On the one hand, parents are very concerned about the job market that their children may face, and on the other hand, the anxiety here is quite blind. The future is so uncertain, and the knowledge structure of parents is often too lagging behind. The "double subtraction" does not really touch the essence of pressure, because the source of pressure is employment. If you don't go to the cram school, you have to spell something else, and there are so many good jobs. Double subtraction does not change the essence of the "zero-sum game".
The book Dark Knowledge contains an insight that has a certain cognitive help for future employment. Through a thorough analysis of the new continent of AI knowledge, it can also tell you how employment in the AI era is different from employment in the Internet age.
The author proposes a very profound new concept - dark knowledge.
So what is dark knowledge? Let's take a look at the diagram above, from the perspective of perceptivability and expressability, to make a division of knowledge. Roughly speaking, it can be divided into four quadrants:
The positive direction of the Y axis represents knowledge that is easy to feel and thus understand. The opposite direction of cognition is knowledge that is difficult for human senses and intuition to feel and understand.
The positive direction of the X-axis represents easy expression (conceptualization, formalization...). words, formulas, etc.), and the opposite direction belongs to the kind of knowledge that is difficult to express.
The knowledge on the right side is easy to spread because it is easy to express. What is taught in school and written in papers is this kind of knowledge. The knowledge in the upper right is easy to feel and understand, and it has been exposed to a lot at the middle school and undergraduate levels. The knowledge in the lower right often conflicts with our intuition, and observation often requires special experimental equipment, which is much more difficult to understand.
Because the knowledge in the upper left is not easy to express, it can often only be transmitted through the limited way of the master and the apprentice, that is, the kind that is not easy to say, but can be understood.
The dark knowledge we're focusing on is located in the lower left, which is formally the way modern AI is best at.
Before I introduce it, let me have an intuitive understanding of it.
2016-3-13, under the temptation of a reward of one million US dollars, Google's "Alpha Dog" and South Korean Go superman Lee Sedol launched a world-renowned "man-machine war", dogs have no taboos, five shoulders to make experts shake their heads. I never thought that the dog's move and style were not flowery fists and embroidered legs, but hidden in the jianghu. In the five chess games, Lee Sedol lost three games first and has already lost. Only in the fourth inning of the consolation match, "One of the Gods Digs" and hits the dog's "BUG". The dog began to "drunken chess" like eating "ecstasy", which finally left a little face for humans.
Since then, an upgraded version of the bug-fixed pseudonym MASTER Sweeping Monk has washed the entire jianghu of the world of Go with 60 consecutive victories online.
The technology behind the dog is a deep artificial neural network that has achieved the supremacy of Go in a way that enhances learning.
This KIND of AI was once one of the five major artificial intelligence technologies, and in 2012 it sprung up, almost destroying the Eastern Evil and Western Poison and the Southern Emperor And Beiqi (Four Gate Sect).
This technique is a simulation of the brain's nervous system, characterized by its ability to discover knowledge that humans cannot comprehend.
For humans, reason and thinking are difficult to separate from language. Language's description of the real world is always incomplete, and it relies on a generalization (representation) of reality with concepts. AI based on artificial neural networks (deep learning) does not have such representational limitations on real-world learning. It uses a super-dimensional parameter space to learn (code) low-dimensional real-world data, and to achieve this dimensionality reduction advantage requires super-computing power, which has been formed in 2012.
Behind Google's dog is an array of thousands of GPU clusters. As a frontier exploration of deep learning, such computing power is indispensable. Once you have acquired a domain-knowledgeed primary training model, a much smaller parameter subspace is sufficient for application.
For example, the "dog coach" currently used by Go players is only a PC equipped with one or two GPU boards.
Now imagine the future employment, such as a doctor, who can have a "medical aid" similar to a "dog coach" on his mobile phone. Insufficient computing power can be solved with cloud computing. "Medical aid" is the cub under the "digital doctor".
The idea, the secret knowledge that the doctor knows is like the coach used by the Go players today, omnipotent. The embarrassment of the various Go dojo bosses today may appear in the field of your child's employment.
Learning to deal with those "domain dogs" may be the norm. In the eyes of AI, the difference between Go and general practitioners is not as big as you think. Isn't that a terrible thing to think about? That's what I mean by a digital assistant. Whether you're a doctor or a lawyer or something, the digital assistant in your field is a "king bomber."
Note: The picture is from the network and is original only.