Today is December 31, 2024. The author can't sleep at night, so I will summarize my experience in 2024. History will definitely remember 2024, the year when AI truly began to subvert the world, computer science began to be reborn, and the importance of basic mathematics was truly re-recognized.
In early December, I was shocked and heartbroken to hear the obituary of Professor Yan Weimin from the Computer Science Department of Tsinghua University. The author once took the "Data Structure" course taught by Teacher Yan himself. Teacher Yan's strict training laid a solid technical foundation for the author's lifelong career. To this day, I use C++ to develop geometric algorithms almost every day. I have spent decades optimizing various complex combinations, topologies, and geometric data structures based on pointers, and solving various partial differential equations on manifolds. Whenever I am overwhelmed by pointer operations, I always recall Teacher Yan’s strict clarity and meticulousness. Nowadays, my classmates in the Computer Science Department of Tsinghua University have long been separated from each other. From Beijing to Shenzhen, from Silicon Valley to New York, I often visit my old classmates. Everyone is living a good life. Whenever we recall the past, everyone will be deeply grateful to the teachers at Tsinghua University, grateful for their original career choice, and grateful for the generous dividends of the times. However, the development of AI in recent years is breaking the pampered life of workers in the computer field and bringing deep hidden worries to the information industry.
The computer science education I received back then was very broad, but after decades, I have realized that there are only two basic cores: program = data structure + algorithm; the principle of algorithm design is top-down and gradual improvement. Basic data structures in computer science include linked lists, trees, queues, stacks, and graphs. A large number of basic algorithms are designed around the creation, update, maintenance, and elimination of these data structures. In my daily scientific research, I often use a large number of algorithms in graph theory: spanning tree, shortest path, maximum flow minimum cut, depth-width first search, etc. Equally basic are numerical algorithms and optimization algorithms, such as QR decomposition, singular value decomposition, conjugate gradient, simplex method and various planning methods, etc. The development of AI in 2024 will already be able to completely automatically design these basic data structures and generate these basic algorithms. This has brought disruptive changes to the direction of computer science education and the development model of the information industry. The author has personally found that 70% of the current C++ code can be efficiently completed by AI, and further improvements by humans account for about 30% (this improvement will occupy 90% of development time). The author believes that in the classic computer science principle "program = data structure + algorithm", the data structure can basically be replaced by AI, but the algorithm part still relies on human developers.
Machine-translated from the Chinese original 2024年终总结：AI开始颠覆世界 (published 2025-01-02). WeChat original ↗