This is an edited volume: every chapter is signed by a different group and the whole has no argument. Inside one cover you get a controller for solar plants, clickbait detection with transformers, drone path planning around no-fly zones, swarm-inspired optimisation, the conductance of DNA molecules, blockchain in food supply chains, and a chapter applying machine learning to numerology. That is not an exaggeration: it sits in the table of contents between chapters four and five.
What I took from it
Little, and it is worth saying so. Two chapters on nature-inspired optimisation reminded me of something the language-model era tends to forget: there are assignment, routing and scheduling problems that a classical optimiser solves better than anything that holds a conversation. When a client asks for artificial intelligence to plan visits or balance a workload, that is usually what they need, and it is cheaper and easier to explain.
The other thing I took is a fast discard rule: if a volume mixes unrelated subjects and one chapter argues that something about a person's life can be predicted from their number, there was no editorial filter, and that contaminates any trust you might place in the chapters that could have been serious.
What I left out
Practically all of it. Volumes like this exist because academia needs to publish, not because anyone needs to read them; they cite each other and rarely reach a real decision. I write it down because a reading list on which everything is excellent tells you nothing, and because the time lost on a book like this is time not spent on a good one.