Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World

by Cade Metz

New York Times Silicon Valley beat reporter Cade Metz has an insider's perspective on the greatest tech story of our time--a story that no one else has been in a position to tell.

Long dismissed as unattainable, artificial intelligence was, for decades, ignored by the big tech giants and universities. The small community of AI researchers worked on the fringes of the scientific community. Then, in 2008, two chance meetings changed everything. The first was between a university researcher and a Microsoft programmer. The second was at Google, where another academic convinced a Google exec to pay closer attention. Within months both Microsoft and Google achieved breakthroughs that vastly improved their speech recognition and image recognition capabilities, respectively. And they both realized the same AI--real artificial intelligence--was far closer than they realized. The first company to achieve full AI would dominate the next several decades of tech. The race was on.

Google immediately launched an in-house artificial intelligence lab. They also went after the few AI researchers, both in academia and those running startups. Microsoft did the same but stumbled due to messy internal bureaucracies. Facebook took notice and quickly poached Microsoft's team. Then Elon Musk jumped in and founded OpenAI, stealing talent from both Google and Facebook. These companies began waging an unprecedented arms race--which now includes China and Russia.

The new AI tech breakthroughs not only drew the interest of companies but of governments, too. Metz's epic story dramatically presents the fierce competition among national interests, shareholder value, the pursuit of scientific knowledge, and the very human concerns about privacy, security, bias, and prejudice. This technology is already guiding our leisure time, our retail choices, our cars, our factories, our markets, and our military weapons. How far will we let it go?

Like a great Victorian novel, this world of eccentric, brilliant, often unimaginably yet suddenly wealthy characters draws you into the most profound moral questions we can ask. What does it mean to be smart? To be human? What do we really want from life and the intelligence we have or might create?

Genres: Artificial Intelligence, Nonfiction, Technology, Business, Science, History, Biography

360 pages, Kindle Edition

First published March 16, 2021

Reviews

[Will Byrnes · Rating 4 out of 5]

[In 2016] Ed Boyton, a Princeton University professor who specialized in nascent technologies for sending information between machines and the human brain…told [a] private audience that scientists were approaching the point where they could create a complete map of the brain and then simulate it with a machine. The question was whether the machine, in addition to acting like a human, would actually feel what it was like to be human. This, they said, was the same question explored in Westworld.

AI, Artificial Intelligence, is a source of active concern in our culture. Tales abound in film, television, and written fiction about the potential for machines to exceed human capacities for learning, and ultimately gain self-awareness, which will lead to them enslaving humanity, or worse. There are hopes for AI as well. Language recognition is one area where there has been growth. However much we may roll our eyes at Siri or Alexa’s inability to, first, hear, the words we say properly, then interpret them accurately, it is worth bearing in mind that Siri was released a scant ten years ago, in 2011, Alexa following in 2014. We may not be there yet, but self-driving vehicles are another AI product that will change our lives. It can be unclear where AI begins and the use of advanced algorithms end in the handling of our on-line searching, and in how those with the means use AI to market endless products to us.


Cade Metz – image from Wired

So what is AI? Where did it come from? What stage of development is it currently at and where might it take us? Cade Metz, late of Wired Magazine and currently a tech reporter with the New York Times, was interested in tracking the history of AI. There are two sides to the story of any scientific advance, the human and the technological. No chicken and egg problem to be resolved here, the people came first. In telling the tales of those, Metz focuses on the brightest lights in the history of AI development, tracking their progress from the 1950s to the present, leading us through the steps, and some mis-steps, that have brought us to where we are today, from a seminal conference in the late 1950s to Frank Rosenblatt’s Perceptron in 1958, from the Boltzmann Machine to the development of the first neural network, SNARC, cadged together from remnant parts of old B-24s by Marvin Minsky, from the AI winter of governmental disinvestment that began in 1971 to its resumption in the 1980s, from training machines to beat the most skilled humans at chess, and then Go, to training them to recognize faces, from gestating in universities to being hooked up to steroidal sources of computing power at the world’s largest corporations, from early attempts to mimic the operations of the human brain to shifting to the more achievable task of pattern recognition, from ignoring social elements to beginning to see how bias can flow through people into technology, from shunning military uses to allowing, if not entirely embracing them.


This is one of 40 artificial neurons used in Marvin Minsky’s SPARC machine - image from The Scientist

Metz certainly has had a ringside seat for this, drawing from hundreds of interviews he conducted with the players in his reportorial day jobs, eight years at Wired and another two at the NY Times. He conducted another hundred or so interviews just for the book.

Some personalities shine through. We meet Geoffrey Hinton in the prologue, as he auctions his services (and the services of his two assistants) off to the highest corporate bidder, the ultimate figure a bit startling. Hinton is the central figure in this AI history, a Zelig-like-character who seems to pop up every time there is an advance in the technology. He is an interesting, complicated fellow, not just a leader in his field, but a creator of it and a mentor to many of the brightest minds who followed. It must have helped his recruiting that he had an actual sense of humor. He faced more than his share of challenges, suffering a back condition that made it virtually impossible for him to sit. Makes those cross country and trans-oceanic trips by train and plane just a wee bit of a problem. He suffered in other ways as well, losing two wives to cancer, providing a vast incentive for him to look at AI and neural networking as tools to help develop early diagnostic measures for diverse medical maladies.


Marvin Minsky in a lab at M.I.T. in 1968.Credit...M.I.T. - image and caption from NY Times

Where there are big ideas there are big egos, and sometimes an absence of decency. At a 1966 conference, when a researcher presented a report that did not sit well with Marvin Minsky, he interrupted the proceedings from the floor at considerable personal volume.

“How can an intelligent young man like you,” he asked, “waste your time with something like this?”

This was not out of character for the guy, who enjoyed provoking controversy, and, clearly, pissing people off. He single-handedly short-circuited a promising direction in AI research with his strident opposition.


Skynet’s Employee of the month

One of the developmental areas on which Metz focuses is deep learning, namely, feeding vast amounts of data to neural networks that are programmed to analyze the incomings for commonalities, in order to then be able to recognize unfamiliar material. For instance, examine hundreds of thousands of images of ducks and the system is pretty likely to be able to recognize a duck when it sees one. Frankly, it does not seem all that deep, but it is broad. Feeding a neural net vast quantities of data in order to train it to recognize particular things is the basis for a lot of facial recognition software in use today. Of course, the data being fed into the system reflects the biases of those doing the feeding. Say, for instance, that you are looking to identify faces, and most of the images that have been fed in are of white people, particularly white men. In 2015, when Google’s foto recognition app misidentified a black person as a gorilla, Google’s response was not to re-work its system ASAP, but to remove the word “gorilla” from its AI system. So, GIGO rules, fed by low representation by women and non-white techies. Metz addresses the existence of such inherent bias in the field, flowing from tech people in the data they use to feed neural net learning, but it is not a major focus of the book. He addresses it more directly in interviews.


Frank Rosenblatt and his Perceptron - image from Cornell University

On the other hand, by feeding systems vast amounts of information, it may be possible, for example, to recognize early indicators of public health or environmental problems that narrower examination of data would never unearth, and might even be able to give individuals a heads up that something might merit looking into.

He gives a lot of coverage to the bouncings back and forth of this, that, and the other head honcho researcher from institution to institution, looking at why such changes were made. A few of these are of interest, like why Hinton crossed the Atlantic to work, or why he moved from the states to Canada, and then stayed where he was based once he settled, regardless of employer. But a lot of the personnel movement was there to illustrate how strongly individual corporations were committed to AI development. This sometimes leads to odd, but revealing, images, like researchers having been recruited by a major company, and finding when they get there that the equipment they were expected to use was laughably inadequate to the project they were working on. When researchers realized that running neural networks would require vast numbers of Graphics Processing Units, GPUs (comparable to the Central Processing Units (CPUs) that are at the heart of every computer, but dedicated to a narrower range of activities) some companies dove right in while others balked. This is the trench warfare that I found most interesting, the specific command decisions that led to or impeded progress.


Rehoboam – the quantum supercomputer at the core of WestWorld - Image from The Sun

There are a lot of names in The Genius Makers. I would imagine that Metz and his editors pared quite a few out, but it can still be a bit daunting at times, trying to figure out which ones merit retaining, unless you already know that there is a manageable number of these folks. It can slow down reading. It would have been useful for Dutton to have provided a graphic of some sort, a timeline indicating this idea began here, that idea began then, and so on. It is indeed possible that such a welcome add-on is present in the final hardcover book. I was working from an e-ARE. Sometimes the jargon was just a bit too much. Overall, the book is definitely accessible for the general, non-technical, reader, if you are willing to skip over a phrase and a name here and there, or enjoy, as I do, looking up EVERYTHING.

The stories Metz tells of these pioneers, and their struggles are worth the price of admission, but you will also learn a bit about artificial intelligence (whatever that is) and the academic and corporate environments in which AI existed in the past, and is pursued today. You will not get a quick insight into what AI really is or how it works, but you will learn how what we call AI today began and evolved, and get a taste of how neural networking consumes vast volumes of data in a quest to amass enough knowledge to make AI at least somewhat…um…knowledgeable. Intelligence is a whole other thing, one of the dreams that has eluded developers and concerned the public. It is one of the ways in which AI has always been bedeviled by the curse of unrealistic expectations.


(left to right) Yann LeCun, Geoffrey Hinton, Yoshua Bengio - Image from Eyerys

Metz is a veteran reporter, so knows how to tell stories. It shows in his glee at telling us about this or that event. He includes a touch of humor here and there, a lightly sprinkled spice. Nothing that will make you shoot your coffee out your nose, but enough to make you smile. Here is an example.

…a colleague introduced [Geoff Hinton] at an academic conference as someone who had failed at physics, dropped out of psychology, and then joined a field with no standards at all: artificial intelligence. It was a story Hinton enjoyed repeating, with a caveat. “I didn’t fail at physics and drop out of psychology,” he would say. “I failed at psychology and dropped out of physics—which is far more reputable.”

The Genius Makers is a very readable bit of science history, aimed at a broad public, not the techie crowd, who would surely be demanding a lot more detail in the theoretical and implementation ends of decision-making and the construction of hardware and software. It will give you a clue as to what is going on in the AI world, and maybe open your mind a bit to what possibilities and perils we can all look forward to.

There are many elements involved in AI. But the one (promoted by Elon Musk) we tend to be most concerned about is that it will develop, frighteningly portrayed in many sci-fi films and TV series, as a dark, all-powerful entity driven to subjugate weak humans. This is called AGI, for Artificial General Intelligence and is something that we do not know how to achieve. Bottom line for that is pass the popcorn and enjoy the show. Skynet may take over in one fictional future, but it ain’t gonna happen in our real one any time soon.


Review posted – April 16, 2021

Publication dates
----------Hardcover - March 16, 2021
----------Trade Paperback - February 15, 2022


I received an e-book ARE from Dutton in return for…I’m gonna need a lot more data before I can answer that accurately.



This review has been cross-posted on my site, Coot’s Reviews, all in one piece. Stop by and say Hi!

==========In the summer of 2019 GR reduced the allowable review size by 25%, from 20,000 to 15,000 characters. In order to accommodate the text beyond that I have moved it to the comments section directly below.

[Max · Rating 4 out of 5]

Metz traces the development of AI, focusing on artificial neural networks which simulate the way the brain processes information. The current iteration, called deep learning, learns from data to identify and produce images and speech, translate languages, drive cars, perform medical diagnostics, beat chess, go and Jeopardy champions and much more. Attempts at neural network AI began in the 1960s with the Perceptron which learned to analyze images although its capacity was very limited. Many twists and turns followed before landing on the current technology of deep learning. But this is more than a technology story, it’s also a story of the developers, their personalities and quirks. These were people driven to make neural network AI a useful technology. We meet people like Geoff Hinton who Metz calls the “founding father of the deep learning movement” along with other top people in the field such as Yann LeCun and Demis Hassabis and many more who made significant contributions to deep learning. Metz describes their small tight knit global community as they fed each other new ideas culminating in the dazzling capabilities of today. My notes follow.

By 2012, deep learning was recognized as necessary to achieve dominance in AI and the few people who really understood it were at a premium. Google bought Geoff Hinton’s company bringing him, Ilya Sutskever and Alex Krizhevsky on board. These three would transform Google’s AI program. Google quickly took the lead launching its deep learning speech engine on its Android phones in 2012 surpassing the capabilities of Apple’s Siri. Goggle’s technology spread quickly. In 2014 Amazon would release Alexa. Baidu, Microsoft and Facebook would also follow with personal assistants. Google spent hundreds of millions of dollars to build a state-of-the-art AI research lab, $130 million just for GPU chips. This was not only to speed product development but to attract top researchers. The race to acquire AI talent was intense. Company leaders like Google founders Larry Page and Sergey Brin and Facebook founder Mark Zuckerberg got personally involved. Zuckerberg invited top researchers to his home for dinner and in 2013 hired Yann LeCun to lead a new AI research lab. In 2014 Google spent $650 million to acquire the AI company DeepMind led by world leading AI researcher Demis Hassibis.

In 2014 Ilya Sutskever working at the Google AI lab, Google Brain, developed the technology that would power efficient language translation, a much more complex task than recognizing an image or word. In translation many varying sequences had to be related, for which Sutskever developed a system of vectors, touted as “a breakthrough for any AI problem that involved a sequence.” It would take eighteen months to turn this method into Google’s translation service that would be relied on by millions of people. Alex Krizhevsky took Google’s stagnant driverless car program and applied deep learning to it. The result was the spinoff company Waymo. To prove the power of deep learning, Google used its acquisition DeepMind led by Demis Hassibis to develop AlphaGo to take on the world’s go champion. Go has many more possible moves than chess and both Sergey Brin and Hassibis were fans. In 2016 AlphaGo beat the player of the decade Lee Sedol and in 2017 World champion Ke Jie in China. Two months later China announced its plan to be the world leader in AI by 2030. Meanwhile Google’s DeepMind was also developing technology to read MRI, CT and eye scans to diagnose diabetic retinopathy and a host of other diseases. It’s first effort, DeepMind Health, working with the NHS, was stopped by British regulators concerned by sharing of personal health information.

In 2014, Ian Goodfellow joined Google bringing with him an idea called generative adversarial networks or GAN. He became known as the GANfather. In GAN one neural network learned from another. If one created an image that was supposed to look real, the other would look for something wrong with it. This back and forth would go through countless iterations until an image was produced that couldn’t be distinguished from real. With the election of Trump in 2016 and the avalanche of misinformation, GANs were looked at in a new light. They were the perfect vehicle to create pictures, even of people that didn’t exist, that could not be distinguished from real. Pictures were no longer proof. Conversely researchers increasingly found ways to fool AI algorithms. Deep learning AI learns from the data it is fed, typically real-world data. For a driving application, countless street scenes showing diverse activity are fed in. But, for example, in one test researches fooled AI by sticking several post-it notes on a stop sign. AI doesn’t learn the way humans learn. Serious flaws may remain hidden. After 2016 Goodfellow redirected his efforts to AI security.

Another problem that surfaced in AI was discrimination against minorities and women. Keeping in mind that deep learning is taught by feeding it data such as pictures for image including facial recognition, who is selecting and labeling the data matters. Seems it’s predominantly white males. An incident where a Google photo categorization service labeled a picture of a black woman a gorilla made headlines. There is also the problem of the use of AI for military purposes. Undoubtedly autonomous weapons will be developed and almost certainly already have. Many if not most of the AI researchers leading development are opposed to this use. In the U.S. which contracts private firms for weapons development, this opposition may slow military use. But in countries like China and Russia opposition isn’t tolerated.

In 2018 Google introduced its personal assistant that could call a restaurant or hair salon to make a reservation or appointment. It sounded completely human and could handle many nuances in conversations with whoever answered the phone. The useful but limited scope of the product raised questions about the future of AI. The holy grail is general intelligence (AGI) that can be used in any situation the way humans do. But the most immediately productive path seems to be developing niche applications slice by slice. As robust as deep learning is, it still relies on being fed data to learn. This approach is not seen as conducive to AGI. Breakthroughs will need to be made. Therein lies a conflict. Companies want deliverables and want their teams focused on them, while many of the very best researchers are only interested in finding that breakthrough leading to AGI.

In 2018 Google released BERT, a “universal language model.” These massive neural networks learn language from reading enormous amounts of text written by humans. BERT read everything on Wikipedia and entire libraries of books of every genre. It listened to countless dialogues between people and digested thousands of questions and answers. A system using BERT passed a twelfth-grade science test. Bert and its kin were the next step forward and could greatly improve the bot world, but they were still far distant from AGI. In 2019 OpenAI, an independent lab cofounded by Elon Musk, demonstrated a robotic hand that could solve Rubik’s Cube, even with two fingers tied together and wearing gloves. The race to find a robotic solution to warehouse picking was on. Amazon was keenly interested. The savings would be enormous, although it would be at the expense of many thousands of workers. Pieter Abbeel, OpenAI’s prize roboticist, left the company in 2018 and with his former students formed a new company, Covariant. Geoff Hinton and Yann LeCun were convinced by what they saw and invested. In 2019 A German electronics company installed Covariant’s system. It picked and sorted ten thousand different components with 99% accuracy. One robot took the jobs of three warehousemen. Other picking companies were following suit. Automated picking and related technologies will be increasingly adopted, although how fast is not clear.

This is a great read for those interested in AI, how it got here, and where it’s headed. It is not technical. AI is described in general terms. Metz goes into more detail about the developers and Silicon Valley executives than he does explaining computer science. It’s fascinating to see how one idea led to another, how the big tech companies scrambled to take the lead or just stay even, and how quickly new ideas led to significant new products. AGI may be in the uncertain future, but based on what has transpired in the last decade, we can expect more breakthrough technologies and game changing products in the decade to come.

[Moritz Mueller-Freitag · Rating 4 out of 5]

How does it feel to see your life’s work go up in smoke? In the early 2000s, the computational linguist Chris Brockett had a sudden panic attack when he realized that a new crop of machine learning methods would make his research obsolete. The anxiety set in when it dawned on him that he had wasted nearly seven years of his life writing down linguistic rules for natural language processing. His colleagues thought he was having a heart attack and rushed him to the hospital. “My fifty-two-year-old body had one of those moments when I saw a future where I wasn’t involved,” he later reflected.

Many AI researchers experienced a similar shock in 2012 when Geoff Hinton and two of his grad students showed that deep neural networks could beat state-of-the-art AI systems in image recognition. Hinton belonged to a small group of academic contrarians – the “neural network underground” – who bet their careers on a concept that was long dismissed as a technological dead end. “Neural networks are for people who don’t understand stats,” they were told. But Hinton’s gang had the last laugh – much to the dismay of their detractors who had invested themselves in “shallow learning” methods.

Progress, of course, didn’t stop with image recognition. Since 2012, neural networks have achieved similar breakthroughs across previously intractable problems, ranging from machine translation and voice synthesis to solving the conundrum of protein folding. These advances have changed the technology industry in profound ways and set off a global arms race for top AI talent. It has also led to a fundamental shift in how software is being developed: instead of programming software by writing explicit instructions, we now increasingly train software by showing labeled examples. The new mantra is to throw just enough training data at a problem until it’s solved. I’ve witnessed this shift myself over the years when I co-founded a company with one of Hinton’s former doctoral students.

Cade Metz’s new book, Genius Makers, chronicles the AI miracles of the past decade from the vantage point of its creators. It’s a very readable and informative history of modern AI aimed at a general audience. The great strength of the book is that it avoids the common pitfall of tipping into hyperbole. Instead, it reminds us that technology always reflects the values, biases, and incentive systems of its makers. Although the narrative holds few groundbreaking revelations for people who are active in the field, it’s still fun to read about a subject when you’ve met many of the key protagonists in the flesh. And let’s be honest: Hinton’s oft-quoted wry sense of humor is worth the price of admission alone.

[Patrick Pilz · Rating 5 out of 5]

I think Cade Metz writes an important book here. As a top journalist, he covers in this latest book the the story of the people who made Artificial Intelligence what it is today. This is rather somber reporting, in which Cade Metz just lays out the facts along with condensed memoirs of all the main actors who brought us to where we are today. His writing is stellar and the journey interesting.

Most importantly, Cade choses to keep the technical details in the background, which makes this book very accessible for anyone with any background. It does an ok job on balancing the rewards and benefits while also outlining some dangers and limitations. You can certainly tell that he is more in the camp of proponents of AI, but he is not ignorant of the risks either.

All in all a book that deserves top spots on the non-fiction bestseller lists, just like "Tools and Weapons" by Brad Smith and Kai-Fu Lee's "AI Superpowers", probably the great read of the year on this subject.

[Mal Warwick · Rating 4 out of 5]

A closely-linked network of several score brilliant men and a few women are pushing the boundaries of artificial intelligence research. You’ll meet many of these high-achieving and sometimes eccentric individuals in the pages of Genius Makers. You’ll get a glimpse inside Google, Facebook, Baidu, and other major institutions where most of the cutting-edge AI research is underway. And in these pages, you’ll gain perspective on the issues and uncertainties that trouble this rarefied community. In a more general sense, Genius Makers will also show how the shifting currents of peer pressure influence the course of scientific research.

One approach among many

The principal theme in this important new book is the emergence of the promising approach to artificial intelligence that has become dominant in the field. Called deep learning, it’s grounded in artificial neural networks, which are loosely modeled on the human brain. In a neural network, scientists link together units or nodes called artificial neurons. The patterns they form allow the machine to learn from experience in a way analogous to learning in humans. Scientists “train” neural networks by exposing them to massive amounts of data. For example, to “teach” a neural network to recognize cats, they might feed it millions of images of cats. In the process, the neural network acquires an accurate enough picture of cats that it’s able to produce a credible cat image of its own. It doesn’t “understand” cats, but it will recognize an image of one.

The emergence of deep learning

For decades, deep learning had few adherents in the sixty-year-old field of artificial intelligence. A competing approach called “symbolic AI” held sway. “Whereas neural networks learned tasks on their own by analyzing data, symbolic AI did not.” Only a handful of maverick scientists stubbornly persisted through the dark ages beginning in the 1970s before more powerful computers allowed their work on neural networks to live up to its promise. Suddenly, the barrier in AI research was broken. The key was an important peer-reviewed article in 2012. It was “one of the most influential papers in the history of computer science,” attracting more than 60,000 citations.

The high-profile events that have brought AI to the world’s attention in recent years are all based on deep learning. For example, the defeat of the world’s top chess masters and Go champions. The increasing facility of machines in understanding spoken language. The advances made in self-driving cars. And the now-widespread use of face recognition. In Genius Makers, New York Times technology reporter Cade Metz profiles the scientists who made all this possible—for good or ill.

Seven key players

In an appendix labeled “The Players,” Metz lists sixty-one of the characters whose names appear in the book. Seven of these—all men—play central roles in the drama, but one stands out above the others. I’ll start with him, then list the other six in alphabetical order by last name.

Geoff Hinton

British-Canadian cognitive psychologist and computer scientist Geoff Hinton (born 1947) Is the grand old man of the researchers profiled in Genius Makers. If anything, he is the central figure in this story, the founding father of the deep learning movement. As Metz puts it, “Hinton and his students changed the way machines saw the world.” It was he who stubbornly continued to advocate for the use of neural networks in developing artificial intelligence in the face of near-universal disapproval within the field. A 1969 book by MIT legends Marvin Minsky and Seymour Papert was the cause. The book savagely attacked AI research using that approach and turned the tide against it for decades.

In Metz’s words, Hinton is “the man who didn’t sit down.” A back injury had prevented him from sitting for seven years when Metz arrived to interview him in December 2012. And Metz describes the elaborate arrangements Hinton must make when he travels. It’s quite remarkable that the man could function at all.

Hinton teaches at the University of Toronto. He joined Google in 2013 but lives and continues to work with his students in Canada. Hinton received the 2018 Turing Award, together with Yoshua Bengio and Yann LeCun, for their work on deep learning. He is the great-great-grandson of logician George Boole (1815-64). Boole’s work in mathematics (Boolean algebra) much later helped ground the new field of computer science.

Yoshua Benguio

French-born Yoshua Benguio (born 1964) is a computer science professor at the Université de Montréal. Along with Geoff Hinton and Yann LeCun, Benguio advanced the technology of artificial neural networks and deep learning in the 1990s and 2000s when the world’s AI community had turned their backs on the technique.

Demis Hassabis

Demis Hassabis (born 1976) is, in Metz’s words, a “British chess prodigy, game designer, and neuroscientist who founded DeepMind, a London AI start-up that would grow into the world’s most celebrated AI lab.” DeepMind was acquired by Google in 2014. Hassabis and his team at DeepMind developed the extraordinary AI named AlphaGo. In 2016, AlphaGo beat Lee Sedol, the world’s champion at Go, which many consider the world’s most difficult game. But the AI’s programming wasn’t limited to playing games. In 2020, DeepMind made significant advances in the problem of protein folding, expanding the boundaries of AI research in medical science.

Alex Krizhevsky

Alex Krizhevsky, born in Ukraine and raised in Canada, was a brilliant young protégé of Geoff Hinton at the University of Toronto. He played a leading role in developing computer vision, which is now central to face recognition and numerous other applications. Krizhevsky was one of Hinton’s partners in a startup they sold to Google in 2013. Together with Geoff Hinton, he “showed that a neural network could recognize common objects with an accuracy beyond any other technology.” Krizhevsky joined Google Brain and the Google self-driving car project but left the company in 2017. His work is widely cited by computer scientists.

Yann LeCun

Yann LeCun was born in France in 1960 but for decades has worked in the United States, first at Bell Labs and then at New York University, where he holds an endowed chair in mathematical sciences. In addition to teaching in New York, he also oversees Facebook’s Artificial Intelligence Research Lab. Like others portrayed in these pages, LeCun long collaborated with Geoff Hinton on deep learning before signing up with Facebook.

Andrew Ng

Andrew Ng, born in Britain in 1976, is an adjunct professor at Stanford University with a long colorful history in machine learning and AI. He co-founded Google’s deep learning research team Google Brain; managed Baidu‘s Silicon Valley lab as the Chinese company’s chief scientist; co-founded the pioneering MOOC (massive open online course) company Coursera, through which he taught more than 2.5 million students online. And since 2018 he has run a venture capital fund that backs startups in artificial intelligence.

Ilya Sutskever

Canadian computer scientist Ilya Sutskever, another of Geoff Hinton’s brilliant young protégés, gravitated with him to Google Brain when they sold their startup company to the Silicon Valley giant. But he left Google to join OpenAI, an AI lab in San Francisco backed by Elon Musk to compete with Google’s London-based DeepMind. Sutskever has made important contributions to the field of deep learning, among them co-inventing AlphaGo.

About the author

Cade Metz covers AI research, driverless cars, robotics, virtual reality, and other emerging technologies for the New York Times. He had earlier worked for Wired magazine. Genius Makers is his first book. James Fallows’ review of the book, along with a second one on AI by another Times reporter, recently appeared in the paper’s Sunday Book Review. Fallows explains, “Much of Metz’s story runs from excitement for neural networks in the early 1960s, to an ‘A.I. winter’ in the 1970s, when that era’s computers proved too limited to do the job, to a recent revival of a neural-network approach toward ‘deep learning,’ which is essentially the result of the faster and more complex self-correction of today’s enormously capable machines.”

In the notes at the conclusion of the text, Metz describes the research he conducted in writing Genius Makers. “This book is based,” he writes, “on interviews with more than four hundred people over the course of the eight years I’ve been reporting on artificial intelligence for Wired magazine and then the New York Times, as well as more than a hundred interviews conducted specifically for the book. Most people have been interviewed more than once, some several times or more.”

[Hugh · Rating 3 out of 5]

Like an ultra-modern take on Isaacson's The Innovators, this book is full of interesting characters doing very interesting things. Full of approachable science and theory, and very readable.

I found it a bit bland, as though Metz was reluctant to really engage with the controversies around AI. He writes about Cambridge Analytica, the inherent racism in facial recognition, and militarization of the tech, but seems to not want to assign any responsibility. Almost as if the author was afraid to lose access to the people he's writing about

A good book and a very thorough introduction to the science and history in modern AI research, and a surface-level discussion of the implications of the tech in society.

[Rishabh Srivastava · Rating 4 out of 5]

This is a phenomenal overview of the evolution of machine learning in the last 3-4 decades. Would recommend to anyone who is in the field or is interested in it.

My main takeaways were:

1. A lot of good ideas are ridiculed until they get blindingly obvious. When you know you’re on to something, stick with it even when others won’t. Yann LeCun, after his network for recognising digits was implemented in the real world, was then asked by AT&T to work on other more profitable projects. He refused. “I don’t care what you want. I want to work on computer vision”

2. You need to be a good marketer of your ideas to get them traction. Both in terms of removing previous stigma (titling papers deep learning instead of neural networks because neural networks had a stigma attached to them at the time). And to attract good talent

3. Find people who cover for your weaknesses. Geoff Hinton was more interested in ideas than in hardcore math or CS. So he found collaborators or graduate students who could take care of that

4. Ask people who aren’t believers, “what would convince you?”

5. “AI that plays games or works with video makes for a very good demo. Demos sell software, and sometimes companies”

6. Be mindful of the tribes you recruit from. Even as deep learning began to have inroads, a lot of Microsoft researchers didn’t believe in it because they harboured biases — pointing out issues like black box decision making despite its efficacy

7. Yoshua Bengio: “The robots of the future would need to sleep. They would need to sleep because they would need to dream”

[Ferhat Elmas · Rating 4 out of 5]

Well-written thriller-like deep learning history you would say I should have read this before.

[Abhilash · Rating 3 out of 5]

It's hard to write a review of a non-fiction book. It's always a mismatch of expectations and reality.

It's a good history book about AI from both academic and corporation pov. It covers almost everything. But it doth not offer insights or make predictions. The author is a journalist and hence he never planned to or make claims about the path AI is to take.
If you are excited about AGI, this book brings you back on the ground.

Microsoft's response to AI vs that of Google and Facebook comes out really well in this one. Also, covers China's plan to dominate AI by 2030 and it's scary.

[Aaron Deutsch · Rating 2 out of 5]

Interesting topic, I wish it had been more technical and less of a story. The flow was strange and a bit hard to follow. It often jumped around different time periods, companies, and/or topics without connecting them. The book gets in it’s own way by trying to mention every single scientist and engineer responsible for every development which just ends up feeling exhausting. Would’ve preferred if it focused on just a handful of main players.

[pugs · Rating 4 out of 5]

edit: i kept thinking of this simpsons quote while reading. -- Mr. Burns: I'll keep it short and sweet. Family, religion, friendship. These are the three demons you must slay if you wish to succeed in business. -- exemplified by musk and the like. carry on.

a lot of history here, i knew almost none of it, looking back. i heard the names tossed around over the years, but this overview goes to show just how competitive the tech field is, and how so much of it is bullshit. there's a lot of white supremacist, patriarchal machismo getting in the way of actually expanding ai, i.e., capitalism getting in the way of itself. especially when limiting immigration. so many strange people in this field. ego, oddity, and capital will continue to clash, and china will come out on top. that's probably for the better. the book completely neglects the fact nothing is truly revolutionary unless - everyone - has access to it, not just the select few. 'genius makers' is a puff piece for the ai industry, but seeing as there's no other book of its type, detailing the field's history (from a western perspective), it gets an extra star. did i mention how weird these people are?

[Florin Pitea · Rating 5 out of 5]

Excellent. Highly informative. Very well documented. Extremery relevant. Recommended.

[Raghu · Rating 3 out of 5]

A popular history of the third (and ongoing) AI summer (with some context from the previous two summers and winters). Brings to life the mad scramble and heady days of the post-ImageNet deep learning gold rush (for AI talent) with juicy and detailed anecdotes behind the popular headlines. Little to no technical details and also a seeming lack of exploration of the personal motivations of the main players despite (or because of) the inside access to them.

Covers good ground from the early successes in speech recognition and (Schmidhuber's) highway sign detection leading up to the ImageNet moment and the subsequent arms race between the Big Tech companies (surprised to learn that Baidu was in that early) to acquhire top talent and the subsequent stream of unexpected successes (video games, Go, language translation, deepfakes, robotics, just missing protein folding), with the sidelined critics continuously moving the goalposts on what constitutes AI (but also trying to cash in on it).

The initial story hook about a bunch of long-ignored outsiders (Hinton/LeCun/Bengio mainly) transforming the technology landscape with their old-new idea is compelling but there is still no solid ending (unlike Bad Blood or the Social Network) just a fizzle of issues such as weaponization, bias, limited success in self-driving cars and robots, etc. However, that won't stop Hollywood from optioning this book. Expect to see Tom Hanks as Hinton soon.

[Rick Wilson · Rating 4 out of 5]

Pretty well done history of mainstream AI at the big tech companies.

[Ishan Mukherjee · Rating 3 out of 5]

Due to Goodreads' character limit, the full review is on Substack.

Overall, a useful and brisk history of artificial intelligence. The majority of the book focuses on neural networks, with only the first couple chapters devoted to GOFAI. Here are some of my notes from the book.

Open-source economics

One reason for open-sourcing your work is that people build upon it, and you can profit off of the improved performance. In Genius Makers, Metz explains Google’s rationale for open-sourcing the ML library TensorFlow:

After deploying the software in its own data centers, Google had open-sourced this creation, freely sharing the code with the world at large. This was a way of exerting its power across the tech landscape. If other companies, universities, government agencies, and individuals used Google’s software as they, too, pushed into deep learning, their efforts would feed the progress of Google’s own work, accelerating AI research across the globe and allowing the company to build on this research.

As long as it sets an industry standard, companies need not fear open-sourcing software they have spent millions of dollars writing. On Dwarkesh Patel’s podcast, Mark Zuckerberg said:

We take a lot of the low-level infrastructure and we make that open source. Probably the biggest one in our history was our Open Compute Project where we took the designs for all of our servers, network switches, and data centers, and made it open source and it ended up being super helpful. Although a lot of people can design servers the industry now standardized on our design, which meant that the supply chains basically all got built out around our design. So volumes went up, it got cheaper for everyone, and it saved us billions of dollars which was awesome.

If Meta open-sources its models, developers will build upon them for free. If models become 10% more efficient as a result, Meta saves "billions or tens of billions of dollars".

The second reason companies open-source software is to commoditize the complement. TensorFlow makes AI research easier and Google can profit off of this research, yes, but TF also drives revenues for Google's TPUs. Ditto for JAX. And as Metz writes in Genius Makers, Nvidia has had a deep learning research wing since early on. Nvidia's hope was that if its research made deep learning easier/more fruitful, demand for its GPUs (which are used in deep learning) would increase. Ben Thompson of Stratechery explains that by open-sourcing models like Llama and supporting the development of the ML library PyTorch, Meta makes AI-generated content cheaper to produce. This is great news for its social media companies, from Facebook to Instagram.

Similarly, Google's open-sourced Angular framework makes it easier to build websites, and it's websites people need a search engine for in the first place. Google Chrome's richly-featured Dev Tools is also a push towards making web programming easier. Another browser, Mozilla, hosts one of the world's best collections of resources on web programming, MDN. As of March 2023, the top financial contributors to MDN are "Google, Microsoft, Igalia, Canva, and JetBrains". All these companies stand to gain if more people build websites. Google and Microsoft make browsers and search engines, Igalia makes browsers, Canva helps create graphics and wireframes, and JetBrains makes IDEs, which are essential infrastructure for web development.

In this post, Byrne Hobart explains how ElevenLabs commoditizes its complement. Its core product "converts text into voice, for arbitrary voices". So why is it launching a tool to isolate voice content from background noise? In part because "[i]f ElevenLabs can dominate text-to-voice (still very much an open question) it's in their interest to make voice samples as abundant and cheap to acquire as possible". It's these samples that will get fed into ElevenLabs' core product. A more whimsical example from another post: Dungeons & Dragons released "a generic role-playing system that could be grafted on to other themes, like superheroes, spies, and Star Wars". While this increased competition (the number of new tabletop gaming companies tripled in two years, while the number of employees per company halved), it made a lot more people interested in tabletop gaming and the basics (20-sided dice, storytelling) much more widely known.

A final reason why open-sourcing work helps is that talent is more likely to work for you. AI academics, used to measuring their impact in publications, may feel ill at ease in a closed-source company.

Metz notes Zuckerberg was able to nab LeCun for Facebook AI Research (FAIR) despite not being able to articulate a use case for AI within the company. Earlier, he had tried and failed to acquire DeepMind for this very reason: Demis Hassabis couldn't see what Zuckerberg was trying to do with AI. One of the reasons Zuckerberg fared better with LeCun was his almost ideological commitment to open-sourcing AI research, and allowing LeCun to keep in touch with academia by spending part of his time at NYU.

Frank Rosenblatt, inventor of the perceptron

After inventing the perceptron — arguably the first nontrivial machine learning algorithm — Frank Rosenblatt pivoted to running brain experiments on rats.

After one group of rats learned to navigate a maze, he would inject their brain matter into a second group. Then he would drop the second group into the maze to see if their minds had absorbed what the first group had learned.

And, yes, this did mean killing rats. From Rosenblatt’s paper:

To obtain this extract, the donor rats were killed by decapitation. … At least some of these [recipient] rats appear to be suffering adverse after-effects from the injection (including injuries and phenol poisoning in several cases).

Genius foundries

Both Rosenblatt and Marvin Minsky were from Bronx Science, the NYC public high school that has produced “eight Nobel laureates, six Pulitzer Prize winners, eight National Medal of Science winners, and three recipients of the Turing Award.”

Minsky doles out high praise to Bronx Science — compared to Phillips Andover and even Harvard, the work was more challenging and the students were “people you could discuss your most elaborate ideas with and no one would be condescending”.

A bit like how both John von Neumann and his collaborator Eugene Wigner happened to be contemporaries at the Fasori Gymnnasium in Budapest. Fasori was pretty influential to von Neumann — if only to the extent that it brought him the math teacher László Rátz, who got von Neumann world-class tutors from the University of Budapest.

Geoff Hinton's remarkable pedigree

You may already know that Geoff Hinton, one of the grandfathers of neural networks, is the great-great-grandson of George Boole (of Boolean algebra). But the rest of his family was star-studded too: his great-grandfather was a science fiction writer who coined the word “tesseract”, his great-uncle got inspired by his father’s four-dimensional fantasy to invent the jungle gym, and Geoff Hinton’s cousin was one of the few women to work on the Manhattan Project. Sir George Everest, after whom the mountain is named, was also a relative.

Geoff Hinton’s father was impressive too — he was an entomologist and a Fellow of the Royal Society. But he doesn’t seem to have been a model parent. “Work really hard and maybe when you’re twice as old as me, you’ll be half as good,” he told Hinton.

Of course, Hinton did okay. When his AlexNet paper had 60,000 citations, he said it was fifty-nine thousand more times than any paper his father ever wrote. He won a third of the Turing awarded for deep learning. The $1 million in prize money was split with the co-winners, Yann LeCun and Yoshua Bengio, but Hinton got to keep the extra cent. But far before AlexNet, when Hinton had only just finished his thesis, “[t]he old bastard died … Not only that, he got a cancer with a high genetic linkage. The last thing he did was increase my chances of dying.”

Hinton's puzzle

Geoff Hinton has a delightful puzzle about arranging two pieces to form a pyramid. Here are the pieces from an NYT article:



Hinton shows how to solve the puzzle in this talk.

In general, Hinton comes across as hilarious. He calls the issue with his spine, which prevents him from sitting down, a "long-standing problem".

Yann LeCun designed the first AI accelerator

I didn’t know that Yann LeCun has an impressive background in electrical engineering. In fact, he may have designed the first bespoke chip for training neural nets:

[In the mid-1990s] LeCun and his colleagues also designed a microchip they called ANNA. … Instead of running their algorithms using ordinary chips built for just any task, LeCun’s team built a chip for this one particular task. That meant it could handle the task at speeds well beyond the standard processors of the day: about 4 billion operations a second.

And where did he do this research? That most hallowed institution of American dynamism — Bell Labs, profiled in Jon Gertner's famous The Idea Factory.

It would be almost 25 years before Nvidia started making GPUs (which exploited ML's embarrassing parallelizability) and Google invented the TPU (which speeded up matrix computations by dealing in integers rather than floating-point numbers).

Speaking of neural networks pioneers who were also electrical engineers, I recently learned from a Diff issue about Marcian Hoff, who helped invent the world's first microprocessor.

The only* truly biology-inspired neural network

Despite their name, neural networks have little to do with the architecture of the animal brain. A few years ago, a machine learning engineer worked through a neurobiology textbook and wrote a series of blog posts on comparisons between biological and artificial neural networks. It's worth checking out. The gist is: you can make hand-wavy analogies between them, but improvement in artificial neural nets has been driven by research on the nets themselves, rather than by sifting through biology textbooks for inspiration.

Except, a specific flavor of neural nets -- convolutional neural nets (CNNs) -- were inspired by biology. First, I'll note how important CNNs are. They're the primary architecture for working with images. Innovations made in CNNs have often trickled into other architectures tailored towards different problem classes. It was a CNN named AlexNet that destroyed the competition in a 2012 image recognition contest and ushered in an AI summer.

A CNN works by chopping up an image into squares, and analyzing each square separately to find broader patterns in the big picture. Consider, for example, the task of classifying 128 x 128 pictures as "cat" or "dog". It turns out that rather than looking at the whole image at once, if a net looks at all possible 5 x 5 patches within images separately, it can learn to recognize edges and splotches of color. These edges/splotches can make up ever more complex patterns, like eyes/whiskers/ears. 3Blue1Brown has a great video on this.

The precursor to the CNN — the neocognitron (objectively the coolest name in all ML) — was inspired by the human visual cortex, as “different sections of the visual cortex process different sections of the light captured by your eyes”.

Amusingly, the clever CNN architecture was built and forgotten in the 1980s. The problem was that LeCun’s nets could recognize digits in 32x32 pixels on checks, but they didn’t work for higher resolutions. Classifications also took inordinately long. LeCun was despondent: he called ANNA “the first (and maybe the last) neural net chips to actually do something useful.”

Obviously, he was wrong. We live in an age in which xkcd 1425 is comically outdated. But what changed? The bitter lesson reared its head. Hardware improvements, like GPUs, made training large nets much faster and cheaper. Geoff Hinton put it this way: “No one ever thought to ask: ‘Suppose we need a million times more [computing power]?’”

Fostering a domestic technical industry

It’s strange how Canada dropped the ball on deep learning. Geoff Hinton worked in Toronto and attracted the likes of Ilya Sutskever and Alex Krizhevsky, who went there for grad school. Yoshua Bengio worked in Montreal and drew Ian Goodfellow (who had competing offers from Stanford and Berkeley). Bengio, in particular, is incredibly keen on staying in Canada — he has repeatedly refused offers from big tech companies like Microsoft to move to the US and work in the industry. Hinton, too, had an affinity for Toronto: his salary was lower than what his pension would have been, meaning that he was effectively paying the university to be allowed to teach.

Despite the devotion of these first-rate researchers, today, the US is undoubtedly the world's leading AI country. Next come the UK and China, then Canada — maybe.

Canada did try some things. In 2017, Hinton's Vector Institute was established with $130 million in funding from the likes of Google and Nvidia to incubate Canadian AI startups. Justin Trudeau promised $93 million in support of AI research centers. Canada is also less cagey about granting visas to AI researchers, especially those who come from Middle Eastern countries like Iran, such as Sara Sabour. Yet, all this seems to have been too little, too late.

Why was it hard to convert the academic hubs into successful AI companies? What, specifically, went wrong? A detailed case study would have lessons for countries trying to kickstart their own AI industries, most notably the UAE and India.

The book provides hints on what can work, when it describes why Baidu was confident it could put self-driving cars on the road faster than Google could, despite starting launching its project years after Google.

This wasn’t because Baidu had better engineers or better technology. It was because Baidu was building its car in China. In China, government was closer to industry. As Baidu’s chief operating officer, [Qi Lu] was working with five Chinese municipalities to remake their cities so that they could accommodate the company’s self-driving cars.

(Another reason: China is less cagey about providing its companies its citizens' data.)

Government intervention in industry, when handled by competent people, can be good. For a specific example, Byrne Hobart's Lessons from the East Asian Economic Miracle is a great look at how MITI helped Japanese industry flourish.

The PR of deep learning

It seems that the phrase “deep learning” started getting used mainly to glam up the field of neural nets, which hadn’t until then produced practical results:

When Hinton gave a lecture at the annual NIPS conference, then held in Vancouver, on his sixtieth birthday, the phrase “deep learning” appeared in the title for the first time. It was a cunning piece of rebranding. Referring to the multiple layers of neural networks, there was nothing new about “deep learning.” But it was an evocative term designed to galvanize research in an area that had once again fallen from favor.

I don't know exactly how much the rebranding helped, but here's an illustrative example from the book. Russ Salakhutdinov, who was to become “one of the world’s leading connectionist researchers”, got drawn into deep learning in part by Hinton’s use of the evocative phrase “deep belief networks” for neural nets. Objectively, deep learning became sexy. By 2016, the Microsoft vice president was complaining that acquiring an AI researcher was as costly as acquiring an NFL quarterback.

On branding, consider the motivation for OpenAI's early projects: a world-class Dota model, and a robotic arm to solve the Rubik's cube.

Both projects were also conspicuous stunts, a way for OpenAI to promote itself as it sought to attract the money and the talent needed to push its research forward. The techniques under development at labs like OpenAI were expensive—both in equipment and in personnel—which meant that eye-catching demonstrations were their lifeblood.

[Vinayak Hegde · Rating 4 out of 5]

One of the rare books that focuses on the people who are building a new technology - AI or more specifically Neural Networks/Deep Learning. The narrative follows the development of Neural Networks leading to the skepticism and the long AI winter where there were few wins. This is followed by some amazing breakthroughs in terms of both software (Convolutional Neural Networks, Generative Adversarial networks, AlphaGo, AlexNet) and hardware (Who knew GPUs would be so good at this specific form of mathematical computation). The book also takes a long hard look at the theoretical (without going too technical), philosophical and ethical under underpinnings of AI through the eyes of different corporations such as Google (Brain and DeepMind), Microsoft, Facebook and OpenAI. The hype, the rivalry, the breakthrough and the competitiveness is seen through the eyes of AI's many whimsical protagonists such as Geoff Hinton, Yoshua Bengio, Yann LeCunn, Ian Goodfellow, Ilya Sutskeyer and Sam Altman. I loved that the roughly chronological timeline culminated in the ACM Turing prize. If you have followed the field for a while, you may recognize several seminal events in AI progress covered in the book such as Alexnet winning the image recognition competition, AlphaGos success, founding of OpenAI and the different acquisitons. The book does a great job of tying these together in a engaging story with proper context. While it written for the lay person, any person in the AI field would still love reading this.

[Kelsey Z · Rating 4 out of 5]

After reading this book, my nervousness about the future of AI has been replaced by cautious excitement (although I’m not nearly as excited as Marc Zuckerberg). I definitely think there needs to be regulation, and we need to make sure there’s not a monopoly on the technology (which is what Open AI was apparently aiming for by making ChatGPT available to everyone).

Also, after finishing this book, I’m convinced that at least with initial AI development, unconscious biases were unintentionally built into the systems. This is definitely a field where you want players from different genders and races involved as the technology continues to get trained.

[Ved Gupta · Rating 5 out of 5]

The most exciting book that I have read this year. I work in the same field and it was so thrilling to know that I graduated at a time when the industry was just taking off in terms of talent wars. Some sense of regret has also creeped in that may be a huge opportunity was missed but that is in the past. So much work has happened and at such lightning-fast speed. It was great to finally learn about so much of the industry inside and the people who have made AI what it is today. Must read for anyone in the industry.

[Pete · Rating 4 out of 5]

Genius Makers : The Mavericks who Brought AI to Google, Facebook, and the World (2021) by Cade Metz is an excellent history of machine learning. Metz writes about technology for the New York Times.

The book has four parts. The first on the slow development of neural networks before the breakthroughs in the early 2000s. The second is about Google, Facebook and others building their machine learning (ML) labs and teams. The third covers some of the concerns with ML. The final chapter looks into the future of ML.

The book is very much focused on the personalities involved. Metz writes extensively about Geoff Hinton and Yann Le Cun. Demis Hassabis, Andrew Ng, Ilya Sutskever and many others also make appearances in the book. One thing that is noticeable is how the vast majority of the leaders in ML were not born as Americans. However most work in America and it is American companies that are driving and profiting from ML. It should be particularly depressing for citizens of the EU given how many of the ML researchers started in Europe.

The book has little technical content. ML could be replaced by ‘magic’ and the meaning of the book would change little.

The character of Geoff Hinton is central to the book. He is truly an inspiration. Geoff Hinton has a back so bad he can’t work at a desk. He also finds traveling incredibly difficult. If he can achieve, then we all need to cut our excuses. He has also had two wives die from cancer.

Genius Makers is a really good history of the people involved in the development of Machine Learning. Metz writes really well and knows how to tell a good story.

[Ali · Rating 4 out of 5]

Great read on the history of AI and intriguing insights in the progress of various approaches and their pioneers. It reads very much like Isaacson’s Innovators.

[San To · Rating 4 out of 5]

A fun book for anyone who works in ML. It'll make you go 'So that's where that came from' a lot.

[Sam Motes · Rating 5 out of 5]

A behind the scenes look into the characters and events that have thrust AI on the world. We are on the bleeding edge of the brave new world AI will bring about.

[Max Greenbury · Rating 5 out of 5]

This was awesome. As it was published pretty recently, I really enjoyed getting background on the deep learning landscape almost up to the modern day. Obviously loads has happened since it was published but loved learning more about the history. From the hardware innovations, race to hire the top researchers and commercial applications, it was super insightful. Excited for what comes next!

[Kathleen · Rating 3 out of 5]

I’m not sure what to rate this book. I listened to it (& even listened to some chapters more than once)… but very little of it actually penetrated my brain. Probably more of a me problem than a Metz problem. I was hoping to come away with a better understanding of AI and it’s implications, but this book was more focused on all of the names, deals, and tech building blocks that came together to get AI to its current status.

[Skye · Rating 1 out of 5]

This book literally put me to sleep - had a hard time following the storyline, and I can't bear the author putting in irrelevant details on how deepmind bought rising desks and which colour of t-shirts they wore in the office. horrible storytelling and the flow is just strange while jumping around a bunch of German and English scientists whose name I'm sure would not be remembered by most of the readers.

[Arthur Kipel · Rating 2 out of 5]

The book describes history of developing theory and applications of neural networks. It describes careers and personal details of the main heroes of this journey and their relationships. Also there are a lot of interesting and funny stories.
But for me the book lacks some integrity: there are a lot of jumps from one story and person to another, it confuses a bit.

[Alicia · Rating 4 out of 5]

Also lol @ the part that talks about how Geoff Hinton doesn't like taking derivatives for backprop so he gets other people to do it for him

[Rafi · Rating 4 out of 5]

This book motivated me to finally start studying deep learning seriously :)

[Shoaib · Rating 5 out of 5]

As someone who's knowledge on AI was mostly informed by 'The Terminator', 'The Matrix' and 'Blade Runner', I am glad that this book set the record straight. Terms like 'AI', 'Neural Networks' and 'Deep Learning' were terms I had heard but never really bothered to look into. To me, AI just meant machines becoming more intelligent and over time taking over the world (*cough* Skynet *cough*). So yes, I was in dire need of a book like this.

Genius Makers is a riveting history book. I call it a history book because it is a historical account of events and people involved in developing this technology and taking it forward. The author, Cade Metz, does not go into the technical aspects of the subject too much and as a reader who isn't a computer scientist, neurologist or mathematician, I am grateful. The book flows across the decades starting from the 1950's till 2020 and documents the events, the major players involved, the struggles of the community, their achievements , challenges and the continuous development of this technology. The author steers clear from giving his opinion, instead choosing to make this a report of the event and conversations as they happened.

A very interesting topic was about the different camps of AI researchers. This book talks about 'connectionists' (who believe machines can learn like humans) and people who champion 'symbolic AI' (who believed that humans would have to define discrete rules for every situation a machine would encounter) . The connectionists are banished to the fringes of the research community after symbolic AI becomes mainstream. Their funding dries up and the community does not take them seriously. This is the case for decades. But some of them still believe. And this belief (and hard work) eventually pays off. This book focuses on the connectionists but I would love to read another one just about the various factions of this community which is divided on belief (almost like a religion).

There are many embarrassing notions I had about AI before I read this book, especially about artificial general intelligence but this was an eye-opener. There are years of research and hundreds of interviews which have gone into this book and it shows. One aspect I wanted it to focus more on were the darker aspects of the technology and community. But I guess that is a topic for another book entirely. Maybe a sequel?

[Nick Frazier · Rating 5 out of 5]

Short history of the people and the concepts that brought us to modern A.I. The book spends time discussing two things that cut across the entire book:

Money and ethics. The few qualified talent in the world lead to a corporate arms race in buying up the talent. In addition, many corporations realized that large sums of money would be required to jumpstart fledgling AI efforts. Many companies know that data, talent, and applied use will drive value in the near future.

From an ethical standpoint, there is a large debate between the future of General A.I. (or what we think of as a computer that can think as well or better than a human). Some futurists are concerned about the ethics and danger of creating General A.I. The other camp? They don't think G.A.I. will arrive anytime soon. Instead, we should focus on specific A.I. to solve nice problems.

Overall, good read to help put into context some of the major concepts of AI with the history/politics of the sub-groups.