博客來In December 2021, a neural network discovered something no human had ever seen-a hidden relationship in knot theory that led to a new mathematical theorem. For the first time in history, machines weren’t just calculating. They were discovering.
This is the story of a revolution unfolding right now at the frontier of human knowledge.
For ten thousand years, mathematics has been humanity’s most distinctive intellectual achievement-a tower of pure thought reaching toward eternal truths. Now, artificial intelligence is beginning to participate in mathematical discovery, finding patterns we missed, suggesting theorems we never imagined, and forcing us to ask: What does it mean to understand?
Future Mathematics takes you inside this transformation. You’ll meet the mathematicians grappling with AI tools that challenge their assumptions about creativity and insight. You’ll learn how neural networks actually work-not through vague metaphors, but through clear explanations that reveal why these systems can do what they do. And you’ll examine the real discoveries: breakthroughs in knot theory, representation algebra, and olympiad geometry that have already changed mathematics.
What you’ll discover:
- How a neural network found a theorem that mathematicians had overlooked for decades
- The geometric intuition behind deep learning-explained so you can visualize it
- Why transformers and attention mechanisms revolutionized AI’s ability to reason
- The billion synthetic proofs that taught a machine to solve olympiad problems
- What AI still cannot do, and why the hardest problems remain beyond reach
- The philosophical questions that working mathematicians are confronting today
Written for curious readers with no mathematical background required, Future Mathematics is both a clear-eyed assessment of AI’s current capabilities and a meditation on what mathematics means in an age when machines have begun to dream in theorems.
Perfect for readers who loved: The Man Who Knew Infinity, Gödel, Escher, Bach, The Code Book, and Life 3.0
