Transition from perceptual intelligence to cognitive intelligence?New trends in artificial intelligence

Wan Gang, vice chairman of the National Committee of the Chinese People’s Political Consultative Conference and chairman of the China Association for Science and Technology, said at the opening ceremony of the Fifth World Intelligence Conference that the new generation of artificial intelligence is gradually transforming from perceptual intelligence to cognitive intelligence. “Perceptual intelligence is that machines have the ability to perceive vision, hearing, touch, etc., structure multi-dimensional data, and communicate and interact in a way that is familiar to humans.” Wan Gang believes that cognitive intelligence is based on brain-like research and cognitive science. Drawing inspiration from it, combining cross-domain knowledge graphs, causal reasoning, continuous learning, etc., to endow machines with human-like thinking logic and cognitive abilities, especially the ability to understand, summarize and apply knowledge.

The three stages of artificial intelligence

The industry generally divides the development of artificial intelligence into three stages, namely computational intelligence, perceptual intelligence and cognitive intelligence.

The first stage is computational intelligence, where machines store and compute information. Machines can have “memory” and “computing” capabilities like humans, that is, they can store and process massive amounts of data. For example distributed computing, neural network.

The second stage is perceptual intelligence, which refers to perception capabilities such as vision, hearing, and touch. At this stage, the machine captures the signals of the physical world through sensors, such as cameras, microphones and other devices, understands some intuitive physical worlds, and efficiently completes the work related to “seeing” and “listening”. For example, cameras that can recognize faces.

The third stage is cognitive intelligence, where machines have the ability to think and learn like humans, and are able to make decisions and act autonomously. Such as fully autonomous driverless cars. This stage is still in the early stages of exploration, and there is still a long way to go.

What is the distance between perceptual intelligence and cognitive intelligence? In the process of human language communication, there are many scenarios where memory and logical reasoning are used. If two people had a conversation yesterday, when they mention the information of yesterday’s conversation again today, they can quickly discover the logical and temporal connection between the two conversations; but for machines, can they have the same “memory”? And what about “logical reasoning”?

Why is it so hard for AI to think like humans?

Can machines reach the logical and cognitive abilities of human beings? There are also the following issues that need to be addressed urgently.

The first is subject perception. Thematic perception refers to the perception that people directly perceive, and it is the process in which consciousness directly senses “I”. For example, when a person recalls what he ate last night, he can immediately think of what he ate yesterday, which is thematic perception. Although this problem is complex, it is likely to be solved in the future after certain research; for machines, what is really insurmountable is human target perception.

Goal perception refers to a person’s perception of a specific goal. For example, when people feel thirsty, they want to drink water. This is a kind of goal awareness. However, in addition to actively doing some purposeful things, humans will also do some meaningless things, which is “unreasonable” for current machines.

The purpose of the machine’s calculation is to ask for the optimal solution of a process. For example, AlphaGo, a robot that plays Go, is designed with the only computational goal of winning against its opponent. However, throughout the history of the development of human Go, there are some periods where the goal of “drawing” is as follows. For humans, sometimes the pursuit of the optimal solution is not the ultimate goal, and human intentions are often complex and unpredictable.

Where will the artificial intelligence industry go?

At present, the goal of making machines think like humans and have free will is still far away. However, at this stage, it is more practical for researchers to let machines better replace human work with the blessing of artificial intelligence.

For example, the intelligent robotic arm of an automobile factory, with the blessing of artificial intelligence, greatly saves labor and ensures the accuracy of repeated operations; in the application of big data, the addition of artificial intelligence algorithms can more conveniently and quickly provide users with the required information, and improve the operation efficiency of application software; in mobile phones, adding better artificial intelligence algorithms to intelligent voice assistants can also reduce the burden on people caused by the huge flow of information on the Internet.

How will perceptual intelligence change people’s lives?

In the foreseeable future, perceptual intelligence will surely follow the trend of technology and have a long-term impact on people’s lives. We may not be able to accurately predict when perceptual intelligence will truly be integrated into our daily lives, but just like the Internet and mobile phones that are indispensable in today’s life, perceptual intelligence may also become an inseparable part of human society.

In fact, the AI ​​industry should probably not aim to make machines more human-like, but to free up productivity and replace human jobs. If artificial intelligence can be improved to allow machines to perform more complex and dangerous tasks, it can become a driving force for social progress and a catalyst for human happiness. (Sheng Zhenming)

The Links:   G121X1-L03 LQ104V1DG5B