What does it mean to be smart? This is the opening question posed in Cade Metz’s Genius Makers: The Mavericks Who Brought AI to Google, Facebook and the World. Genius Makers was a page turner, a fun romp through the life of the notable men (yes, it’s almost exclusively men) behind our current AI moment and our algorithmic future.
Unlike other books about artificial intelligence that focus on the technological breakthroughs in the field, Metz takes a look at the people behind the technology. More specifically, the tale is centered on Geoffrey Hinton. Hinton is an “OG” of AI research and one of the Godfathers of Deep Learning. His connectionist approach to AI and the development of back propagation or back prop, has led to our current obsession with machine learning. His many protégés include Yann LeCun, and Yoshua Bengio (now both AI pioneers in their own right), as well as Alex Krishevsky (of Alexnet fame) and Ilya Sutskevar. The book is a “six degrees of Geoffrey Hinton” take on the story of modern AI.
Genius Makers is organized into sections and chapters that move the reader along the highs, lows and lingering questions in AI. Each chapter is a one word description of a phase in the story - Promise, Ambition, Rivalry, Hype, Religion. There are other words that came to mind for me in reading this tale, namely Privilege, Probability and Patriarchy. Beyond the question of defining smart, for man or machine, there’s another question lurking in the background - a question about who gets to determine the future. In the case of AI, it’s a small, elite and fairly homogenous group of people. Yet the technologies they have built impact all of us.
When we think about what it means to be smart, we tend to think of those who have some level of expertise in a particular field or domain. We might equate being smart with education, holding a PhD and doing award winning research. In the world of business we equate it with financial success and running a successful organization. We live in a culture that ascribes all of this - book smarts or business smarts - in terms of meritocracy. We herald individual choices and individual abilities as being the defining aspects of success. In reading Metz’s positioning of AI through the stories of the men behind it’s current trajectory, this idea of meritocracy, of individual genius, is reinforced and celebrated. The book reveres the people that make up the human genius behind artificial intelligence.
What struck me throughout the book is just how much privilege is behind the humans who’ve built AI. I don’t deny the talent, hard work, dedication or sense of “worth” any of these individual contributors have brought to the field of AI. At the same time, it seems painfully obvious just how many people will never have a chance to create and define the future if they don’t already start from a position of relative privilege. The Genius Makers are a group of people who attended elite schools, were in a social and financial position to pursue PhDs and by and large were the progeny of academics or upper middle class professionals. Those are huge advantages! These were people who were able to take unpaid internships or poorly paid research assistant roles to work on interesting projects early in their career, which then paid off with job opportunities at big tech firms and the financial perks of six and sometimes seven figure salaries. This privilege also yielded personal connections to tech titans - the Peter Thiels, Elon Musks and the like - to fund startup ventures which in many cases were then sold to big tech companies for vast sums of money. This tale of privilege stands in stark contrast to the usual Silicon Valley narrative of meritocracy and disruption. Silicon Valley’s founding story which has been peddled, packaged and sold as a cultural gospel is that anybody can change the world, but Genius Makers illustrates the opposite.
One of the tales in the book that caught my attention is not a primary storyline but a somewhat random plot point in the life of the central figure of the AI renaissance, Geoffrey Hinton. Hinton has dedicated his entire career to the field of AI. His story is one of believing in something unequivocally and sticking with it for the long haul. He may have been content to be an eccentric academic. However, in 2012, Hinton finds himself becoming a startup founder and tech multi-millionaire in his late sixties through the sale of a company setup at the behest of his grad students in order to house their groundbreaking research on neural networks. The book launches with the auction of this company, DNN Research, to Google for a hefty $44M!
But the part of Hinton’s story that most intrigued me was a sad personal detail. Both of his wives, Rosalind Zalin and Jackie Ford, died of cancer at relatively young ages. The fact that Hinton found himself at the University of Toronto in the 1980s, where funding for AI was kept afloat by the Canadian government at a time when it was horribly out of fashion elsewhere, was due to his first wife, Rosalind Zalin. Zalin was a molecular biologist and a self described “confirmed socialist” whose political views were at odds with Reagan era politics. Hinton’s second wife, Jackie Ford, an art historian, gave up her own career in order to move to Canada to support his dream - a fact he gratefully acknowledges during his acceptance speech for the 2018 Turing award.
One of the reasons organizations are investing so heavily into AI is that it promises to be a prediction machine. Through calculating probabilities it provides a crystal ball, a level of certainty in a particular outcome. Yet, I couldn’t help but wonder, what are the odds that both of someone’s life partners would die of cancer at a young age? And that this person is also the pinnacle figure in the story of modern AI - a technology that promises to predict the odds and provide some greater level of certainty? What are the odds of that??!
There are other moments in the book that reflect this element of randomness. Ian Goodfellow - known at the GANfather - gets drunk at a Montreal bar, makes a bold statement about training a neural net with another neural net, goes homes, can’t sleep and instead does an all night coding binge that results in the creation of generative adversarial networks. This fits the lone genius narrative - but even Goodfellow acknowledges being lucky. He says if it had not worked that night he probably would have given up on the idea.
There are eight women mentioned in the players list of notable figures in Genius Makers - Anelia Angelova, Timnit Gebru, Fei-Fei Li, Meg Mitchell, Lily Peng, Sara Sabour, Deb Raji and Joy Buolamwini. Some of these women were mentioned only in passing as contributors to research led by one of the men who made up the central characters in the book.
Most of the women mentioned in the book are concentrated in one chapter entitled Bigotry. Their work is primarily centered on calling out the harms or biases in AI. They are leaders and pioneers in the field of AI ethics. This is important work (my own bias acknowledged!). I was sad to see this topic only received a mere 10 pages of a 311 page book. This chapter feels begrudgingly included - a part of the story that can’t be fully ignored, but at the same time is downplayed. To be fair, Metz does raise ethical issues in the later part of the book. However, the stories that get told are from the perspective of the usual big tech companies.
There are other notable figures excluded from this story. For example, Rich Sutton, one of the fathers of reinforcement learning, a technology that is a key component of the DeepMind and Demis Hassabis story, is barely mentioned. Similarly, the University of Alberta, is glossed over save for a very brief mention as part of David Silver’s story. That work is not given its fair recognition in the historical account.
Finally, I wonder what has become of Alex Krishevsky, Hinton’s grad student and the namesake of AlexNet. Alexnet was the neural network that kicked off the current AI revival in 2012. After making millions in the DNN Research deal alongside Hinton, joining Google, then quitting in 2017 and joining the startup Dessa (now part of Square), Krishevsky seems to have evaporated from the AI scene. But he never really thought of any of it as AI anyways. It was just pattern recognition, “non-linear regression” and applying math to new situations. The book paints Krishevsky as a contrarian character, who states that “deep learning should not be called AI” and downplays his own role in the AI renaissance as being in the right place at the right time. It’s not a view that makes headlines, sells books or warrants the label genius maker.
By Katrina Ingram, CEO, Ethically Aligned AI
Sign up for our newsletter to have new blog posts and other updates delivered to you each month!
Ethically Aligned AI is a social enterprise aimed at helping organizations make better choices about designing and deploying technology. Find out more at ethicallyalignedai.com © 2021 Ethically Aligned AI Inc. All right reserved.
Metz, C. (2021). Genius Makers: The Mavericks Who Brought AI to Google, Facebook and the World. NYC: Dutton.