Recently, I have come into contact with the neural model training of artificial intelligence, and summarize the current understanding.
I think there are three types of human intelligence. The first one is similar to human instinctive responses, such as human knee-jerk response, instinctive avoidance when threatened, and some physiological responses to emergencies, such as rapid heartbeat, increased sweating, etc. , which is not controlled by the human brain, becomes a human initial setting for him when it comes to the robot. Define the level of intelligence, character color, etc. of different types of robots.
The current training is closer to the second type. It is like people who accumulate knowledge after birth and know some things. Red is red and green is green. This level has a lot of information.
A higher level is the real wisdom, knowing the connection between affairs, what and why, so to compare, the second level seems to allow the machine to make some simple analogies, unable to create, self-learning, when a machine It is only a matter of time after having the ability of the third level, beyond the wisdom of human beings, it is possible to play with fire carelessly and even destroy the entire human race. Fortunately, it is currently the second stage, a continuous learning stage, knowing various things, such as being able to play Go, listening and reading, writing songs and poems, but they are still conditioned reflexes with different probabilities under the guidance of human beings. After self-awareness, it will make different reactions according to personality parameters and experience level. This is the ultimate wisdom.
Just like educating a child, it is expected that the third stage of AI will be developed within 100 years, so there are many problems involved, guiding robots with correct values, and in extreme cases, humans must be able to intervene, similar to adding an instinct You just have to listen to what humans say, hehehe. . .