Paper on the History of Artificial Intelligence Human Development (History of Artificial Intelligenc

1. Artificial Intelligence Image Robot

Author | Wang Jianzong, Qu Xiaoyang Source | Original text of big data DT | Must know! Four images show the significant milestone in the development history of AI. Figure 1 shows the overview of the development of artificial intelligence. The development of artificial intelligence has gone through a long historical accumulation. As early as 1950, Alan Turing proposed the Turing testing machine.

2. Artificial intelligence image generation

The main idea is to place people and machines in a small black room and talk to people outside. If the people outside cannot distinguish whether the interlocutor is a human or a machine, then this machine has intelligence like a human.

3. High definition artificial intelligence image materials

▲ Figure 1 Origin and Development of Artificial Intelligence Subsequently, at the Dartmouth Conference in 1956, the concept of "artificial intelligence" was first proposed. In the following decade, artificial intelligence reached its first small peak in development history, with researchers rushing in crazily and achieving a number of remarkable achievements. For example, in 1959, the first industrial robot was born; In 1964, the first chat robot was also born.

4. Artificial intelligence image creativity

However, due to the serious shortage of computing power at that time, in the 1970s, artificial intelligence entered its first cold winter. In the early days, artificial intelligence mostly executed specific problems through fixed instructions and did not have true learning and thinking abilities. Once problems became complex, artificial intelligence programs became overwhelmed and became unintelligent.

5. Artificial intelligence image materials

Although some people took the opportunity to deny the development and value of artificial intelligence, researchers did not stop their progress. Finally, in 1980, Carnegie Mellon University designed the first expert system - XCON - which has a powerful knowledge base and reasoning ability, and can simulate human experts to solve specific domain problems.

6. Artificial intelligence image high-definition

From then on, machine learning began to rise, and various expert systems began to be widely used. Unfortunately, as the application fields of expert systems became wider, problems gradually emerged, and expert systems had limited applications and often made mistakes in common sense problems. Therefore, artificial intelligence ushered in a second cold winter. In 1997, IBM's "Deep Blue" computer defeated the world chess champion Kasparov, Becoming an important milestone in the history of artificial intelligence.

7. Artificial Intelligence Picture Cartoon

Afterwards, artificial intelligence began to develop steadily and upward. In 2006, Professor Li Feifei realized that experts and scholars had overlooked the importance of "data" in the process of researching algorithms, so he took the lead in building a large image dataset - ImageNet. The image recognition competition began with this. In the same year, due to the continuous development of artificial neural networks, the concept of "deep learning" was proposed. Later, Deep neural networks and convolutional neural networks are constantly catching people's attention.

One of the methods for artificial intelligence images is to use object detection

The development of deep learning has once again sparked a wave of research in artificial intelligence, which is still ongoing today

One of the methods for artificial intelligence film reading is to use object detection

Figure 2 lists some important events in the history of artificial intelligence. Since its birth, machine learning has undergone considerable development and has now been applied to a wide range of fields, including data mining, computer vision, natural language processing, biometric recognition, search engines, medical diagnosis, detection of credit card fraud, securities market analysis, DNA sequencing, voice and handwriting recognition, strategic games Art creation and robotics, as well as a major trend in the future development of machine learning and deep learning that we are particularly concerned about - automated machine learning and deep learning (AutoML and AutoDL).

The process of artificial intelligence film reading is reflected in

▲ Figure 2: Major Events in the Development of Artificial Intelligence 02: Next Generation Artificial Intelligence. Let's first review the development process of artificial intelligence through Figure 3.

▲ Figure 3: Development History of Artificial Intelligence So far, according to the overall upward development process, artificial intelligence can be roughly divided into four development stages, namely the birth period of intensive farming, the industrial period of eager for quick success, the explosive period of gathering small pieces to form small pieces, and the future development period of gradually using AutoML to automatically generate neural networks

In the early days, due to the limitations of computer computing power, machine learning was in a slow development stage, and people paid more attention to endowing computers with logical reasoning ability and human summarized knowledge. However, with the development of computer hardware, especially the application of GPU in machine learning, computers can learn various data features from massive amounts of data, thus effectively completing various basic tasks assigned to them by humans.

At this point, deep learning began to achieve great success in fields such as speech and images. Various deep learning networks emerged one after another, and the accuracy of completing related tasks continued to improve. At the same time, deep learning neural networks were advancing towards deeper depths and more ingenious and complex structures. The research and application of GPUs continued to move forward rapidly with the increasing computational power requirements of neural networks.

Figure 4 shows the development of major neural networks in recent years.

In 2012, AlexNet innovatively designed the deep neural network into two parts to fully utilize the computing power of multiple GPUs, allowing the network to be trained on both GPUs.

In 2013, ZFNet further solved the problem of feature map visualization, taking the understanding of deep neural networks a big step forward. In 2014, VGGNet achieved higher accuracy by further increasing the depth of the network; In the same year, the invention of GoogLeNet introduced the repeated module Inception Model, further improving accuracy.

In 2015, ResNet developed the idea of repetitive modules to a deeper level, thereby achieving resolution beyond human level. However, due to the continuous deepening of the layers of deep neural networks, the number of parameters that need to be trained was too large. In order to reduce the number of parameters that need to be trained without sacrificing accuracy, DenceNet emerged in 2017.

With the continuous development of deep neural networks and the continuous invention and utilization of various models and novel modules, people gradually realize that developing a new neural network structure is becoming increasingly time-consuming and laborious. Why not let the machine itself create new neural networks in the continuous learning process? With this idea in mind, Google launched AutoML in 2017

——An AI network capable of independently designing deep neural networks was subsequently released in January 2018 and opened up as a cloud service, known as Cloud AutoML. Since then, artificial intelligence has further developed, and people have begun to explore how to use existing machine learning knowledge and neural network frameworks to enable artificial intelligence to independently build networks suitable for business scenarios. Another door to artificial intelligence has been opened.

For more fresh news and high-quality content, please follow our CDA official mini program, as well as professional courses and free trial classes. Don't miss it!