Why can't we create strong artificial intelligence?

Artificial Intelligence (AI) is a branch of computer science, and its purpose is to study and develop a new technology science that simulates, extends and expands the theory, method, technology and application system of human intelligence. (from Baidu Encyclopedia)

Does artificial intelligence have an end? I think there should be strong artificial intelligence. There is such a joke: a man asked: what may be the last invention of mankind? Answer: automatic invention machine ... the emergence of strong artificial intelligence can be said to be a qualitative change in the discipline of artificial intelligence. If an artificial intelligence can be invented in the future, then the science of artificial intelligence may come to an end.

1950, alan turing raised the question whether the machine can think. He believes that if there is no difference between a computer's answer to a question and a person's answer, then a computer can be considered as a thinking machine. This experiment is called Turing test. Turing also put forward a view of strong artificial intelligence, that is, a Chlid machine, which first makes an agent with a small amount of intelligence, called a black list, and then makes it accept knowledge. The thinking of starting with small and low-level intelligence is consistent with that of heuristic thinkers. Then, a child is like a notebook filled with blank forms, but it is also a mechanism for storing knowledge.

There is an intelligent representation of human brain simulation. As early as 1949, before the advent of the digital computer era, William Gray Walter had designed an early robot, which was called Machina Speklatrix. The main principle is to simulate the working process of the brain with analog electronic circuits. That is to say, if a conductive material is connected according to the way neurons in the brain are connected, can it be conscious? Or the existence of wisdom? Connectivists believe that, at least as far as intelligence is concerned, the constituent materials are irrelevant as long as they can accurately replicate what happens in the human brain. But its problem is that there are as many as 80 billion neurons in the human brain! Moreover, these neurons form a highly complex network system with extremely subtle connections. Each neuron and unit group has a sensitive automation mode, which is regulated by various known or unknown biological factors. Not only that, human cognitive models are also strongly related internally, and human beings themselves may be different. In a word, even today's supercomputers can't imitate the whole human brain. Unless the next generation of computers have amazing computing power, this scheme is not feasible.

There is a deep learning method at present. According to the definition, deep learning is "learning by using neural networks with deep hidden layers and different connection structures", which is a definition of strong artificial intelligence. But the problem is that most of the current deep learning networks fail to break through the category of "feature detection". At present, deep learning actually only simulates the sensory system in the complex human nervous system, which is far from the decision-making brain. At present, no one can know exactly how to break the game, but one possible direction is to establish a neural network based on "neural network", which is simple to say but difficult to do. How to train a single neural network? This will continue to learn from the way human beings learn knowledge.

Summary: Strong artificial intelligence is still far away. Humans can't create real artificial intelligence with current computer technology, and in essence, it won't bring direct economic benefits, and the possibility of failure is also great. 20 13 Kenneth Salzberg of the Artificial Intelligence Laboratory of Stanford University made a robot that can buy coffee. It can identify the transparent glass door and find the door handle to open the door. Find the elevator and go downstairs to the coffee shop to buy coffee. Finally, it took the robot 40 minutes to deliver the coffee to the office, which was far less efficient than one person spending 10 minutes to buy coffee, but this was only the beginning. With the development of computer technology and the new idea of artificial intelligence in the future, the obstacles encountered at present may be solved.