The Harvard Business Review featured an article by Baidu’s chief scientist Wu Enda this week. The article describes what the current artificial intelligence can do and what it can't do.
The following is the full text of the article:
Many industry executives will ask me what artificial intelligence can do. They want to understand how artificial intelligence will transform their industry and how they can use artificial intelligence to reinvent their own company. However, with regard to the power of artificial intelligence, recent media have portrayed some unrealistic future scenarios. (Perhaps artificial intelligence will take over the world soon!) Artificial intelligence is already changing the search, advertising, e-commerce, finance, logistics and media industries. As the founding head of the Google (microblogging) brain team, the former director of the Stanford Artificial Intelligence Lab, and the current leader of Baidu's 1,200 artificial intelligence team, I attach great importance to cultivating the world's leading artificial intelligence organizations and developing a variety of energy. Artificial intelligence products used by hundreds of millions of users. After understanding the impact of artificial intelligence, I can say that artificial intelligence is transforming many industries. However, this is not magic. To understand the impact of artificial intelligence on your business, let's take a look at the hype and see what AI can actually do now.
Surprisingly, despite the wide-ranging impact of artificial intelligence, deployment is still extremely limited. Recently, almost all artificial intelligence developments have been of the same type: providing some input data (A) and quickly getting simple feedback (B). Examples are as follows:
The technology of getting B from A will revolutionize many industries. The software for developing B with A is called supervised learning. However, this software is still far away from the perceptual robots depicted in science fiction. Human intelligence does more than just get B from A. This system of obtaining B from A is rapidly being optimized, and the current best technology is based on a deep learning system, or deep neural network. This system draws on inspiration from the human brain. However, such systems are still far from reaching the level of science fiction. Many researchers are still exploring other forms of artificial intelligence, some of which have proven useful in certain environments. Perhaps we will see breakthroughs that make higher levels of intelligence possible, but there is still no clear way to achieve this.
Today's supervised learning software has its own "Achilles' heel", which requires a lot of data. You need to show the system a lot of A and B examples. For example, developing a photo identification tool requires tens to hundreds of thousands of photos (A) and a label (B) about whether there are people in the photo. The development of a speech recognition system requires tens of thousands of hours of voice recording (A) and corresponding text scripts (B).
So what can this system from A to B do? There is a rule of thumb that explains the revolutionary significance of this system:
If an ordinary person's thinking task can be completed in less than one second, then at present, or in the near future, we can use artificial intelligence technology to automate it.
Many valuable tasks that need to be done manually, such as checking surveillance videos to determine suspicious behavior, determining whether a car will hit a pedestrian, identifying and deleting abusive content on the network, can be completed in less than a second. . These tasks are ready to be automated. However, they are often placed in larger environments or business processes, and it is equally important to analyze the association of these tasks with other aspects of the business.
Developing artificial intelligence requires careful selection of A and B, providing the necessary data to help artificial intelligence find relationships from A to B. Selecting A and B in a creative way has revolutionized many industries, and this will revolutionize more industries.
After understanding what artificial intelligence can and can't do, the next step for industry executives is to incorporate artificial intelligence into their strategies. This means that they need to understand where value is created and what is hard to replicate. The artificial intelligence community is very open, and many top researchers publish and share concepts, even open source code. In this open source world, the scarce resources will be:
1. Data. Many leading AI teams are likely to replicate each other's software for a maximum of one to two years. However, the most difficult thing is to get the other party's data. So data, not software, will be a defensive barrier for many companies.
2. Talent. Simply downloading open source software and "applying" it to your data won't work. Artificial intelligence needs to be customized for your business environment and data. Therefore, the industry is competing for artificial intelligence talents who are qualified for this job and are scarce.
We have written a lot about the potential of artificial intelligence. This may bring help to humans and may also bring disadvantages. For example, we have seen artificial intelligence to bring comfort to the lonely, and we have seen that artificial intelligence shows racial discrimination. However, the biggest threat that artificial intelligence may pose to individuals in the short term is to steal jobs, because the work we can now automate through artificial intelligence is becoming more and more. As leaders, we have a responsibility to ensure that we build a world where everyone has the opportunity to grow. Understanding what artificial intelligence can do and how to adapt to your strategy is just the beginning of the process, not the end.
wrist watch accessories,watch parts,custom watch accessories,watch parts customize
Dongguan Yingxin Technology Co., Ltd. , https://www.dgyingxintech.com