This article was first published on the WeChat public account: New Wisdom. The content of the article belongs to the author's personal opinion and does not represent the position of Hexun.com. Investors should act accordingly, at their own risk.
|
Ji-won new compiler Source: beta.theglobeandmail.com Author: Jeff Gray Translation: Zhang Yi Wen Fei Liu Xiaoqin
[New 智元导读] Geoffrey Hionton, known as the "Godfather of Neural Networks", is a legendary life. Gao Zeng's grandfather was a Boolean who invented Boolean algebra. He had studied in Cambridge for a month and then transferred to the building, but only lasted for one day. After returning to physics, I felt that mathematics was too difficult. I went to study philosophy. After one year of study, I was full of antibodies to philosophy and had to practice psychology. After confirming that psychologists have no clue about human consciousness, this turned to Edinburgh to study AI.
Geoffrey Hionton, a professor of computer science at the University of Toronto, was born in the UK and is now known at the age of 69. Neural networks (neural networks) involve designing computer systems that mimic the human brain to enable them to learn. As some experts say, neural networks will completely subvert our lives – in fact, this has already happened – just as electricity overturned human life in the 20th century.
Over the years, Professor Hinton's field of research has not only been obscured, but has been isolated from the mainstream academic circle of computer science. His neural network is considered to be "weak nonsense" by unscrupulous people who use traditional methods such as manual programming to create artificial intelligence. Professor Hinton said that academic journals used to reject contributions to neural network papers.
But in the past five years or so, the students he has brought have made a series of amazing breakthroughs, and the neural network has become hot. He also became a leader in the calculation of a new era. Most voice recognition software on your phone works on neural networks. The neural network can identify pictures and distinguish the breeds of dogs with almost the same accuracy as humans.
And the world's largest technology company is investing millions of dollars in neural network research, hiring many former students of Hinton professors who are now responsible for or conducting AI research at companies such as Apple, Twitter, Google and Facebook.
According to Hinton, AI can range from driverless cars to smart phones that can diagnose skin cancer better than any dermatologist. The new wave of technology is expected to cause some damage to the industry and make people who develop and control AI technology make a fortune.
Professor Hinton spread his time between the University of Toronto and Google's Toronto office. At Google he is an engineering scholar and helps guide a new AI lab. He has just been appointed as the Chief Scientific Adviser to the newly formed Vector Institute, which will fund artificial intelligence research to transform Toronto into a global AI hub.
Despite all the achievements, Professor Hinton seems to be a defensive stance. The interview took place in his narrow Spartan-style office in Google. There was no chair in the room, and the whiteboard on the wall behind him was covered in the equation. He has almost no chills, and the struggle between talking about neural networks and more traditional AI methods continues.
He said that his own university, in the past, has been delaying the rehiring of new neural network professors, although it can get $1 million in revenue from Google. Now, with more research funding, many people have jumped into the field of neural networks.
“When you see the effect of the neural network, the industry and the government have begun to call the neural network directly AI. The people who have been laughing at the neural network all the way in the AI ​​field are now happy to say that the neural network is AI and do everything possible. Make money inside," said Professor Hinton.
The traditional concept of AI is to rely on logic and rules to program a computer to "think." Professor Hinton said that when the neural network was “destroyed†and smeared as an alternative in the 1960s, most of this work was theoretical and not implemented. The traditional model was considered correct at the time.
However, the breakthroughs of the past few years – partly because of the dramatic increase in computing power – have changed everything. In 2009, two graduate students at Hinton won a speech recognition contest using neural networks, using neural networks to defeat other more mature methods, which were later upgraded and incorporated into Google's Android phones. In 2012, his other two students easily won the image recognition competition ImageNet, the recognition error rate reached 5%, which is about the same level as humans.
To explain how neural networks work, Professor Hinton uses examples of translation programs. He uses neural networks as translators, which involves providing a bunch of words and word fragments to a computer network. The system calculates the meaning of a sentence and then sends it to another neural network, which then outputs sentences in another language, all without programming or language rules. Neural networks can even automatically learn the difference between active and passive.
"No one has told the neural network what is active or passive. Like a child, you won't say: "Look, Johnny, this is initiative, this is passive. "It's not like this. After a while, they naturally understand it," Professor Hinton said. "The same is true for neural networks. They are learning."
After a rotation at several universities in the United States, Hinton decided to come to the University of Toronto in 1987, when he considered two factors. One is the funding from the Canadian Institute of Advanced Studies for his AI. The other is more politicized: "I don't want to take money from the US military. Most of the US AI funds come from the military."
Hinton was born in Wimbledon and grew up in Bristol, England. His mother is a math teacher and his father is an entomologist who likes beetles. Hinton's grandfather was a 19th-century logician, George Boolean, who was also the inventor of Boolean algebra, the basis of modern computing. Hinton is what he calls a second-level private school (called a public school in the UK): "I am not particularly good at school. I like physics and football."
He first went to Cambridge to study physics and chemistry, but only lasted for a month, then transferred to architecture, this time only lasted for one day. Then he revisited physics and physiology, but found that mathematics in physics was too difficult to learn philosophy and squeezed the two-year course for a year.
Professor Hinton said: "That is a very useful year, because I have a very strong antibody to philosophy, so I want to understand how the mind works."
To this end, he changed his psychology, but the result only confirmed that "psychologists have no clue about human consciousness." He worked as a carpenter for a year before entering the University of Edinburgh Graduate School to study artificial intelligence under the door of Christopher Longuet-Higgins. Higgins students include Nobel laureate John Polanyi, University of Toronto chemist and theoretical physicist Peter Higgs.
At that time, Professor Hinton was convinced that the concept of neural networks would be the right way forward. But his mentor has turned to the traditional AI approach.
“My graduate career is like a storm, and there is a quarrel every week,†Hinton said. “I always said to the mentor, 'Okay, give me another six months, I will prove to you the role of the neural network. 'And then at the end of six months, I will say again, 'Okay, I am almost able to prove it, give me another six months'. Then I always said, 'Give me another five years. 'Everyone said, 'You have been doing it for five years and five years, and the neural network has never worked!' But in the end, finally, it worked."
He never doubted that the neural network would one day prove its superiority. He said: "I never doubted that because the brain must work in some way, and the brain determines that someone is not giving it some programming rules. Only work."
When asked about the typical "Will the robot take over the human world?", he agreed that AI needs some restrictions. He recently signed a petition requesting the United Nations to ban artificial intelligence deadly weapons. This is an organization called "The Stop Killing Robot Movement", which reads: "I think this is the most terrible thing. And it is not a distant future... but now!"
He predicted that the better future that AI will achieve is that doctors use neural networks to diagnose disease or skin cancer. AI will also be a human assistant, not only reminding you to make appointments on time, but also using "common sense" to observe your behavior. If you forget your appointment, it may decide to interrupt what you are doing.
Large banks, cable companies and many other companies are looking to use AI to analyze sales data and better interact with customers, Steve Irvine said. He left Facebook and returned to Toronto to create a startup called Integrate.ai to help companies achieve this.
"I don't think he can praise him." Speaking of Professor Hinton, Irvine said. "Because he stayed in the AI ​​field for days when he didn't see hope, it made him look like a crazy scientist. People never thought that AI would be what it is now. Now these talk about 20 or 30 years. Everything is happening. I think this is a good reward for him. Now the world is in hysteria, and he is the godfather. This is definitely not a overnight success."
Faux Leather for contract, contract vinyl, pvc leather for contract
Contract Vinyl,Contract Vinyl For Seats,Synthetic Leather For Contract,Artfical Leather For Contract
Vigor Plus Co., Ltd , https://www.vigorplusx.com