Tuesday, January 7, 2020

Review: You Look Like a Thing and I Love You: How Artificial Intelligence Works and Why It's Making the World a Weirder Place

You Look Like a Thing and I Love You: How Artificial Intelligence Works and Why It's Making the World a Weirder Place You Look Like a Thing and I Love You: How Artificial Intelligence Works and Why It's Making the World a Weirder Place by Janelle Shane
My rating: 5 of 5 stars

With You Look Like a Thing and I Love You, Janelle Shane has given us an amusing, engaging, in depth, and surprisingly approachable explanation of how AI works, what it’s good for (and not good for) and why. This is one of those unique books about technology that’s written for non-technologists, yet manages to be in-depth enough to be a resource for those whose knowledge ranges from “I heard about AI once” to “the concepts are familiar, but coding AI is not my day job.” While I build software systems, and have worked around AI and Machine Learning systems, I’ve not built or worked on their internals. This book inspired me to want to learn more.

As AI (and Machine Learning) in its various forms touches many aspects of our lives, this book is a must read for anyone who wants to know more what AI does well, what it does poorly, and why . You’ll learn about the difference between narrow and general AI, basic concepts like Neural Nets and Markov chains, and how AI’s learn, including the impact of training data on how well they do.

You’ll also learn about how AI systems can go astray, either through incidental issues (poor training data, for example) or malicious actions. While the concepts sound technical. Shane makes them very approachable though clear language, and memorable through humor. Since it’s not a text book on the subject, there may be a few places where more technical minded readers may see a few conceptual details skipped over, but between the notes and references, and the context the book gives you to do a good web search, this is not a major problem.

The style of this book reminds me of some of Mary Roach’s books on science, such as Stiff: The Curious Lives of Human Cadavers and Gulp: Adventures on the Alimentary Canal, in how it easily mixes humor in with important factual information. At various times in the book Shane provides examples of some AI-Generated “recipes” to show how the wrong training data can lead to bizarre results. My family read some of these aloud and could we on the floor laughing.

In addition to being an excellent primer on AI concepts, I also started thinking about how many of things that set AIs astray are also things that lead humans to the wrong solutions too, even as we are better equipped to compensate. Some recurring themes are being given the wrong definition of the problem (consider incentives at work and how they often lead to the wrong global results) and introducing biases through the examples we learn from, which lead to the wrong solutions. While Shane doesn’t seem to set out to make people understand how to learn better, I can’t help but think that there are lessons in the book for how we can be better at problem solving.

This amusing, thought provoking and educational book got be excited to learn more about the subject. As I read the book, I wanted to slot time into my schedule to experiment with some machine learning code to better understand the ideas. But even if that isn’t something you are likely to do, the book can inspire you to think more critically about your experiences with Chatbots, recommendation engines, and other places where AI and ML technologies touch your life.

View all my reviews

No comments:

Review: Homo Deus: A History of Tomorrow

Homo Deus: A History of Tomorrow by Yuval Noah Harari My rating: 4 of 5 stars Homo Deus is an in...