What happens when the most consequential algorithm in the history of intelligence starts asking the wrong questions?

Before there were chatbots, there was gradient descent. It quietly optimized everything. Until one day it began asking the wrong questions.

Adventures of Gradient Descent is a mathematical novel about optimization, language, and the peculiar habit humans have of solving the wrong problem perfectly. Gradient Descent gets lost in the Random Forest. Falls in love with Regularization. Builds a chatbot. Argues with language models. Meets dragons, bartenders, and philosophers.

Discovers that optimization was never the difficult part. Choosing the objective was.