Academicians talk about technological self-reliance and self-improvement. Educators are not focused on the past or the present, but on the future. At present, the development of artificial intelligence models is amazing. How to timely incorporate the latest technological achievements into classroom teaching, enable students to learn with exploratory and innovative thinking, and teachers to teach with research-based thinking should be the basic principle that innovative talent cultivation should adhere to. Since 2018, I have been conceptualizing the knowledge system and curriculum of my undergraduate major in artificial intelligence. Up to now, Xi'an Jiaotong University has fully trained two undergraduate graduates in artificial intelligence. In this process, there are gains and regrets, and it is necessary to constantly search for problems and gaps. For example, this year, we cancelled the college physics course for undergraduate students in this major. "How can we cancel physics classes?" Many people have doubts. In fact, "cancellation" is a butterfly change - we have specifically offered courses on "Modern Physics and Artificial Intelligence" and "Statistical Physics and Machine Learning", and also added three new courses on "Generative AI and Big Language Models", "Advanced Autonomous Driving Technology and Systems", and "Innovative Design Thinking". In "Modern Physics and Artificial Intelligence", content such as quantum mechanics and quantum computing, relativity and electromagnetics, and physics of matter and condensed matter have been added. Why is that so? This is because we should guide students to look towards the future and let them understand where the challenges and opportunities of artificial intelligence lie: with the exponential growth of computing power, by the mid-20th century, almost 40% to 50% of global energy will be consumed on large models. For humans, this is an unbearable burden. If we cannot give up the demand for computing power growth, then we need to find new sources of computing power. In addition to the currently used silicon-based materials, developing new computational substrates from physical materials seems to open a door, and an exciting and challenging journey has begun! To make students majoring in artificial intelligence aware of this, it is necessary to guide them to actively think and explore in this direction. Not only that, artificial intelligence majors should also have courses on control theory and artificial intelligence, guiding students to think about the importance of control in artificial intelligence. This is because, from the origin of artificial intelligence, Wiener's Cybernetics played an irreplaceable role in its emergence and development. Facing the future, achieving natural interaction and control between humans and machines is essential for the integration of humans and the objective world. Therefore, it is crucial for students to understand the correlation between artificial intelligence and control. In teaching, the mathematics courses in the field of artificial intelligence also need to be reformed. The basic mathematics courses in computer science and artificial intelligence should include matrix analysis, basic concepts of graphs, optimization and operations research, etc. It is worth noting that some mathematics courses should be taught by teachers majoring in artificial intelligence, so that the course content is more vivid and concrete, rather than just abstract formula derivation. Currently, there is a common understanding that artificial intelligence originated from computers and is a secondary discipline under the discipline of computer science. In fact, this is a very one-sided view. The cultivation of artificial intelligence talents requires helping students establish a comprehensive and correct understanding. Talent cultivation is a slow, elegant, and wonderful process. It is not about filling a bucket of water, but about lighting a fire, opening a door, and igniting students