What else is possible besides neural network artificial intelligence? What is the possibility of artificial intelligence in the future? How will it develop? Zhang Jiayuan, chief technology officer and deputy general manager of Quanta Computer, and director of Quanta Research Institute, believes that it is necessary to understand the nature of the problem and find the third type of talent needed for AI. For the cognition of the brain, the superficial neural network is only one of them Returning to the essence, neural network analysis is one of the keys to promote this wave of artificial intelligence. However, this is not a new term. It was proposed more than 30 years ago, including neural network, parallel computing, etc. have already existed. Zhang Jiayuan observes that the biggest difference between the present and the past is that the higher the degree of digitalization in society, the greater the amount of data required, and the emergence and upgrade of GPUs and servers have greatly improved the computing power of computers. "Under the situation where such a large amount of computing becomes possible, artificial intelligence based on data and neural networks has exploded," Zhang Jiayuan mentioned. This is the artificial intelligence that everyone is familiar with now: the pursuit of sufficient and complete data, Interpret the "wisdom" revealed in the data, and use machine learning or deep learning to select an appropriate neural network model to realize it. Because this conforms to the needs and current situation of the industry, and the tools are relatively complete, it develops relatively rapidly; and the entire education system also trains talents around this form, which has become a prominent school. But to him, it's just one branch. Is artificial intelligence just neural networks? "It's far more than that!" Zhang Jiayuan said frankly that human beings' understanding of the brain is actually very superficial, and there are still many unknowns about the operating model of the brain; in addition, the understanding of the mathematical model in the nonlinear system is not deep enough, so in terms of technological development. That being said, there is still a long way to go. Back to basic thinking, what exactly is artificial intelligence? Zhang Jiayuan believes that artificial intelligence is still bionics to a certain extent. The artificial neural network is the skin and nerve layer of the human brain. However, our thinking and emotions are not only nerves, and the brain is not only neural networks, but also brain cells. , "How do brain cells convert into a mathematical model that humans can understand?" Further thinking, after the development of the mathematical model of brain cells, if coupled with the development of quantum computing, is there a chance to create the next wave of qubit artificial intelligence? wisdom? At present, many biologists and cognitive scientists have invested in related research, which is quite different from Taiwan's pursuit of rapid commercial application of AI. Yang Chun, a navigator who has stuck to his space post for 45 years The general consensus among scientists is that human beings have a complete understanding of the brain and the understanding of the mathematical model of artificial intelligence, which is the key to the next development of AI, and this involves the key issue of whether AI can be explained. Zhang Jiayuan, who holds a doctorate in aerospace, mentioned that, for example, there was a debate 30 years ago on whether "artificial intelligence neural networks can be applied to aircraft or space." If you really want to apply to these fields, you need to prove that neural networks Lu is capable of stabilizing the flight safety of the aircraft. Of course, there is no way to prove it mathematically, so the aircraft is still controlled by the formula derived from traditional Newtonian mechanics.
Forum Role: Participant
Topics Started: 0
Replies Created: 0