Top 10 books for 2021 on human intelligence for curious AI folks
We have to understand biological intelligence to understand artificial intelligence. We might want to find inspiration from AI and use an artificial learning algorithm to help us think about how our brains might work. We might want to make better predictions about AI’s progress and to understand what AI actually has to do to beat a human, consistency, all the time, in every circumstance. Or we might want to design an AI-enabled application so that it strikes the correct balance between augmentation and automation of human activity. These all require a deep appreciation of what is uniquely human, something that science is still working on.
Covid gave us a chance to read—a lot! As we prepare for the holidays (and factoring in supply chain issues and shipping delays) we thought we’d share the books that have had the most influence on our ideas about the emerging human-machine community this year.
Starting at the beginning—and I really do mean, the beginning at 3 billion ya—Peter Sterling’s What is Health? (2020) builds humans up from our basic biology via millions of years of evolution and explains with absolute clarity the mismatch between our biology and modern life. For curious AI folks, it opens the door to numerous inferences relevant to human-centered AI. A key concept laid out in the book is the biological root of culture and ritual, which allow strangers cohere to share knowledge and which underpins our social decision-making.
How We Learn by Stanlislas Dehaene (2020) explains why humans are better than machines (for now). The real art of this book is the chimeric journey he takes, weaving together what matters to humans and what’s possible with machines. This book organizes so effectively around learning as a concept that you come away reframing everything important, everything that matters, as learning.
Being You, by Anil Seth (2021) is remarkable. Consciousness as a subject can feel impenetrable and perhaps irrelevant. After all, who really wants to spend time talking about whether reinforcement learning agents suffer? Anil Seth’s clear and no-nonsense thinking makes consciousness science into an everyday wonder and source of curiosity. We haven’t totally figured out how to incorporate all his insights into our work (yet) but we have an intuition that how he thinks about human thought is going to be a fascinating thread to pull in the world of AI design.
The Extended Mind, Anne Murphy Paul (2021). This book is a triumph of meta-research. Riffing off of Andy Clark’s essay from 1997, she takes a central idea—that we do a lot of thinking outside our own brains—and explores all the ways this happens. We store information in the world, we rely on others for knowledge, we think when we move, we think when feel, we think when notice our bodies. The idea of embodied cognition gets a turbo charge with this easily accessible and enjoyable book.
Know Thyself by Stephen Fleming (2021). I went through an entire box of mini post-it notes on this book. Concerned with the science of self-awareness or meta-cognition, this book explains how humans are good at “thinking about thinking.” Closely related in many ways with consciousness, meta-cognition is fascinating for AI folks because it speaks to the “agents in the room.” If we are in an arms race between human and machine learning, how can find better, faster ways to use machines to help us know the limits of our own learning and respond with faster learning? This book deepens the conversation on what it might mean that machines know us better than we know ourselves.
In Seven Lessons About the Brain, Lisa Feldman Barrett takes a hammer to common myths about cognition and brains and in a short space of time manages to get through a huge amount of “need to know” stuff.
Rationality, Steven Pinker (2021) assumes that the reader needs a crash course in logic and statistical reasoning. Great! Many of us do! It’s an excellent reference and works nicely with Tim Harford’s, The Data Detective (2021) which is a modern version of How To Lie With Statistics from 1954. I will also add Framers, Kenneth Cukier, Victor Mayer-Schonberger and Francis de Vericourt, which takes a novel approach at problem solving and counterfactual reasoning.
Finally, Noise by Daniel Kahneman, Oliver Sibony, and Cass Sunstein is a must-read for anyone who follows DK and has been thirsting for the sequel to Thinking Fast and Slow. This time, how human reasoning is flawed by variability as well as bias.
These books join our top shelf reading! (see below for the full back bar and previous vintages).
We’ll be taking a break next week. See you again in December.
Just like our cocktail bar, the best stuff is on the top shelf. These are our “top shelf” reading for mixing the perfectly balanced cocktail of human and machine reading.
If you can read only one book on aligning human and machine values:
Brian Christian, The Alignment Problem
If you can read only one book on the science of the socially aware algorithm:
Michael Kearns and Aaron Roth, The Ethical Algorithm
If you can read only one book on how to think about applying AI to intelligently automate tasks and transform a business model:
Agrawal, Gans & Goldfarb, Prediction Machines
If you can read only one book on the advantages of serendipity and all the ethical calls that algorithms can’t provide:
Edward Tenner, The Efficiency Paradox
If you can read only one book on capital, labor and power in the age of automation:
Carl Benedict Frey, The Technology Trap
If you can read only one book on making sense of statistics in the age of big data:
Carl Bergstrom and Jevin West, Calling Bullshit
If you can read only one book on how we think communally and never think alone:
Steven Sloman and Phillip Fernbach, The Knowledge Illusion
If you can read only one book on the art and science of prediction:
Philip E. Tetlock and Dan Gardner, Superforecasters
If you can read two books on keeping an open mind, confidence and continuing to learn:
Adam Grant, Think Again
Don Moore, Perfectly Confident
If you can read three books on human thinking and behavior:
Robert Sapolsky, Behave
Daniel Kahneman, Thinking Fast and Slow
Lisa Feldman-Barrett, How Emotions Are Made
If you can read only one book on how to think about the disruptive forces shaping our lives:
Yuval Noah Harari, 21 Lessons for the 21st Century
If you can read only two books on human futures, AI, power and the limits of knowledge in a data-driven society:
Shoshana Zuboff, The Age of Surveillance Capitalism
Sun-Ha Hong, Technologies of Speculation