You Look Like a Thing and I Love You
You Look Like a Thing and I Love You: How Artificial Intelligence Works and Why It's Making the World a Weirder Place is a 2019 nonfiction book by optics research scientist Janelle Shane. The book documents experiences the author and others have had with machine learning programs, and discusses what "intelligence" means in the context of "artificial intelligence".
Overview
The main title of the book refers to a phrase generated as a pickup line by a neural net that Shane trained on pickup lines gathered from the Internet.Shane discusses the dangers of "artificial stupidity", describing for example a 2016 crash at a city street intersection, which Shane attributes in part to Tesla Autopilot being trained for highway use and therefore failing to properly perceive a blocking flatbed truck from a side view. Shane provides "Five Principles of AI Weirdness", including "AIs don't understand the problems you want them to solve" and "AIs take the path of least resistance to their programmed goal". Shane gives many examples of AI "shortcuts", including the legend of an AI that appeared to reliably recognize tanks from photos, by noticing whether the photos were taken on a sunny or a cloudy day. Another of Shane's examples is a hypothetical scenario where a simulated AI evolved to keep people from entering a hazardous hallway during a fire emergency, learns the optimal strategy is to just kill everyone so they cannot enter the hallway. Because AI lacks general intelligence, Shane is skeptical of efforts to power self-driving cars or to detect online hate speech using artificial intelligence. Shane also pushes back against concerns artificial intelligence will replace people's jobs.