So here is the lowdown: sometimes when life gets strange, the best way to cope is to lean into the weirdness, engage a child-like sense of wonder, and suspend judgment. Janelle Shane’s book “You Look Like A Thing And I Love You” introduces a little levity to our current (and appropriately serious) collective circumstance to show us “How Artificial Intelligence Works and Why It’s Making the World a Weirder Place”. The result is a sublime journey through AI that only the author behind www.aiweirdness.com could pull off.
Through truly bizarre examples that make you laugh out loud while secretly learning, the author talks about structured and unstructured machine learning, natural language processing, deep learning, ethics/bias, and much more. In fact, her step-by-step approach to explaining neural networks is one of the best I’ve seen. Think a “chicken and marshmallow” sandwich is good? A neural network without a well-tuned deliciousness activation function might lead you astray.
Make sure your activation function is set [pg 71]
The author also portends a few of the more troubling elements of AI. Adversarial attacks could introduce ways to fool imaging, perverse instantiation may turn simple tasks into catastrophic outcomes, and not understanding the difference between humans and zombies might cause a self-driving car to slow down at a crosswalk (rather than speeding up as you might hope with slow-moving zombies).
Explaining bag-of-features modeling… with zombies [p 207]
If you want an AI book without code or math, but with examples that allow you to understand complex problems through stories about phantom giraffes, failed recipe ideas, and other curiosities - this is the book for you.