I have been posting a lot about artificial intelligence technology here, mostly about autonomous cars. When I study an unfamiliar topic like this, I like to find a good reference book to set the stage for learning more. Something to anchor my knowledge, as opposed to a bunch of links heading in different directions.
Thanks to a recent Vox article, I found that book. As you can see in the graphic, the indispensable reference book is Rebooting AI, by Gary Marcus and Ernest Davis. It is not quite what I expected, which is probably a good sign that it is what I needed.
You can see that the Vox article gives more examples of Elon Musk giving bad takes on AI, based on progress in self-driving cars and ChatGPT. Gary Marcus offers to bet him a large amount of money that Elon is wrong about the timeframe (the 2020s) for an overarching artificial general intelligence. After reading this book, you will realize how safe that bet really is.
Because I read the Vox article first, I thought that Marcus was strictly negative about AI. On the contrary, the book makes clear that he has some very high hopes for AI. But as an expert in the field, he feels responsible for debunking the hype generated about current AI-based products and demos.
Rebooting AI is a powerful book precisely because it is not an AI debunking project. The authors have much higher aspirations for AI than trying to sell the latest “breakthroughs”. They give a powerful behind-the-scenes tour of what components are working well, and what components are missing in the quest to produce a robust AI that can be trusted to fulfill the promise of intelligent assistants, robots and vehicles.
This book also gave me the background and terminology to understand how AI of today connects with the long-ago aspirations for AI that I remember from 40 years ago. The authors discuss some early, or “classic” approaches to make progress in AI, as well as more recent progress in some areas (deep learning, machine vision, and robotic manipulation).
As a career software engineer and technology enthusiast, I found it an easy and quick (just 200 pages before references) read. It was also an entertaining read, as Marcus is an engaging storyteller. If you want to understand where AI started, where it is now, and where it aspires to be someday, you should read this book.
