The Word Play Problem of Artificial Intelligence
A deep mistake in the fields of AI and autonomous driving is that they simply understand the problem of human cognition as "word play". The so-called "image recognition" in the field of AI, the result is only a word. According to the top-5 accuracy standard, five words are recognized by the neural network, as long as one of them matches the marked text label, it is considered "correct". But all of this is words without touching the essence (nature) of things. Just like that, they say that neural networks surpass human vision. But what is the practical significance of such "correctness" and "superhumanity"? It is impossible for the field of autonomous driving to use this technology to drive a car.
In daily life, when a person sees an object, what comes to mind is not its name. Only when he needs to communicate with others, he will recall its name. What many people do not realize is that one cannot decide what to do with something just by knowing its name. Human cognition includes a more direct and deeper understanding of things. He does not need to know the name of a thing in order to decide quickly what to do with it. He can directly know the nature of things from sensory signals, but he does not need to know its name. He can even know how to treat things he has never seen. In fact, not only humans, but all animals have this ability. Other animals have no language ability, they don't know what things are called, but they know what kind of properties they have.
It’s not just this AI boom, the AI field has been based on such misunderstandings since its inception. The so-called AI is completely different from human intelligence (animal intelligence). Stop being superstitious about these technologies, take a closer look at all kinds of life around you, and understand their magnificence and greatness.