This article is produced by NetEase Smart Studio (public number smartman 163). Focus on AI and read the next big era!
[NetEase smart news January 20 news] Although the term artificial intelligence was formally invented in the 1950s, the concept of artificial intelligence (AI) can be traced back to ancient Egyptian robots and early Greek robot myths. The celebrities in the AI ​​field tried to define AI through the 1956 Dartmouth Conference and the Turing Test, and enthusiastic AI advocates insisted on explaining this to the world in a distinguishable and understandable way. concept.
Artificial intelligence is a mysterious, magical, seemingly endless theme. However, for the general public, it is still an elusive thing. In the prediction of people's future, it is often portrayed as a negative image.
In order to counter the vicious cycle of fear caused by artificial intelligence in Hollywood movies, we need to clearly understand what artificial intelligence is.
How to tell if something is AI
In the broadest sense, artificial intelligence may possess all human cognitive abilities, including the ability to learn. A machine needs only one minute of these skills and can be counted as artificial intelligence. Artificial intelligence is a characteristic of a machine. It is usually a computer program that exhibits intelligent behavior. In this case, "smart" means the ability to achieve goals in different environments or conditions. Correspondingly, in the field of computer science, the field of artificial intelligence refers to the research of designing intelligent systems.
Based on this technical definition, artificial intelligence does not need to have the ability to learn. In the most extreme case, all smart behavior in the machine can be implemented by programmers by writing hard code. As long as the preset algorithm can achieve its goal, the machine can still meet the definition of artificial intelligence. Many current artificial intelligence systems are actually system types based on rules, in which engineers can provide the system with the intelligence it needs.
In addition, machine learning is a science that allows the machine to exhibit intelligent behavior without explicit programming. Specifically, it provides the system with the ability to automatically learn from the data and make improvements without the engineer changing the program code.
To put aside the technical aspects, you can say that artificial intelligence is the goal, and machine learning is one of the ways to achieve this goal - let the machine solve the problem by itself. In many cases, machine learning involves learning and improving models using previously collected data. With data, machines can make empirically driven predictions or decisions. By constantly updating the model, the machine will learn to adapt itself to the changing environment.
AI is not inherently better than humans
In order to understand the capabilities of artificial intelligence, we need to explain what it can do. Although engineers can operate on artificial intelligence and provide all intelligence, machine learning becomes more and more important when creating artificial intelligence systems. This is because machine learning promises to reduce the time for manual engineering while finding an unknown solution, even for domain experts. However, in many cases, the engineering time only changes from the time when the artificial intelligence is directly designed to the time when a machine learning algorithm is designed to learn the solution itself. Human engineers are still indispensable.
At first glance, this is a perfect solution: we create an artificial intelligence that we can learn, show us how to learn to solve a task, and then it can come up with solutions to any related problems, right? It seems that big companies like Google, Microsoft and Apple think so: They use this intuitive expectation to convince people that their artificial intelligence system will solve many customer problems. They invested heavily in artificial intelligence and made major commitments.
In the past ten years, the learning system has perfectly solved the problems of object recognition, speech recognition, speech synthesis, language translation, image creation, and game play. The ability of these algorithms is advertised as having breakthrough capabilities, and it is true. In machine learning, people who do not have a deep technical background often think of the improvement of the machine in performing such special tasks as the combination of different artificial intelligence. This is not entirely correct.
Every day, algorithms are learning how to solve new tasks and get better in other ways. AlphaGo of Google’s DeepMind beat Li Shishi, one of the best Go players in the world. Knowing this, a customer with an engineering background said, "We now have a general-purpose artificial intelligence that has already learned to transcend humanity in go-ahead - then it certainly can optimize the design of automotive exhaust systems." This reasoning is based on the assumption that once the machine learning algorithm is developed to solve a problem, the same algorithm can be easily applied to solve a different problem.
but it is not the truth.
In reality, each of these breakthroughs is achieved through highly specialized machine learning algorithms that have been developed by the smartest people on earth for many years. They are designed and adjusted to solve their specific tasks - just this task.
There are some basic methods, such as deep learning, that can be applied repeatedly in different application areas. However, for most applications, it is necessary to combine different machine learning methods. The generated machine learning system needs to be adjusted according to the data of a specific application, and the training algorithm needs to be adjusted to find a high-performance solution. Each step requires a machine learning expert (usually more than one), supplemented by software engineers and domain experts.
It needs an army
AlphaGo is the result of a multi-year project that involved at least 17 people, several of whom were in the leading position in their respective machine learning areas. According to third-party sources, AlphaGo used 1920 CPUs and 280 GPUs in its games with Sedol.
Large artificial intelligence companies have several world-renowned teams of machine learning experts and work with software engineers. In many cases, each team is focused on a specific application area. The goal is to study incremental methods to improve the current best machine learning methods.
Modern artificial intelligence is more like a mollusk than an omnipotent machine. Biology provides inspiration for today’s artificial intelligence. Biologists have studied a mechanism by which animals change their response to a particular environment after experiencing something that can change the meaning of a certain environment. One word to describe is - learning.
Common subjects are sea urchins (ie, molluscs or jellyfish): Specifically, scientists study genes that determine how neurons burn. Based on their genetic structure, the two sea lions reacted differently to the same experience (ie data). Now, machine learning is roughly at this level - experts have modified the program code of the learning algorithm (similar to the sea rabbit's genetic code), changing its ability and adapting to various experiences.
The state of development of machine learning may be closer to that of invertebrates, such as the sea apes, rather than the advanced cognitive abilities of mammals or humans.
In the past two years, researchers have begun to develop machine learning techniques to adapt to new tasks. However, the methodology has only just begun. In the words of a DeepMind scientist, "Recent research on memory, exploration, compositional representations, and processing architecture has provided us with optimistic reasons." In other words, we have reason to believe that the goal of achieving a wider range of artificial intelligence is feasible. of.
The core message here is that you can't simply pour the raw data into general-purpose artificial intelligence and expect something meaningful to happen—this kind of artificial intelligence doesn't exist yet. In addition, if you provide a rock solid problem definition, the machine can only learn the right solution. For a successful story, you need a good plan, a mathematically sound problem statement, enough training data, a lot of machine learning knowledge and software development capabilities.
(From: CodeBurst compilation: NetEase smart participation: Li Qing)
Pay attention to NetEase smart public number (smartman163), obtain the latest report of artificial intelligence industry.
The 8-inch tablet will have a big impact on the 7-inch and 10-inch tablet market. Because the portability of an 8-inch tablet is stronger than that of a 10-inch tablet, and the usable area is larger than that of a 7-inch tablet. The most important thing is that the price is more moderate, which is much cheaper than a 10-inch tablet. It can be said that the 8-inch tablet computer has a good balance between portability and screen display area, and is more likely to be favored by the majority of users.
8 Inches Tablet Pc,Tablet Computer,8 Inch Android Tablets,8 Inch Tablet
Jingjiang Gisen Technology Co.,Ltd , https://www.gisengroup.com