This article is produced by NetEase Smart Studio (public number smartman 163). Focus on AI and read the next big era! [NetEase smart news January 4 news] The birth of the field of artificial intelligence can be traced back to a long time ago, but many people think that it was a group of scientists gathered in 1956 Dartmouth University discussed together. In the past few decades, computers have been developing at an alarming rate, and now their computing speed is much faster than the human brain. From an optimistic perspective, given the incredible progress in the field of computers, it is reasonable to have such achievements. Alan Turing, a talented computer scientist who had been in the field of artificial intelligence several years ago, has already proposed the idea of ​​machine thinking. Scientists also have a rather simple idea about this: intelligence can be analogized to a mathematical process, and the human brain is also a machine. After mastering this process, you can have a machine simulate it. This problem does not seem to be difficult. Just as the scientists in Dartmouth wrote: “If we allow a group of carefully selected scientists to sit down and discuss together throughout the summer, we think they can make significant progress on one or more of these issues.†I think By the way, this research proposal contained the earliest part of the term artificial intelligence. They have many ideas for this. For example: The neuron model that simulates the human brain can realize and teach the importance of the machine's abstract rules about human language. Scientists are optimistic about this, and not surprisingly, their efforts have finally paid off. Before that, they had a computer program that could understand and understand human language, and it could solve algebra problems. And people confidently predict that humans will build an artificial intelligence machine in the next two decades. What we can predict is that, shortly after the emergence of the artificial intelligence industry, real bionic artificial intelligence robots will also appear and be more like us. In fact, its speculation can be said from Turing’s first paper on the “thinking machine†in which he predicts that the Turing test (a test that allows humans to believe that machines have human characteristics) will be After 50 years of formal adoption, this kind of machine may appear around the year 2000. Of course, people like Ray Kurzweil, the famous scientist, are still predicting the possible progress of artificial intelligence in the next 20 years. In many different experts and analysis surveys today, you almost want to know if artificial intelligence researchers are just trying to give an automatic reply: “I’ve known the questions you are about to ask, no, or yes, I cannot Really predict what you have to say." The problem with trying to predict the exact date when humans have artificial intelligence at the same level is that we don't actually know how far it is to achieve this goal. This is not like Moore's Law. Moore's Law states that the machine can double its information processing capacity every two years, which can make very specific predictions for a very specific phenomenon. From this we can get a general idea of ​​how to achieve the goal, such as improving the engineering design of silicon wafers: we know that we are not at the basic limit of our current method (at least until you try to transform the chip at the atomic scale). But artificial intelligence and what we said above cannot be analogized. They are completely different things. Common mistakes Stuart Armstrong's research investigates these predictions about artificial intelligence trends. Specifically, he is mainly looking for two major cognitive biases. The first is that AI experts predict that true AI will occur before they die (and prevent them from dying). This is what people have criticized Kurzweil's "nerdy fury" because they believe that Kurzweil's predictions are based on fear of death, on the immortality of life, and fundamentally irrational. This makes the ability to create super intelligence as a project with personal beliefs. Of course, there are also criticisms from the artificial intelligence development staff who have encountered the setbacks and limitations of today's artificial intelligence. Second, most people think that real artificial intelligence will appear in the public's view within 15 to 20 years. This period of time is actually enough to make people believe that they are doing something that can quickly prove to be a new revolution (people will not be very impressed with what they are trying to behave in a regular way, for centuries it has been), and so far it has not Say these proofs are wrong. Of these two deviations, the Armstrong study has more data for the latter? - People are very happy to choose to predict artificial intelligence after death, although most people do not, but in historical predictions There is a clear "15-20 years" deviation. Measuring progress Armstrong pointed out that if you want to assess the feasibility of a specific forecast, there are many parameters before you can refer to it. For example, the idea of ​​developing human intelligence by simulating human brains will at least provide you with a clear path for you to assess progress. Whenever we get a more detailed map of the brain, or successfully simulate another part of the brain, we can see that we are moving toward this ultimate goal, and this goal may be reached after the emergence of human-level artificial intelligence. On this path we may finish in less than 20 years, but at least we can use scientific methods to assess progress to consolidate this. If a network is complex enough and gives enough processing power to people who "emerge" artificial intelligence, this may be the human intelligence and so-called "consciousness" that we imagined in the course of evolution -- despite human evolution Billions of years, not just decades. But the problem is that we have never had empirical evidence: We have never seen consciousness reflected in a complex network. Not only do we not know if this is possible, we do not know how far we are from it because we cannot even measure its development potential and value. To understand which tasks are difficult for the development of artificial intelligence, there is still a long way to go from the day of its birth until today or even later. As long as you look at the original research plan, you can compare it from understanding human language, to randomness and creativity, and self-improvement. We have good natural language processing, but our computer can understand what they are dealing with? We have artificial intelligence. It can be "creative" at random, but is it creative in itself? The exponential self-improvement process that the singularity often depends on seems to be far from the goal. We also have difficulty understanding the meaning of intelligence. For example, AI experts have been underestimating the ability of artificial intelligence to go. Many people believe that humanity will still explore this issue in 2015 and even 2027. But in the end, it only took two years, not 12 years. But does this mean that artificial intelligence can write "a great American novel"? Does this mean that it is closer to conceptual understanding of the world around it? Or does it mean that it is closer to what we call human-level agents? These can not be explained clearly for the time being. Not human but smarter than human But perhaps we have been looking at such issues from the wrong perspective. For example, the Turing test has not yet passed. Artificial intelligence cannot convince humans in conversation that it is an agent that can be similar to humans. Of course there is also computing power, and the ability to perform pattern recognition and other tasks such as driving a car may soon be far beyond the level of humans. Due to "weaker" artificial intelligence algorithms making more decisions, IoT communicators and technology optimists are trying to find more ways to provide more data or more algorithms, so this "artificial intelligence" The impact on society can only be a growing trend. Maybe we haven't set the relevant management mechanism for human intelligence yet, because we don't know how long we can use existing algorithms to influence it. National automation will disrupt society and fundamentally change the risk investigation on it, because we don't have many assumptions about some ambiguous super intelligence. Some people have pointed out that we should worry about the birth of artificial intelligence for other reasons. It's just because we can't determine whether artificial intelligence at the same level as human beings will be developed in this century, or it will never exist in our lives, but that doesn't mean that we shouldn't be optimistic about its predictions. It is the right possibility to prepare. We need to ensure that people's values ​​are incorporated into these algorithms so that they understand the value of life and act in the proper manner as moral standards. Phil Torres, author of The Future Human Prosperity Plan, expressed this in my interview. He pointed out that if we suddenly decide to integrate intelligent robots into our lives, as an evolving society, we must solve such ethical issues in the future (deciding on the right and wrong to incorporate machines into our lives). And when can we do this? Therefore, we should question these predictions. And keep in mind that the problems foreseen by the pioneers of artificial intelligence are much more complex than expected, and so is today. At the same time, we cannot prepare for the emergence of artificial intelligence. We should understand the risks and take preventive measures. When those scientists held collaborative talks at Dartmouth University in 1956, they did not know what they would face in the future. After 60 years, we still do not know how far we should go in the development of artificial intelligence. But we are sure that we will reach somewhere in it. At that time, we need to work together and think to welcome it. Gear Sensor has been widely used in the automotive and industrial field, which is important to the measurement of velocity, angel, angular velocity, direction of rotation. Gear Sensor,Custom Gear Sensor,Gear Sensor 3 Pins,Good Gear Sensor Yuheng Optics Co., Ltd.(Changchun) , https://www.yhenoptics.com