The development of AI can't wait for Ma Yun to call the best CEO or robot in the future.

It is better to do a lot of things, or work, to the robot to complete, such as dangerous handling, high-altitude work, etc., and a lot of boring and repetitive work to the AI ​​will be easier to operate. For example, these reviews of Google have greatly encouraged the innovators to increase the research and development of artificial intelligence. Ma Yun is the best CEO or robot in the future.

According to Wired magazine, there are such a group of people in the United States whose daily work is to open a computer to watch Youtube videos. These people specifically look at the video for content such as violence and hate speech, and accordingly make the video "offensive" or "sensitive" and other different ratings. They are Google's so-called "ad quality rating staff", helping Google make up for temporary workers with inadequate machine algorithms. Today, when the automatic algorithm is not perfect, Google needs the help of these manual rating staff.

Google’s video site Youtube serves millions of ads every day. Google uses an automated system to arrange ad serving, and advertisers don't know or decide which video to show with their ads. However, this uncertainty has recently caused trouble for Google. Many reports show that Youtube is actually advertising on videos that promote hate and terrorism. For this reason, advertisers such as Wal-Mart, Pepsi and Verizon have abandoned the platform, and Google's broader advertising network has also experienced a crisis of trust.

Google responded that the media exaggerated the advertising scandal. A company spokesperson said that ads with extreme video impressions were less than one-thousandth of the total number of advertisers’ impressions. Philipp Schindler, vice president of global sales and operations at Google, also said that the advertising problem involves only a very small part of the video. But the advertising rating staff revealed that Google is preventing the problem from worsening by increasing the team of artificial rating staff.

Advertising revenue accounts for 90% of Google’s total revenue, so finding a video of the problem and putting an end to the advertising scandal is especially critical. But nearly 600,000 hours of new videos are uploaded to Youtube every day, requiring a city's population to watch around the clock to ensure full manual review. This is obviously unrealistic. Google said it is working hard to develop artificial intelligence to help identify problem videos. "This problem cannot be solved by manpower and should not be solved by manpower," Schindler said in an interview with Bloomberg.

The problem is that AI training is inseparable from labor. Therefore, at this stage, Google still relies on manual grading to identify and mark offensive videos to build a database for AI learning. Eight current and former rating staff told Wired magazine that Google once became increasingly dependent on the work of the manual rating staff. The lack of communication with Google and the lack of stability in their work have jeopardized their ability to do their jobs.

Manual rating

Technology companies have a tradition of hiring content administrators. As people upload and share more and more content online, this position is becoming more and more important to the Internet giant. The rating staff interviewed by "Connect" said that in addition to reviewing the video, they will also review the comment area and mark rude interactive messages. The rating staff examines the various websites covered by the Google advertising network to ensure they meet the company's quality standards. They sort websites by category (such as "retail" or "news") and then click on the link in the page ad to see if it works. And, as the name "Ad Quality Ratings" shows, they also rate the quality of the ad itself.

After the March advertising scandal hit the customer, Google asked the raters to suspend other work and focus on "high priority rating projects." This new project means almost exclusively focusing on Youtube, checking videos and even entire playlists based on customer complaints. “This is a huge shift,” said an advertising campaigner.

The rating staff said that the huge workload means that the quantity and speed are placed at a higher priority than the accuracy. In some cases, the system will ask for a video that has been reviewed for a few hours in less than two minutes. Overwhelmed rating people have exchanged various efficiency-enhancing methods on anonymous forums. For example, check the lyrics of Rap MV to quickly search for 亵渎 words, or quickly zoom in and out with 10s. Google uses a timer to monitor the review time of each video, which also increases the pressure on the rating staff. "I am worried that if I review too many videos at once, I will be fired." A rating officer told Wired magazine.

Human-machine relationship

Advertisers interviewed by Wired magazine said that hiring temporary workers to take on the repetitive tasks required to train AI systems is a practice adopted by big companies across the technology industry. One rater talked about his involvement in the evaluation of the Microsoft Bing search engine in the early years, when the manual reviewers needed to browse up to 80 pages of search results per hour. Rating staff revealed that LinkedIn (data annotations) and Facebook (evaluating the fan page sponsor's posts) also hired temporary workers for similar tasks.

(Microsoft declined to comment. LinkedIn cannot confirm the existence of such a project, and Facebook did not respond to requests for comment.)

The overall instability of the work of the rating staff and the widespread involuntary dismissal have made many current and unemployed people dissatisfied. They said that the instability of their work will also cause behind-the-scenes employers such as Google to lose the accumulation of skilled employees. “They spend their time training new employees and then rushing them out,” said a former employee.

However, for technology companies, the constant use of new temporary workers for manual evaluation may be the best way to improve AI intelligence. AI experts say that regular replacement of manual data sources is more suitable for training AI. “AI needs a lot of observations, especially in terms of content that is offensive,” said Jana Eggers, CEO of artificial intelligence startup Nara Logics. She also quotes famous stories of the Supreme Court and pornography. The federal judge said that he did not know how to define pornography, but "I know it when I see it." By the same token, giving the machine more perspective will yield better results.

Although AI researchers generally believe that the trainer's bad habits will not be transmitted to the AI ​​along with the training data. But beyond the working environment and personal experience, some of the more subtle effects are hard to control. “People often hear the idea that a lot of different data is the way to train AI models,” says Cornell University AI professor Bart Selman. “This is a good guideline, but when it comes to moral judgment, the deep-rooted prejudice in the group also needs attention.” For example, people tend to think that men are better at certain jobs than women (or vice versa). “If you use data from a specific group to train AI, the hidden prejudice in the crowd will also be learned by AI.” Similarly, using a group of anxious temporary workers to train AI, temporary work-specific bias may also be absorbed. Into the system.

"You don't want to train the AI ​​ethics module by letting the machine observe the way ordinary people do in their daily lives," Selman said. “You want to get data from people who have deeper thoughts about prejudice and ethical issues.”

Casual suits, white shirts, black trousers... Ma Yun sat in a group of entrepreneurs and guests, thoughtfully playing with the signature pen in his hand - this is April 23 at the Chinese Enterprise Club in Zhengzhou, Henan. The scene that was seen at the annual meeting of the China Green Company. Ma Yun is undoubtedly the focus of the venue. In order to satisfy the curiosity of the participants, the conference organizers deliberately extended Ma Yun’s speech time. Unlike the occasional disappointment of the audience, Ma Yunjian walked onto the podium, and after a word of thanks, he went straight to the subject.

"The CEO is mainly two things, one is to look at the future opportunities; the second is to look at the future disaster." Ma Yun shared with more than 100 entrepreneurs. As one of the best entrepreneurs, Ma Yun goes straight into the game, indicating that in the next 30 years, the changes in this world will exceed the imagination of all people, and most people will be unlucky.

Ma Yun also did not forget to seize the opportunity to attack the real and real disputes. "The real economy has never been done well." He said that the hardship of the real economy stems from the conservative application of new technologies. In another 10 years, including after 90s and 00, there are 5 billion to 6 billion people in the world who use the Internet. "Do you think about the use of today's conflicts?" Ma said.

At the beginning of the game, it will "play chicken blood" for young people.

"Young people like us, there are no rich and powerful parents, there is no relationship, we have no accumulation yesterday, no resources today, the only thing to do is to judge the future" - opening, Ma Yun His own experience made a shot for the young people at the venue.

As the CEO, Ma Yun believes that the two abilities are the most important: one is to look at the future opportunities; the other is to look at the future disaster. There is no way to motivate employees when they don’t see the opportunities in the future; it’s impossible to avoid risks in the future without even finding a once-in-a-lifetime business opportunity.

Ma Yun said that he often thinks about the future. "If you don't use the Internet to develop your own business in the next 30 years, if you don't use the Internet to develop your own business, it will become as terrible as the lack of electricity 100 years ago or decades ago, even more terrible than breaking the electricity. 14 years ago, Ma Yun saw the opportunity for the next 30 years, and today, Ma Yun reminds people.

At the annual meeting of China Green Corporation, Ma Yun once again explained the "5 new theory" and reiterated that new retail, new manufacturing, new finance, new technologies and new energy will have a huge impact on all walks of life. When it comes to new finance, Ma Yun recalled that he once asked him if he thought that Alipay would make money and how to develop. His answer was "I didn't want to make money, it was a request, I hope Alipay can make any old lady." Power is the same as the power of the chairman of the bank."

I noticed that Ma Yun gave such an example: the old lady next door said that she had to go home late at night to cut off the electricity. The electricity bill forgot to pay early in the morning and went to the bank to line up for payment. The big president and the big bankers called the secretaries to pay. money. Therefore, he feels that if anyone clicks on the mobile phone, they can pay the bill, which is called new finance.

People regard Ma’s judgment on business as a standard. If you come back today and become an entrepreneur with nothing, will Ma Yun still insist on the Internet?

"If you start again, I may not be rushing into the Internet industry today because there are too many smart people in the Internet industry. Today, traditional entities have the best opportunities, low cost and low cost, and they are combined with the Internet (the opportunity is greater). Ma Yun said that if he came back, he would definitely rush to the line.

The economy is false, but it should not be opposed

Ma Yun used Alibaba to enlighten China's virtual economy, but it was difficult to get out of it in the "virtual and real dispute."

At the annual meeting, Ma Yun once again attacked the view that the virtual economy and the real economy are opposite. "The world's real economy is getting harder and harder. I think the real economy has never been done well. How many enterprises in China are working hard, leaving a little bit of entrepreneurs, and the hard work they have done is beyond ordinary people." Ma Yun lamented that Only China in the world completely opposes the real economy and the virtual economy.

Speaking of the dispute between reality and reality, Ma Yun is somewhat angry. "I often hear a sentence on the Internet called the entity to serve the country. I always feel that this sentence is very awkward. Entity and virtuality should not be opposed. This perfect combination is the real economy in the future." For the dilemma faced by the real economy, Ma Yun It is thought that backward thinking is elusive, "Thinking not to embrace technology has impacted you."

However, Ma Yun also stressed that the eliminated real economy must be eliminated. "Not all real economies are good, and not all virtual economies are bad." "In the past, especially in the past 20 or 30 years, we have turned people into machines. It is particularly obvious on the assembly line. In the next 20 or 30 years, we will turn machines into people, and machines will do a lot of things. Ma Yun said that every time technology hits employment, it also innovates employment.

Nowadays, artificial intelligence becomes a hot word. There is a definition of artificial intelligence: artificial intelligence is a machine that replaces what people can do. In this regard, Ma Yun said that he did not agree. His view is that he must find ways to make the machine do things that humans can't do, so that there is an opportunity for the machine to become a human partner forever, not the biggest opponent of mankind.

Ma Yun even joked that 30 years later, on the cover of Time magazine, the CEO of the year is likely to be a robot. "It remembers that you are better than you, and you are not angry with your competitors."

At the end of the speech, Ma Yun called on entrepreneurs to pay close attention to "30": first, focus on the next 30 years; second, focus on companies with less than 30 people; third, focus on young people around 30 years old. "For entrepreneurs, the world for the next 30 years is not an Internet company. It is a company that uses the Internet," Ma said.

Coaxial Cable

Rf Coaxial Cable,Digital Coaxial Cable,Coaxial Cable For Internet,Coaxial Cable Types

Zhejiang Wanma Tianyi Communication Wire & Cable Co., Ltd. , https://www.zjwmty.com