Following in May this year, AlphaGo defeated Ke Jie, the world's number one Chinese chess player, in the world-class three-player man-machine battle. Recently, AlphaGo’s emergence of the next-generation AlphaZero has rekindled The world’s attention. In this game, Alpha was able to easily defeat Alpha Dogs with a score of 100:0. Alpha's success has brought artificial intelligence to a whole new stage.
I divided the learning characteristics of people and machines into three phases. The first stage is the stage of learning to learn from the outside world. Just as we want to go to school, we learn from teachers and books. Artificial intelligence robots are also at this stage. The Alpha Dog has a training period of up to several months in order to defeat the world champion of human Go. During this time, it needs to strengthen the learning of self-teach based on the chess steps of millions of human professional players. In another case, IBM-based Watson, a medical-assisted robot developed on the basis of "DeepQA" (depth open domain answering system engineering) technology, has now "learned" in order to better provide doctors with alternative evidence-based treatment options. 42 medical journals in the field of oncology research, more than 600,000 medical evidences and 2 million pages of textual data in clinical trials.
In the second stage, it is a self-learning stage of human and artificial intelligence. Just last week, Nature magazine published an academic paper called "MasteringthegameofGowithouthumanknowledge." The article mentions that compared to alpha dogs, alpha does not need any knowledge of human experience to start learning from scratch, relying solely on self-enhancement learning and enlightenment. Among them, the biggest breakthrough is unsupervised deep learning. This is a learning process that does not require human intervention. Just as we learned elementary mathematics when we were young, we can reason and think independently. Alpha's breakthrough is a breakthrough in the inherent mode of human thinking.
The core technology behind this capability, deep learning, was on the world stage as early as 2012. This year, the father of deep learning led the team to successfully increase the accuracy of artificial intelligence calculations to 85% in the world's most famous ImageNet machine vision competition. This figure was 72% in 2010 and 74% in 2011. Just three years later, Microsoft Research used up to 152 layers of neural networks to increase the accuracy rate to 95.06%, which spanned 94.9% of the human visual barriers. It can be seen that artificial intelligence is continuously creating miracles.
The last phase of learning between humans and artificial intelligence, I call it the stage of creating abilities. Chen Yiran, an expert on artificial intelligence at Duke University in the United States, once said that she thinks that the most interesting aspect of artificial intelligence is that it proves that human experience tends to converge to the local without knowing the limitations of the sample space, and artificial intelligence learning can break through. This restriction.
Today's artificial intelligence has entered a stage of self-learning. This is a new breakthrough. Through Alpha's story, we have concluded that human beings often affect their machine learning ability because of their own limitations. Second, an artificial intelligence that relies on external learning is at most similar to a human being who is particularly good at the brain. The future needs more artificial intelligence machines with self-initiated capabilities.
The next decade will be the decade of artificial intelligence. In the next 20 years, artificial intelligence will be applied to more commoditized scenes to gradually replace the simple and tedious labor of human beings. At the same time, human beings will enter into fields that humans cannot transcend. , so all this is still just beginning and it is worth our looking forward to.
Source: China Economic Weekly
Author: famous professional managers, minimally invasive Tang Jun, chairman of China
smart whiteboard Touch Screen for meeting room,conference video system;Smart board Pro for Business
The smart board Pro is specially designed for enterprise smart conferences. It has functions such as ineractive whiteboard writing, wireless and wired screen projection, smart screen sharing, PPT annotations, etc. It also supports multi-person video conferences and transoceanic video conferences.
smart board Pro for meetting,smart board tv,smart board price,smart active panel,smart flat panel
Jumei Video(Shenzhen)Co.,Ltd , https://www.jmsxdisplay.com