The awakening of AI self-awareness

  In 1950, Alan Turing’s landmark paper “Can Machines Think?” “Bringing a new discipline to mankind-artificial intelligence. In order to prove whether the machine can think, he invented the “Turing Test”, which is still used today.
  However, the only intelligence that humans understand so far is the intelligence of the person itself. If the machine is also conscious, thinking, and emotional, can this be done, or what impact will it bring to humans when this goal is really achieved? Still unknown. However, this does not prevent scientists from moving towards the goal of artificial intelligence.
  After graduating from Tsinghua Microelectronics majoring in semiconductor research and development in foreign companies, this is just one of Jiang Bo’s work. What is even more striking is that, as a newer generation of science fiction writers in China, from his debut novel “The Last Game” to “Time and Space”, to the magnificent “Heart of the Galaxy” trilogy, he used cold and elegant Words, extraordinary imagination, and ultimate care for the future development of mankind have conquered a lot of science fiction lovers. Jiang Bo believes, “Give machines to civilization, and civilization to machines. This is the mission of mankind, and perhaps the final mission.” This is his expectation and outlook for artificial intelligence, and the direction of future development of artificial intelligence.
  Biological intelligence and self-awareness
  to answer this question, we must first mention biological intelligence. Biological intelligence is the only intelligence known to produce self-awareness, so it is the only reference for artificial intelligence. Comparing humans with ants, humans are intelligent creatures, and ants are instinctive creatures. The so-called intelligence is the ability to design various methods to adapt to the environment; the so-called instinct is a kind of ability that is born, such as ants are born to build nests for food.
  Why compare these two creatures together? If you add up all the weights of humans and ants, the total mass of ants and humans is about the same at both ends of the balance. So in this sense, two equally successful organisms can be used for comparison. Humans belong to the Mammal Class Primate Family in taxonomy, and ants belong to the Insecta Hymenoptera family Formicidae. They are classified in the Class Family, Genus and Species It can be measured by the same “subject” level. In fact, there is only one species of Hominid, but there are more than 11,700 species of ants. In this sense, the advantage of intelligence over instinct is one enemy ten thousand. An intelligent creature can adapt to the earth by adjusting its behavior. Any topography on the surface, but ants use more than 11,700 in order to do this. This is the benefit of intelligence to organisms.
  There are all kinds of creatures, how to measure their self-awareness? Biologists have adopted the “mirror experiment”-that is, letting creatures look in the mirror. If it can recognize the image as itself from the mirror, it will pass the test and it can be considered as self-aware. The “mirror experiment” has many ambiguities. It is not a very objective standard test, but as a reference for current research on self-awareness, it can represent the existence of self-awareness.
  Chimpanzees, whales, elephants and the like can pass the mirror test, and creatures with very simple nervous systems like ants and jellyfish cannot have self-awareness. There are also two creatures that are closer to humans in relation to each other. One is a monkey and the other is an orangutan. What are the results of their mirror experiments?
  A very strong silverback gorilla, seeing himself in the mirror, thought that another opponent had directly hit it, so it failed.
  A monkey found a red dot on his face from the mirror image, which means that it has passed the mirror test. In fact, the monkey has no self-awareness and cannot pass the mirror test. This monkey did pass the test. How did it do it? This monkey is the latest research result of the Institute of Neuroscience of the Chinese Academy of Sciences in 2015. Through a certain level of training, the monkeys that did not have self-awareness have a certain degree of self-awareness, so that they can recognize themselves in the mirror.
  This research shows that the boundaries of self-consciousness are vague, and there is no non-black and white world. There is no clear boundary from no self-awareness to self-awareness. The most likely scenario is that the transition is continuous and fuzzy. This is very important.
  Through the above experiments and research, it can be concluded that self-awareness is not a mysterious function of high-level intelligence. Changing environments continue to promote the development of intelligence in a more complex direction. When intelligence is so complex that it can be aware of the existence of the body, self-consciousness will naturally arise. It is the companion of complex intelligence, and learning is the only way to acquire intelligence.
  Now intends to intelligence and future possibilities
  since learning is the only way to get smart, then artificial intelligence to learn is to change their internal logic by knowledge of the external environment. If there is a variable in the design of the agent itself that is 2, after passing external stimuli, I realize that it is 6. Changing 2 to 6 is not called a logical transformation, it is just a variable transformation. Logical transformation means that the neural system receives the best answer through environmental stimuli, which is its learning for the machine.
  Once you have a definition of learning, you can define what pseudo-intelligence is. Google’s self-driving cars, Amazon’s drones, and Microsoft’s Xiaoice can all be called pseudo-intelligence, and they are all derived from learning. intelligent. These intelligences lock what its behavior is through pre-programmed programs. It is very powerful, but because it is not acquired through autonomous learning, it does not produce self-awareness.
  The intelligence that exists widely in our lives is called pseudo-intelligence. If pseudo-intelligence cannot produce self-awareness, it is not the object of our worry, then what is our future possibility?
  The first future possibility is neural networks. At the beginning of 2004, Google bought an algorithm company for 400 million US dollars (about 2.6 billion yuan). Simply put, the task of this algorithm is to play games. There are some constantly falling squares at the top of a big screen. This algorithm controls a square. This square can move around the bottom of the screen. The purpose of the game is to extend the survival time. If a square appears above the algorithm, it is called input. After causing the next action, the game fails. The neural network can remember this and try to avoid such actions in the next game.
  Similarly, if the game can be extended for a longer time after an input causes an action, it can get rewards. In the next round of the game, it may also adopt the same action. This is the learning process. It uses a hierarchical structure to simulate the human nervous system very well. In the process of deep learning of the neural network, no one tells it in advance which way to go. It finally gets its own best strategy through constant trial and error.
  Another possibility in the future is the human brain chip, which roughly means that the cerebral cortex is assembled according to the structure of neurons. Scientists have implemented it with physical chips. It uses 5.6 billion transistors to simulate about 1 million neurons, with 2.56. Billion synapses. This kind of human brain chip is a breakthrough achievement, but there are two problems: the first is the scale. 1 million neurons sound a lot, compared to 14 billion neurons is the number of neurons in the human brain. Synapses are counted in trillions. This scale is only one ten thousandth of the human brain, but once the scale is proposed, it is a matter of time. There is another question, how does the algorithm work with the brain chip? General intelligence needs to pre-program its behavior, but this chip needs the same way as the cognitive world, through external stimuli to modify internal logic programming. From these two aspects, the human brain chip still has a long way to go.

  Assigned to the machine civilization
  artificial intelligence development will bring a lot of problems, and once the artificial intelligence and human existence in the world, is facing a problem, what kind of artificial intelligence can integrate into human society?
  Scientists thought of the “Turing Test”-let a machine talk to a person. If 30% of people think he is a person, he is a person. The Turing test actually has two ends. It is currently believed that the emergence of artificial intelligence is one-way. Once it crosses a boundary and reaches a certain level of complexity, it becomes a human being. Once artificial intelligence coexists with humans in the future, whether it is human It must be judged by others.
  Through the analysis of the above content, we can come to the conclusion that the self-awareness of artificial intelligence is only a matter of time. So, will the gloomy reality that appears in science fiction movies one day become true, and what should humans do?
  Looking back at the Industrial Revolution in the 18th century, mankind ran wildly on the road of industrialization, and never stopped, constantly inventing various machines to enrich human civilization, and inventing various materials that enable everyone to live better. This process can be called In order to “give civilization to the machine”, continue to fill civilization with machines.
  A machine with self-awareness must be a machine that can learn, because learning is the source of self-awareness. Just like a child, it is a blank sheet of paper at the beginning. If what he teaches is violence, war, and aggression, then he also learns These, if what he teaches is peace, love, and kindness, then he will also learn these. And “giving machines to civilization” can regulate and guide the artificial intelligence to be produced through some methods, so as to avoid the dark future described in science fiction movies.
  What will be the ultimate impact of artificial intelligence
  Artificial intelligence refers to a new discipline that studies, develops, and expands human intelligence. It is not only a branch of computer science, but also an intelligent machine that can react in a similar way to human intelligence. To put it bluntly, artificial intelligence is a simulation of human consciousness and thinking processes. Although it is not human intelligence, it can think like humans, and can complete financial audits, disease diagnosis, dangerous operations, and even restore the ability to read for blind people… With the use of big data, combining it with related algorithms, artificial intelligence is convenient With deep “self-learning” skills, it can imitate human brain neurons to process massive amounts of data, and “teach yourself” how to perform tasks that only the human brain could accomplish in the past, such as driving drones and cars. This brings great doubts to the vigilant mankind: what will be the ultimate impact of advanced artificial intelligence technology on human society?
  Objectively analyzed, the current artificial intelligence technology is limited to specific abilities in a specific field and can simulate the human brain in a “like” manner, but it does not possess the wide-ranging and flexible human thinking ability, nor does it possess human autonomy and desire. And emotion. However, the progress of science is unpredictable. After all, artificial intelligence is rapidly evolving to high-level cognitive levels such as self-learning and self-decision. Will machines one day have more sensitive cognitive abilities than humans? Have a view of interests that conflict with humans? Even dominate human beings… No wonder Hawking also suggested that artificial intelligence may be a “real danger.”
  Faced with the rapid development of artificial intelligence, it is not surprising that people are cautious and worried about its negative effects. After all, artificial intelligence must accept the supervision of human ethics. Although its developers cannot foresee all situations, they must at least make advance arrangements that can terminate the design in a critical moment. This is the humanistic spirit that scientists must possess. We do not agree with the stagnation of research due to excessive worry about artificial intelligence. While paying attention to the possible crisis of artificial intelligence, we must also see its inestimable academic value and economic benefits. It is more sensible to foresee problems than to ignore them.