A few weeks ago, I read the book “You Look Like a Thing and I Love You” by Janelle Shane.
When looking at the title for the first time, I thought to myself, “What a peculiar choice for a book about AI; it reminds me more of a sort of rom-com.” Yet, I was sure that Shane had chosen her title extremely deliberately, enticing me to begin reading the first few pages of the book to find out more… I am so glad I did! Not only were all my questions about the title answered (it actually makes a lot of sense), but it also provided me with a hilarious new perspective on artificial intelligence.
With the perfect balance of information, diagrams and funny real-life examples, this book was very much a compelling read. Although it was written in 2019 and AI has since developed in a plethora of ways, the descriptions provided remain relevant as they relate to the main principles behind AI, which the new systems still rely on. Even for the few sections that are slightly outdated, it is still interesting to compare what the characteristics of AI were before 2019 with the current state of the technology.
The theme that resonated most deeply with me was how unpredictable AI can be. In particular, Shane gives numerous examples of when AI sometimes accidentally solves a different problem than we want it to. For instance, when telling an AI to design a robot body that can go from point A to point B, we would expect it to create a robot that can learn to walk, not for it to make the parts pile up into a tower that then falls over to reach point B (although that technically does accomplish the task of getting from point A to point B)! While this unpredictability sometimes makes working with AI more tricky, it also means that AI can devise creative solutions to solve problems.
In addition, in one section, Shane talks about self-driving cars, an AI project that is being worked on. Having long existed in sci-fi, there are a multitude of benefits that self-driving cars would bring. Yet, experts still doubt whether driving is a sufficiently narrow problem to be solved with the technology available today. There are endless situations that the AI could encounter on the road, making it impossible to train it on every single one such that it will react appropriately when faced with them. This means that the AI has to make decisions in circumstances it has never experienced before, sometimes leading it to make mistakes. The current trial of driverless taxis in San Francisco is a prime example. It has caused controversy given the numerous traffic jams that have resulted from the taxis freezing in the middle of the roads, as well as instances of a car colliding with a fire truck and even driving into freshly poured concrete!
When I asked AI to tell me about the book, I expected a very generic concluding sentence, such as "All in all, this book would be a great read for anyone looking to get a better understanding of AI," which could just be inferred by the title. However, I was so wrong. Instead, ChatGPT concluded with: “If you're interested in AI and enjoy a humorous take on the subject, Janelle Shane's book is a delightful read that offers a unique perspective on the world of machine learning and AI-generated content.” Even after reading the book, I am still left in awe, wondering how a non-conscious machine is able to put a string of words together to provide such a deep and accurate summary!
Credit: Janelle Shane