Will neuroscience become the key to artificial intelligence "super evolution"?

Will artificial intelligence replace humans? This issue once triggered a heated discussion among the whole people. Although artificial intelligence is currently making "smart" quickly, it does not mean that it is really smart. On the contrary, many times it is still silly and naive, and still needs to learn from the human brain.

Recently, Simon Ullman, a professor of computer science at the Weizmann Institute of Scientific Research in Israel, wrote that he believes that neuroscience can provide further assistance for the development of artificial intelligence. So, what is the relationship between artificial intelligence and neuroscience? How does neuroscience further contribute to the development of artificial intelligence? What changes will happen to artificial intelligence in deep fusion neuroscience?

Neuroscience and artificial intelligence are homologous

Speaking of the relationship between artificial intelligence and neuroscience, Wang Xiaoli, an associate researcher at the Shanghai Institute of Biological Sciences of the Chinese Academy of Sciences, summed up two sentences: homologous shunting, discipline independence; cross-integration, long-term integration.

Initially, artificial intelligence and neuroscience were two separate disciplines, with different research objects and research methods. From the time origin of the origin of the discipline, the artificial intelligence discipline originated from the 1956 Dartmouth Summer Symposium in the United States; and the birth of neuroscience can be traced back to the 1891 neuron theory. In this way, neuroscience is regarded as the "predecessor" of the artificial intelligence discipline.

Neuroscience focuses more on the laws of neural activity in the biological sense, and analyzes the mechanisms of high-level neural activity including thinking, emotion, and intelligence. The problem of the origin of consciousness is the ultimate goal of neuroscience. Upper neuroscience is an "experimental science" based on the induction of natural phenomena. Artificial intelligence is a new technical science that researches and develops theories, methods, techniques and application systems that can simulate, extend and extend human intelligence. The research object is not intelligent but intelligent control. At present, the research method focuses on complex Phenomenon "computational science" for simulation.

"But the relationship between neuroscience and artificial intelligence can be simply understood as source and flow." Wang Xiaoli told the Science and Technology Daily reporter that the rise and development of artificial intelligence is inseparable from the nourishment of neurological achievements.

As described in the Simon Ullman article, scientists in the early days of artificial intelligence have used the biological nervous system as a reference object to create a "deep network" brain-inspired architecture that has prevailed in recent years. This is a very distinct "source" case. It has also been talked about by scientists in the field of neuroscientists and artificial intelligence. However, some experts in the field of artificial intelligence believe that the deep network is imitation of the brain in the early stage, and developed an independent method in the later stage. Therefore, it is believed that artificial intelligence has its own method system, which can basically abandon brain science. This view is actually worthy of in-depth discussion.

Academician Pu Muming from the Institute of Neuroscience of the Chinese Academy of Sciences once told reporters that in recent years, advances in brain and neuroscience and cognitive science have led to various cognitive tasks observed at different scales such as brain regions, neuromicrocirculations, and neurons. It has become possible to obtain some activity data of brain tissue. It is known that the human brain information processing process is no longer based solely on guesswork, and the human brain working mechanism obtained through multidisciplinary crossover and experimental research is more reliable. Therefore, brain science is expected to provide a reference for the breakthrough of machine learning and brain-like computing.

However, the anti-feeding or feedback effect of artificial intelligence on the development of neuroscience is also objective. In the basic research stage of neuroscience, artificial intelligence can assist researchers in analyzing complex brain signals, brain nerve map experimental data, and constructing and simulating brain model systems. In the application phase of transformation, artificial intelligence can also accelerate the application of brain science results, such as the diagnosis of brain diseases and the clinical transformation of new treatment results.

Open several paths of artificial intelligence "black box"

In fact, without the theoretical breakthrough of neuroscience, there is no understanding of the primitive of intelligent organisms. The concept of "intelligence" in artificial intelligence is likely to have always been a "black box", and intelligent simulation and expansion may have been rotated around the "periphery". . For example, the National Academy of Engineering's "14 Major Technology Challenges for Humans in the 21st Century" report believes that some of the problems currently existing in artificial intelligence stem from the fact that the design does not fully consider the real brain situation. By revealing the secrets of the brain through reverse engineering of the human brain, it is possible to better design computing devices that can process multiple streams of information simultaneously.

At present, neuroscience has several pathways to help the development of artificial intelligence. Wang Xiaoli introduced that in the specific path, the direction of artificial intelligence development of cognitive and empirical ideas can be continued. For example, for artificial intelligence, it is always a specific task to train it, ignoring the process of its contact with other things. If you give the agent a similar growth environment and growth process, will it make it smarter? The wisdom of human beings is based on communication. The current artificial intelligence has no ability to communicate independently. This is also the gap between the current level of artificial intelligence and strong artificial intelligence, and it is also the future development direction.

But it is also possible that Simon Ullmann's reference to the human innate cognitive system makes more sense. Gain an in-depth understanding of the brain's original capabilities to achieve advanced machine logic capabilities. Humans have the ability to learn how to learn. If the agent learns how to learn, then this second-order learning relationship may make it learn faster. If the future agent has imagination and planning ability, then it may be true. It can create something that we humans can hardly create.

In addition, neuroscience has helped artificial intelligence, and there are several directions in the field of artificial intelligence. For example, constructing a new learning theory combining statistical association and feature association, realizing the association between “knowledge-driven” and “semantic-driven”; constructing bionic and natural computing theories such as fusion deep learning and reinforcement learning, evolutionary computation, active learning, lifelong learning, etc. New theoretical framework; large-scale parallel neural networks, evolutionary algorithms and other complex theoretical calculations; universal artificial intelligence systems with autonomous learning capabilities.

The future integration of the two is promising

So what will happen to the artificial intelligence of deep fusion neuroscience?

In this regard, Wang Xiaoli believes that the current fusion of neuroscience and artificial intelligence only accounts for the tip of the iceberg of the biological brain computing principle. Accurately predicting how artificial intelligence will develop in the future is difficult, but if you have an insight into the discipline of neuroscience, artificial intelligence, and the general trend of human economic and social development, it is still possible to outline the future development stage. This is to identify breakthroughs in innovation and to clarify innovation. The direction is very critical. This is also one of the original intentions of carrying out relevant brain science predictions and technical foresight including China.

From now until 2025, neuroscience continues to maintain rapid development, but the subversive theoretical results are still few. During this period, artificial intelligence and big data technology are the "accelerators" of neuroscience development. By 2030-2035, neuroscience will usher in the first major breakthrough, with subversive results in neuro-perception and neurocognitive understanding, thus feeding back and innovating the original algorithmic basis and component basis of artificial intelligence, human society. Enter the substantive brain intelligence research phase.

By 2050, neuroscience will usher in a second major breakthrough, with subversive results in emotional and conscious understanding, developing a multi-scale, integrated, and verifiable brain model theory, brain-like intelligence into an upgraded version, and Promote the super-human evolution of the human brain, integrate neuroscience and brain-like intelligence disciplines, and human society has entered the era of strong artificial intelligence. Of course, there are many scientific theories and social and ethical issues surrounding neuroscience and artificial intelligence, especially strong artificial intelligence.

"We believe that in the future, the field of neuroscience is promising, and the future integration of neuroscience and artificial intelligence is promising." Wang Xiaoli said that from the perspective of human science and technology civilization, neuroscience and artificial intelligence are two sides of the same coin, although independent of each other, All have a common direction: to provide new possibilities for the survival of human beings and the evolution of consciousness.

Source: Technology Daily

Analog System

Analog System,Digital Analog System,Analog System And Digital System,Wifi Intercom

Zhuhai Mingke Electronics Technology Co., Ltd , https://www.zhmkdz-electronics.com