Prediction Machines: The Simple Economics of Artificial Intelligence – by Ajay Agrawal, Avi Goldfarb, and Joshua Gans

Three economics professors based in Toronto, one of the world’s epicenters of artificial intelligence, authored the timely book Prediction Machines: The Simple Economics of Artificial Intelligence (Harvard Business Review Press, 2018). This book’s central tenet is that prediction is the process of filling in missing information and AI with its ability to predict is at the heart of making decisions under uncertainty.

The authors strategically divided the book into five sections to reflect each layer of impact from AI: Prediction, Decision Making, Tools, Strategy, and Society. First, with predictions now being less costly, this paradigm shift increases the value of prediction complements: data, judgment, and action. Another key takeaway is elucidating that prediction is only one of the essential components of a decision, with the other decision elements being judgment, action, outcome, and three types of data (input, training, and feedback). Since humans can enable data, influence action, and render judgment, humans will need to be in synergy with the machines and AI for the best decision making process, at least for now. Lastly, an erudite discussion focuses on AI and its inevitable tradeoffs: more speed and less accuracy; more autonomy and less control; and more data and less privacy. 

The discussion on medical image interpretation and AI is noteworthy. The authors disagree with Geoffrey Hinton’s proclamation that we should discontinue training for future radiologists. They delineated five clear roles in the reconfiguration of human radiologists in the era of deep learning of medical images, at least in the short and medium term, including choosing the image, using real-time images in medical procedures (interventional radiology), and interpreting machine output (and advising primary care physicians). Perhaps in the future there will even be a new type of subspecialist who will have both medical image domain knowledge as well as convoluted neural network and deep learning expertise.

Overall, this book is an insightful framework for anyone who would like to explore the portfolio of AI’s prediction tools in the search for value in health care and medicine. The authors succeeded in bringing both clarity and reality to the reader in the midst of the current AI hype.

 This review originally appeared in AIMed Magazine issue 04, which you can read here. 

Reviewer

dr anthony chang ai medicine healthcare artificial intelligenceAnthony Chang, MD, MBA, MPH, MS

Chief Intelligence and Innovation Officer

Medical Director, The Sharon Disney Lund

Medical Intelligence and Innovation Institute (MI3)

Children’s Hospital of Orange County