Book Review: Architects of Intelligence by Martin Ford

The new title “Architects of Intelligence – The Truth About AI from the People Building It,” by futurist Martin Ford is a real gem and worthy read for several groups of people: data scientists, AI researchers, enterprise decision makers, and while we’re at it, we can also throw in folks who are working to transition into this dynamic field that is in such high demand right now.

Ford clearly spent a lot of time interviewing the industry luminaries who have pioneered AI technologies to get their valuable and insightful perspectives into how the field has grown to such exceptional levels today. The historical tone alone is worth the price of admission because you won’t find this valuable perspective anywhere else in a single resource. Beyond that, readers will get a critical insider’s view for the state-of-the-industry from its most high-profile players. I’ve been doing AI, machine learning, and deep learning for a long time and I loved reading all the well-conceived interviews as they filled many holes in my knowledge.

The book itself consists entirely of transcripts of these interviews, so you’re hearing from these experts directly. Martin’s role is important too, by asking the right questions. I appreciated his well thought out inquireies, which were technical enough to make sure that the reader gets the whole story from each interviewee. In addition, the conversations flow naturally in a way that’s interesting to read, and just about all the topics are easy to understand without a deep technical background.

Twenty-three of the world’s foremost researchers and entrepreneurs working in the field of AI were interviewed in the book: Geoffrey Hinton (University of Toronto and Google), Ray Kurzweil (Google), Yann LeCun (Facebook), Demis Hassabis (DeepMind), Yoshua Bengio (University of Montreal), Fei-Fei Li (Stanford and Google), Andrew Ng (Stanford Adjunct Professor, founder Coursera, CEO Landing AI, founder AI Fund), Jeff Dean (Google), Rana el Kaliouby (Affectiva), Nick Bostrom (University of Oxford), David Ferrucci (Elemental Cognition), Barbara Grosz (Harvard), Judea Pearl (UCLA), Rodney Brooks (Rethink Robotics), Daphne Koller (Stanford, Coursera), Stuart Russell (UC Berkeley), Gary Marcus (NYU), Daniela Rus (MIT), Oren Etzioni (Allen Institute for AI), James Manyika (McKinsey), Cynthia Breazeal (MIT), Josh Tenenbaum (MIT), and Bryan Johnson (Kernel).

My favorite interviews were with Yann LeCun, Yoshua Bengion, Andrew Ng and Geoffrey Hinton just because these are the experts I routinely encounter in my work. I wish that Ian Goodfellow was part of the group.

One particular section of the Hinton interview was particularly intriguing. Often called the “inventor” of the backpropagation algorithm, Hinton corrects the record in a very humble way: “… it’s something I feel I’ve got too much credit for. I’ve seen things in the press that say I invented backpropgation, and that’s completely wrong … My main contribution was to show how you can use it for learning distributed representations, so I’d like to set the record straight on that.”

The book was filled with many other insights just like that, and I feel this resource represents a valuable record of AI going back to the beginning all the way up to current day research. If you’re just climbing aboard the AI bandwagon, I would read Ford’s book along with one of the best new books on the technology, “Deep Learning,” by Goodfellow, Bengio, and Courville. This one-two punch will give you a good perspective for the current state-of-the-art along with a solid history of the field.

Ford also conducted an informal survey, where he asked each interviewee “to give me his or her best guess for a date when there would be at least a 50 percent probability that artificial general intelligence (or human-level AI) will have been achieved.” Sixteen of the experts agreed to take a guess anonymously, and two people took a guess with their name attached. Ford points out, “the average date of 2099 is quite pessimistic compared with other surveys that have been done … most other surveys have generated results that cluster in the 2040 to 2050 range for human-level AI with a 50 percent probability. It’s important to note that most of these surveys included many more participants and may, in some cases, have included people outside the field of AI research.” Any way you look at it, the experts feel that AGI is a long way off.

I’ve done a lot of interviews over the years for insideAI News and I know how popular this form of communication is for our readers. In this book, you have the best of the best all under one roof. I wholeheartedly recommend this book for anyone interested in learning more about this dynamic and growing field.

Contributed by Daniel D. Gutierrez, Managing Editor and Resident Data Scientist for insideAI News. In addition to being a tech journalist, Daniel also is a consultant in data scientist, author, educator and sits on a number of advisory boards for various start-up companies. 

 

Sign up for the free insideAI News newsletter.

Speak Your Mind

*