Book Review by Frank Cerwin:
A Human’s Guide to Machine Intelligence
By Kartik Hosanagar – published by Viking/Penguin LLC, 2019
Technology algorithms that power artificial intelligence assist in making our decisions about what products to buy, where we eat, what news we see, whom to date, and who we hire. The author explores the world of algorithmic decision-making and reveals its potentially dangerous biases. Based on his professional algorithm design experience, he dives into details about how they work, what drives our trust in them, and the ramifications of algorithmic decision-making. Hosanagar begins the book with a bit of history about XiaoIce, the first chatbot launched in China in 2014, and Tay.ai, Microsoft’s chatbot that was launched in 2016 and shut down within 24 hours due to its negative behavior. At its basic composition, he defines an algorithm as simply a set of steps to follow to get something done. He further defines modern AI algorithms as those that have evolved to take in data, learn completely new sequence of steps, and self-generate more sophisticated versions of its algorithm. Advancements now include the ability to reason, understand multiple languages, navigate the visual world, and manipulate objects. The author presents several unanticipated consequences of modern algorithms that include unforeseen benefits, perverse results, and unexpected drawbacks. He asserts that data-driven algorithms that direct our decision making may achieve their objectives but may have side-effects that might warrant a warning label.
Hosanagar provides good contrasting descriptions and examples of “content-based recommendation” (based on similar characteristics), “collaborative filtering” (based on someone else’s preferences), and “filter bubble” (a type of echo-chamber where we hear and see information narrowed by our own perspectives). He describes studies that the same input data can have varying results depending on the algorithm acting on it. He summarizes chatbot learning as “nature versus nurture” whereby “nature” equates to application code and “nuture” equates to data. I was amused with the author’s concluding quote from J.K. Rowling’s book Harry Potter and the Chamber of Secrets: “Never trust anything that can think for itself if you can’t see where it keeps its brain.”
I found this book to be a fascinating and easy read that is thought provoking and entertaining. I recommend it for DAMA members due to its emphasis on the importance of data to AI.