The father of Google Brain: We should worry about AI now

Mercedes trucks can run on the road themselves. Google's computer won over the go-go grandmaster. App appearances can translate sentences with accuracy similar to humans. Those are some notable milestones (artificial intelligence (AI) achieved in 2016 and experts in the field think that AI will continue to change our lives in a way dizzy.

However, many of the world's technology leaders have now expressed concern about the development of AI. Typical as Tesla's CEO, Elon Musk once said, AI can be the "biggest existing threat" to humans and establish a research team to find ways to keep AI technology. always open and useful to the community.

Mr. Andrew Ng.

Notably, Andrew Ng , co-founder of the famous "Google Brain" AI project and currently working as director of scientific research for China's Baidu technology firm, said that many of the risks immediately precede AI did not receive proper attention. "I think the government and society are distracting about moral issues will encounter when using AI," Ng said, "And we should not justify by saying those things will only happen. hundreds of years later ".

Talk a little bit about the "Google Brain" project. This project is part of Google's AI research effort to apply to image search. The main goal of "Google Brain" is machine learning, natural language and many other technologies to create a network of AI capable of finding and learning from things. available. After retiring from the "Google Brain" project , Andrew Ng moved to Baidu's AI research lab in Silicon Valley (USA).

Recently, Mr. Ng had an open conversation with Time's reporter about the future of AI, the importance of moral issues related to AI development and many other things.

Computers are chatting better and better

Besides smart speakers like Amazon Echo or chatbot on platforms such as Facebook Messenger, technology companies are also working hard to make computers talk more naturally with people."I think 2017 will be the year of conversation computers , " Ng said, "we have reliable data to think this will come."

Ng also believes that the key to making a voice-controlled computer in the future is to create an AI software platform that can run on any device , similar to Android. "AI will not only be on a hardware device, it will be available on many devices in many different forms," Ng said. Now Ng's Baidu has created a platform called the Duer OS that can answer user-generated questions and connect third-party services in China.

Leaders of large technology corporations often express concerns about potential long-term AI threats.

The three main ethical issues when using AI

Leaders of big tech corporations like Bill Gates or Elon Musk often expressed concern about potential long-term AI threats. However, Mr. Ng affirmed that there are three main problems when using AI that we should worry from now on: the influence of AI on workers, the sharing of information between professionals. AI and honesty in AI software.

"I believe in the honesty of AI," Ng said, taking an example: If a startup works for carpooling or giving food using AI to predict shipping and receiving times, they will be provided with the most accurate prediction tool possible.

In addition, Mr. Ng also believes that openness is a very important thing in studying AI. AI researchers often publish their findings through scholarly materials to foster the overall development of the industry. However, the fact that the technical details are hidden makes their documents " go against the spirit of openness".

"If you want to publish data, you should do it with the disclosure of the knowledge you have," Ng said, "you don't need to share everything and there are many different ways to share. instead of just through the academic literature system ".

The impact of AI affects every industry

Many technology companies, from Google to Facebook or Microsoft, have emphasized their focus on developing AI in recent years. However, the effects of AI will not only be limited to the information technology industry. We can see, for example, that the IBM AI Watson platform has begun to encroach on the field of doctor diagnosis.

IBM Watson platform wants to rob a doctor's job?

"We are making AI have the same impact as creating the previous electricity," Ng said. "The availability of electricity has changed all industries: from transportation and telecommunications to manufacturing. I think we are at a stage where AI technology has grown to the point where it can make a lot of changes. industry " . Mr. Ng also predicted that AI will have a strong impact on the entertainment, retail and logistics industries.

Researchers hope that AI will soon be able to learn by itself

Many AI software is currently using a training program called "supervised learning." In this training program, researchers will use stretched data and find ways to get AI to produce the expected results. For example, to help AI understand a sentence, researchers will include data about the meaning and encoding of the sound of the sentence so that the AI ​​learns from it.

Now, however, researchers are aiming for an advanced technique rather than "unsupervised learning". Accordingly, the AI ​​will learn from the data without labeling it collected. This will help improve the learning ability of AI as well as save manpower because there is no need to label each data anymore.