethicists don't even agree themselves? What policies should we create to make the transition safe, and will we even have time if the takeoff is as fast as Bostrom's arguments suggest? He makes a convincing case that for the first time we are truly facing "Philosophy with a Deadline".
​

The Second Machine Age
Erik Brynjolfsson and Andrew McAfee

Two MIT professors pair up to give us a thorough look at the economic transformation we are currently undergoing. An initial summary of the technological breakthroughs of machine learning, plus a deep dive into exponential growth and Moore's law, sets the stage for how people and jobs will be radically affected.

The book shines especially because the human capital of the authors allows them to draw from interviews with countless experts in the field. They go beyond traditional economic theories by showing how GDP doesn't cover nearly enough of the things that we actually consider valuable.

​A healthy awareness of how these technologies can increase inequality culminates with sound policy recommendations that have no chance of passing in our current dysfunctional government. That doesn't make it any less urgent, though, to explore and discuss the perils related to job automation and displacement.

 

Machine Platform Crowd
Erik Brynjolfsson and Andrew McAfee

These two authors come together again for a book exploring the economics of machine learning and adding two additional drivers of this "triple revolution": the platform and the crowd.

Platforms like Facebook, Airbnb and Uber have grown so fast because they simply connect people instead of creating their own content or owning their own hotels or cars. Network effects cause them to become more valuable as more people join and other platforms in "second place" are worth less and less. 

The power and potential of the crowd to bring in diverse ideas is demonstrated by case studies about big companies increasingly relying on such a "grassroots" technique. 

A list of questions at the end of each chapter prompt the reader to analyze how they are using, or could use, these technologies in their current organizations. The machine learning section expands from the last book and argues convincingly, but soberingly, that whenever algorithms are available to make predictions, they should trump "expert" opinion. A must-read for tech startups in the 21st century.​


​

 

 Weapons of Math Destruction
Cathy O'Neil

If you consider yourself a data scientist, I would argue that, of all the books I review on this page,  this is the most important one you should read.

​The first reason I believe this is because this topic hasn't been covered nearly as much as the other more common ones. Secondly, many data scientists are directly, but unknowingly, harming vulnerable people with their models. It is our responsibility to learn about the potential harm we could be causing.

​From determining whether someone should stay in jail from their acquaintances’ behavior, to predictive policing that brings more cops to the same neighborhoods, to denying people jobs or firing them, to charging more for insurance from people’s background, to predatory advertisement that targets people in poverty.

​All these models are dangerous because they are obscure, ubiquitous, and create feedback loops that perpetuate the assumptions people make when making models. Well-argued and very revealing book, that, though a bit repetitive at times, is sorely needed. 

​Homo Deus
Yuval Noah Harari

The author of the critically-acclaimed Sapiens comes back with another great book about the future and the past. The book argues that we have are close to solving all the problems that have been on humanity's agenda since our beginning: hunger, disease, and violence. Today, more people die from overeating than starvation, more people die from chronic diseases than infectious ones, and more people die from suicide than from human violence. 

Assuming these concerns will soon be over, the whole book is about exploring what will be our next goals. To find our future agenda, he looks back to see what are the logical conclusions of what we have been doing. The speculations are riveting and Harari's writing style is as simple and enthralling as we've come to expect from him. 

Readers expecting a book full of content about to future should be aware they will be disappointed in that the majority of time is spent looking at the past. However, I agree with him that to understand the new "techno-religions emerging from Silicon Valley" as he calls them, we must look at the monotheistic and humanistic traditions they are replacing. The last chapter, The Data Religion, looks at where our "quantified self", led by algorithms, could lead. Highly recommended for any data scientist or technologist in general, especially if they are interested in living forever. 

Data and Goliath
Bruce Schneier


A book to truly remind you of the quote "If you're not outraged you're not paying attention." Schneier, an expert in security and cryptography, outlines all the different ways we are under constant surveillance. Our phones have GPS devices to track our movements, our web-browsing behavior is analyzed routinely, cameras springing up everywhere are starting to use facial recognition, metadata on phone calls is consistently evaluated by the government, and most of our emails, texts, and social media providers collaborate with the government while being compelled to hide the fact they are doing so.   

Schneier makes a convincing argument that we are heading towards a 1984 world where the government knows everything we are doing. For those who believe "nothing to hide if you're doing nothing wrong" justifications, he reminds us that in many cases our data is being stored indefinitely and future administrations might find our current beliefs criminal. He emphasizes all social justice movements that we take for granted now were resisted by the government at the beginning.

He spends the first few chapters outlining all the ways we are being tracked, both by the government and by corporations. He then makes a moral case that privacy is necessary for a healthy democracy and should be a right of all humans. In the end, he lists policy recommendations that may not all be feasible or justified but at least he argues we must start having this debate. I am now convinced this topic is of supreme importance and would recommend this to anyone who uses technology or cares about democracy.
​

Life 3.0
Max Tegmark

A very important book that summarizes many of the most important questions surrounding artificial intelligence. 

Tegmark addresses both the shorter term issues around AI, like automated weapon systems, jobs, and laws, as well as long term questions about humans cohabiting with artificial general intelligence. He evens goes further into the long term future, where the limits of physics can point to what an intelligence could do once our sun dies. 

 I really appreciated how he breaks down all the words that normally get tossed around without agreement about what they mean: intelligence, computation, memory, learning, and consciousness. 

To explain what the title aludes to, this is how Tegmark defines life's stages and how it has evolved in the last 4 billion years:

Life 1.0 - Life can't update hardware or software - biological evolution is the only way to improve

Life 2.0 - Life can update it's software, but not hardware - cultural evolution, allows for much faster change 

Life 3.0 - Life can update both it's software and hardware - technological evolution, where we're headed

Tegmark's insight as a cofounder of the Future of Humanity Institute provides an excellent inside view into how people are preparing for the future of life, and making sure we think about where we want to ultimately end up. Highly recommended. 

---

​Rationality

---

​Algorithms to Live By
Brian Christian and Tom Griffiths

A unique book combining cognitive science, self-help, computer science, game-theory, philosophy and linguistics to help us improve the decisions we make when dealing with time, space, and other people.

I loved this book because it helps the reader both find heuristics to make better decisions and learn computer science at the same time. Very clear explanations and analogies for technical topics like Big O notation are peppered throughout the book. 

Assuming the goal of monogamy, how many people should we date before we decide to get married? Optimal stopping can help us think through this question. What clothes should we keep more accessible our closet? The last recently used algorithm makes the most sense. How should we deal with a friend who doesn't keep their word? Exponential backoff can serve as a guide that seems more ethical than common "x strikes you're out" strategies.

It's either a self-help book for geeks or a compendium of simplified algorithms for the non-technical. Superb either way.

​

 

​Decisive​
Chip and Dan Heath

This straightforward and organized book is almost certainly going to improve the decisions you make in your life. Case studies, mnemonics and chapter summaries make sure the lessons stick. 

In a nutshell, most of our decision-making processes are sloppy, and our explanations really just rationalizations for our feelings. In a way, this book is a simplified version of Kahneman's classic "Thinking Fast and Slow."

Their WRAP mnemonic summarizes the book nicely. To avoid narrow framing, Widen your options. To overcome confirmation bias, Reality test your assumptions. To reduce the strength of our short-term emotions, Attain distance before deciding. To mitigate overconfidence, Prepare to be wrong. 

Examples from psychology intermixed with business cases grounds their ideas in practical and commonsense suggestions. Scientists seeking to decide what job to take and what to measure in the first place are sure to benefit from this book.
​
​

​The Outer Limits of Reason
Noson Yanofsky

A beautiful and humbling collection of the limits of the human experience. A rare blend of philosophy, mathematics, linguistics, computer science, and quantum physics that works very well together.

The way Yanofsky covers chaos theory, P and NP problems, and the halting problem is very valuable for computer scientists but perhaps a bit more technical than the average reader would prefer. The section on quantum-computing and its philosophical implications is mind-blowing and reminds us how we still mostly think in outdated Newtonian ideas.

From Zeno to Godel to Turing to Einstein, a whirlwind tour of discovery and the undiscoverable. Expand your horizons with this book, or at least be aware of what are the limits of that expansion.
​

​

SuperForecasting
Philip Tetlock


Tetlock is a political scientist that became famous for spending 20 years tracking predictions made by experts in different fields. The common headline in the media is that the average "expert" is as good as predicting the future as random chance. However, Tetlock writes this book to show how that average hides the fact that there were some "superforecasters" who routinely made good predictions. His goal is to make us all better at this.

This Highly accessible and entertaining description of superforecasters makes a  great argument for the dire need of evidence-based decision-making in general. It goes through many of the common biases we hold when we think about the future and the past.

Guidelines in the book are sure to help us make better decisions in the future: being skeptical, being Bayesian in our thinking, being self-critical, seeking out diverse and conflicting views, being willing to update our beliefs based on new evidence, and most importantly, measuring our predictions about the future. 

If we can't measure what we believe, we will never get feedback and improve our skills of looking at the future. Worse of all, hindsight bias will make us feel we're better at predicting than we actually are. I, for one, joined his Good Judgement Project to start tracking my views of the future and see how my biases influence my thinking. 

Everything is Obvious
Duncan Watts

Duncan Watts is a physicist turned sociologist that offers a unique and refreshing perspective on the importance of social networks. He powerfully reminds us how we are entering a new era for the social sciences. Never before have we been able to measure humans at the scale that digitization allows today. 

The results of his own experiments at Yahoo and Microsoft are inspiring and revealing, if a bit outdated by 2017. Constantly reminding us of the importance of social norms on our behavior, the book makes it hard for us to continue thinking our beliefs are somehow our own. 

He makes us question all the typical social science explanations that rely on the reasoning of agents. In a way, it's a more sophisticated, scientific, and nuanced version of popular books like Freakonomics and authors like Malcolm Gladwell.



​

​

---

 

​The Past

---

Sapiens
Yuval Noah Harari

An accessible and comprehensive history of our species. Harari excels in humbly acknowledging and presenting the many competing theories that seek to explain our history. This doesn't mean he's not opinionated - he's just very forthcoming about his biases. 

Even though every chapter is useful to understand the big picture of how technology has transformed humanity throughout our history, the section dedicated to the scientific revolution is the most relevant to machine learning. 

It illuminates how science and technology were seen as completely separate fields until relatively recently. Only after the industrial revolution did politicians and businesspeople start thinking about research as a way to advance their goals. It's a very important reminder of how science funding has almost always been a result of some powerful group's agenda, and not for the interest of pure research.

All science is part of an ideological or religious agenda. Science can tell us what happens or with great difficulty, maybe even why, but it can never tell us what should happen. To be responsible scientists, it is crucial to be aware, and to keep in mind, the uneasy historical partnership between science, imperialism, and capitalism. 

The Structure of Scientific Revolutions
Thomas Kuhn

Anyone who considers themselves a scientist would benefit from reading this classic book about the philosophy of science. Without this underlying perspective, scientists might fall into the trap of believing that what they have been taught is the "Truth". 

Kuhn uses examples in many different sciences to identify the phases that comprise a scientific revolution:

The pre-paradigm stage where there is no consensus among scientists. Someone like Aristotle then comes along and creates a paradigm that becomes a standard, and subsequent "normal science" proceeds to clarify and expand on this first universal model. The next phase in a revolution begins when anomalies start showing the predominant paradigm might not be correct. Though scientists may ignore them at first, eventually enough anomalies and the dying of older scientists allow for new competing paradigms to emerge. Eventually a new paradigm, like Newton's, takes the place of the old one and normal science begins again. Until Einstein upended everything again.

Keeping in mind this never ending change in all sciences let's us maintain an open mind to the possibility, in fact the certainty, of being wrong. This lets us be open minded to things that may not appear possible under our current model. If all past models have been wrong, it seems very naive to think that this time we are actually using a model that is "true". All models are wrong by design - some are just useful. Not an easy read, but a transformational one.
​

​Viral Loop
Adam Penenberg
​

An entertaining, insightful, and straightforward account of the history of various viral products and companies. A book that Alex Schultz, VP of growth at Facebook, says he gives to everyone who joins his team. A good introduction to start thinking about, and measuring, the "viral coefficient" of different products. 

Incredibly inspiring stories throughout the book show how quickly an idea can spread if the viral coefficient is high enough. Examples like "Hot or Not", Hotmail, Netscape, and Facebook are valuable not only to learn about  their strategies but also to learn about the history of many big players in Silicon Valley. Good rules of thumb about the underlying principles commonly found in viral projects would definitely benefit anyone trying to attract users to a platform or product. However, more than a set of guidelines, it's a book to inspire you and make you think big.



​

The Facebook Effect
David Kirkpatrick

​Kirkpatrick's interviews with Zuckerberg and many of the people that were at Facebook from the start give us an inside view of the company from the Harvard-dorm room it started in to 2010 when the book was published. 

Very easy to read, and difficult to put down, Kirkpatrick creates a gripping narrative of constant growth. He provides an inspiring and eye-opening account that still has the courage to ask the tough questions and expand on the missteps made by the company. The underlying issue of privacy is a common thread throughout the book. A much needed correction to the inaccurate and embellished 2010 movie "The Social Network".
 

---

 

​Fiction

---

​The Foundation Trilogy
Isaac Asimov


The whole plot of this science fiction classic centers around predictions made by "psychohistorians" - a future academic discipline which merges history, sociology and mathematics to map out the most probable future of humanity.

The math shows clearly that the galactic empire will inevitably fall and the 12,000 years of galactic rule will come to an end. The politicians' best hope to shorten the interregnum - the dark ages before a new government emerges- is to take create a planet with all of human knowledge. From this Foundation, a new empire will be born.

Predictions about the future, and how knowledge about those predictions affect the people who are being measured, makes this a fascinating read for data scientists. Cited by Elon Musk and Paul Krugman as an important guide that helped them choose their career paths.

​ 

The Circle
David Eggers

​The story of an employee who moves to Silicon Valley to work at a company that is a sort of Facebook-Google conglomerate so powerful it's starting to overshadow the government.

The story slowly builds as the company gains more power, becomes more intrusive, and eventually becomes terrifying. In a way this is the logical conclusion of many of the efforts currently being undertaken at the big tech firms. If we share more of ourselves, and our data is owned by these big companies, what is the role of government? Is privacy something that we should guard, or the last refuge of those who have things to hide? Many parts are reminiscent of Zuckerberg's naïveté about a completely "open" society, ignoring the fact that there still must be someone at the top deciding what information should be presented to whom.

Though the end may appear too far-fetched to occur, it's still a valuable cautionary tale to think about the Orwellian world some of today's beliefs, taken to their extreme, could lead to.