Original Title: GMIS 2017 | Professor Berkley Stuart Russell: Past, Present, and Future of Artificial Intelligence
Originality of the machine heart
The heart of the machine editorial
The Global Machine Intelligence Summit (GMIS 2017) is the first conference organized by the heart of the global artificial intelligence industry information service platform machine. Many top experts from the United States, Europe, Canada, and China attended the conference. The conference consisted of 47 guests, 5 Sessions, 32 lectures, 4 roundtable forums, 1 man-machine battle, taking into consideration the academic and industry, technology giants and startups, and taking a professional and global perspective as the artificial intelligence practitioner. Both lovers and enthusiasts offer a feast of machine intelligence.
On May 28th, the two-day Global Machine Intelligence Summit (GMIS 2017) sponsored by the Heart of Machines entered the second day. The most important part of the day's agenda was the leaders’ summit attended by many important guests, including “Artificial Intelligence: Stuart Russell, author of a modern method, Yang Qiang, co-founder and chief scientist of the fourth paradigm, Hu Yu, CEO of the HKUST, president of the Consumer Business Group, professor at the University of Alberta, and Martin Müller, computer go top expert, Element AI co-founder Jean-Sebastien Cournoyer et al.
Stuart Russell is the founder and professor of computer science at the University of California at Berkeley. He has been focusing on the development of artificial intelligence. At the same time, he is the author of the standard textbook "Artificial Intelligence: A Modern Approach" in the artificial intelligence field. As the first speaker of the day, Stuart Russell led the audience to conduct an exploration of “The past, present, and future of artificial intelligence†to see and understand the mystery of artificial intelligence.
"For a long time, we don't know what artificial intelligence is. We have been thinking about what it should be." Russell said that artificial intelligence is an informative subject involving philosophy, mathematics, and nerves. Science and other disciplines and fields.
At the same time, artificial intelligence is not a new discipline. In the 1940s, people were thinking about how to use some new tools. In the 1950s, there were already many conferences calling for people from all disciplines to work together to create artificial intelligence. "But what happened later? They are all separate because all of these disciplines lack a common mathematical language." Creating a universal mathematics language to cover all disciplines is a difficult task, even though Can not be achieved, but people are moving in this direction. In the 1960s, there were already some visible developments in this area, and people were very optimistic; but by the 1970s, people gradually became disappointed; at the time of the artificial winter in the 1980s, many companies had begun to Losing money, artificial intelligence becomes a word that everyone does not like.
Now people are beginning to take a serious look at artificial intelligence and usher in the miracle of artificial intelligence explosions. Many start-up companies have begun to focus on the development of artificial intelligence, and giant companies such as Google and IBM have also invested in the research of artificial intelligence. At the same time, people have also seen progress in neuroscience and the development of computer resources and big data.
Russell then detailed the supervised learning method. When deep learning requires the use of a convolutional neural network, it is a type of supervised learning that is also part of artificial intelligence. The data here is tagged, and it also involves some hypothetical methods including linear models, logistic regression, neural networks, and decision trees, and then inferring future cases.
Russell used a simple example of object recognition to give everyone a clearer understanding.
So why do you have to do deep learning?
In 2015, the machine has already surpassed the performance of humans, and this intermediate ImageNet data set was indispensable. However, the machine also has inferior performance to humans in terms of visual performance. Various errors may occur, including object occlusion, distortion of the view, multiple object categories in a picture, or confusion of super categories and subcategories. .
In addition, Russell cited the example of AlphaGo's victory over Li Shishi. "AlphaGo contains a lot of technology, one is deep learning, and also uses Monte Carlo search, which can produce very efficient results. It can be traced back to some complex but classic search methods of the 1950s. On the other hand, artificial intelligence Developments in many areas have also contributed to AlphaGo's success, such as supporting deep learning hardware development.†Ke Jie, who recently lost to AlphaGo, once said that playing chess with AlphaGo last year seemed to be playing chess with people. This year he felt that It seems like playing chess with God.
Next, Russell shared some of his views on artificial intelligence.
He believes that although all the developments are very gratifying now, it is still a gimmick. People need to think carefully and do not feel disappointed because of excessive expectations. While everyone is proud of the tremendous progress made in artificial intelligence, it is also possible that the winter of AI will occur. “Our last AI winter was due to the prospects and promises of this technology. Many people may not remember what they were at that time. Some technologies around 1980 were not very effective when they were put into the real world. At that time, deep learning was not very popular, but now we can update modern technology, and it may not be that much higher in the future for training materials and data." In response, Russell said that if you let a child look at an elephant, The picture, he can identify the picture of two or three elephants for the child, instead of training him more than 4,000 pictures. Therefore, when technology is advanced to a certain degree in the future, people’s dependence on it is no longer that great.
The development of AI in the past 50 years has given people a big lesson. Knowledge is very important. Human beings can learn efficiently because at the beginning human beings have a lot of knowledge. Knowledge enables humans to complete learning from very few cases. Here, Russell mentions a probability plan and exemplifies the application of equal-probability planning for nuclear weapons test sites.
Russell also gave his opinion on the current, future and current problems of artificial intelligence.
"Although we have a lot of missing things, we can already foresee that, in the short term, AI systems can have the same capabilities as humans." Some people may think that this is a fantasy, but 1933 In the year, Lord Rutherford, a very famous physicist, once said that anyone who wants to extract atomic transformation ability is in a whimsical way. However, Leo Szilard invented the neutron-inducing nuclear chain reaction the next day.
Artificial intelligence can allow humans to do more things and push human civilization in a more positive direction. However, there is also the possibility of killing machines. It has become a new breed of weapons of mass destruction that has caused people's concerns.
The crux of the problem is that people need to ensure that the functions given to the machine really want it to have it, but it may also be that the functions that people give it are not what people originally expected. Russell raised the story of the king who turned gold to death after food, water, drinks, and even his daughter became gold. For this issue, Russell's opinion is to change the definition of AI, and the AI ​​system can be proved to be of benefit to human beings. In order to achieve this goal, there are three simple methods: "The first point, the goal of the robot is to maximize the realization of the human desire, that is to say, the machine should satisfy the human's wishes rather than let the machine create a kind for us. Humans do not feel comfortable living. Second, robots do not know what value is, we do not give the machine a fixed value system. The third point is that human behavior provides a reference for the machine."
At the end of the speech, Russell proposed two very interesting examples. The first example is about personal assistants who want to help people who need dinner more than they choose to help their hosts prepare dinner.
The second example is when the child feels hungry at home. There is nothing in the refrigerator in the home. The robot chooses the cat in the house as the ingredients to prepare dinner.
Is this robot good or bad? Where will future artificial intelligence go? Russell gave his own answer: "AI needs to contribute to mankind. To do this is a technical issue. I believe we can solve this problem."
For more information on the GMIS 2017 conference, please click "Read original" to view the GMIS theme
Editor:
1000W Digital Tv Transmitter,Wireless Tv Transmitter,1Kw Digital Uhf Transmitter,1Kw Digital Analog Tv Transmitter
Anshan Yuexing Technology Electronics Co., LTD , https://www.yxhtfmtv.com