People and machines will eventually merge into one

We are always looking for a relationship between people and machines.

However, when discussing this relationship, it is necessary to involve historical trends, especially what is inevitable. Only by understanding the trend, we must be able to begin to explore the relationship between man and machine. This is a philosophical problem.

Throughout the historical development, we have experienced technological breakthroughs from prehistoric civilization, agricultural civilization, industrial civilization to information civilization. In fact, the information civilization that was closest to us has not ended. It is the rapid development of information civilization driven by the development of computers and the Internet.

Even today, there are still a lot of scenes in China that are not informatized, such as education. However, there is already a sign at this time, which is the birth of intelligent civilization. When it comes to smart civilization, artificial intelligence is a topic we can’t get around.

Enlightenment brought by AI

Artificial intelligence has broad and narrow meanings. The narrow sense of artificial intelligence is about deep learning after 2012, especially the birth of the Alpha dog in 2016. Artificial intelligence has a long history, from the expert system of the 1960s to the feature engineering, and now, we have begun to use more supervised learning, especially for the processing of voice images.

After deep learning and supervised learning, what we are still doing is called reinforcement learning, or reasoning, but they have not yet reached the practical stage. The deep learning we talked about today still stays in the existing stage of voice image processing. Therefore, we are curious about the future trends.

I want to talk more about how the technology will change.

In fact, the biggest impact I gave to 2016 was the Alpha Dog, which we call an Enlightenment. Compared to the previous Renaissance, it is exploring the relationship between people and people, and between man and God. After the Alpha dog, we are concerned about the relationship between people and machines.

In the game of 2016, I participated more. I was witnessing Li Shishi being defeated by the machine during the live broadcast of Sina. In that game, the guest who was with me was Yu Bin, the head coach of the Chinese Go team. Before the game, he firmly believed that people will definitely defeat the machine, and even think that the machine can play Go is an unreasonable thing.

After the game, Yu Bin’s spirit showed a certain collapse. I heard that Sina editors helped him to take a taxi and he didn’t know how to take a taxi.

When a machine civilization was born, it was unacceptable for us, especially the experts, to be replaced by machines. This change will occur in various fields such as medical care and driving.

Before the game, we had a lot of questions about the Alpha dog. After the game, we found that many young friends called the Alpha dog a dog. It is very acceptable to it. The nine players of the Go game called it “A teacher.” Today, we accepted the machine in an anthropomorphic way. This acceptance will be faster.

Voice is AI entry

What kind of work will be replaced? This is a question that many media people are curious about. I think this is the answer to be solved in the near future. Another question is whether human beings will be replaced by machines. This is a more long-term problem. In fact, the most important issue I have been asked for forever in the three years from 2016 to the present is also the two things, not the technology behind them.

What kind of work will be replaced by machines? We need to understand the power of the current machine itself and its current limitations. In some media or film and television works, the machine is very powerful, imagine it becomes a human shape, can talk to people, and even more powerful than people. In fact, to build such a robot with today’s technical capabilities and level, I am afraid that it can only be in fantasy.

I use two points of view to illustrate the limitations of artificial intelligence itself.

The first point is that today’s artificial intelligence has a strong sense of perception. What perception? It is the perception of low-level processing of information after hearing, visual, and contact with the outside world. Now in image recognition, the face recognition of the machine can surpass people.

The other is the cognitive field, the machine will be much harder, especially related to the language. Language is an abstract understanding of the world, as it is in the brief history of mankind. We are looking for fairies in the forest. This kind of thing has a concept and a story. This thing is not mastered. Plus humans have the ability to migrate and learn, and machines don’t currently have them.

Under this limitation, we can make a simple assertion that the machine can only assist human beings in cognitive problems. For example, when writing a novel or dialogue, the machine can not work independently, but in cognitive problems, In the perception of the problem, the machine can be independently employed, just like the processing of images and voice.

Our approach is to focus on language-related perception and cognition. Perception is the processing of speech and images. Cognitive processing is dialogue, translation and question and answer. In this perception, the machine can reach the same level as people, but it will be very difficult in the cognitive field.

In terms of priorities, the easiest thing to do is machine translation. Even so, we can see that it has a certain gap compared with people. In particular, people think that things that are particularly simple, the machine does not necessarily handle well, and people think that complicated problems, the machine can do very well. This is not the same as perception. In terms of perception, machines can be more comprehensive than people do, and machines are not accurate, and people can’t do well. In the cognitive world, machines and people seem to be different types of people.

In the field of perception, we have done speech recognition. Sogou is the engine with the largest amount of speech recognition in the To C field in China today. Because we have the largest input method in China, there are more than 600 million speech recognition requests every day, including voice modification. The purpose is to use language as the core to make people communicate better with people through perception.

However, we feel that this does not represent the frontier of technology, and we have also recognized the lip language. Previously, we used sound to convert to text. Now, we can convert it into words with mouth shape changes.

In addition to speech recognition, there are speech synthesis, including style migration. On the Internet, in addition to the machine can recognize the user, when the user is expressing, the machine can do more auxiliary work. We can migrate our own sounds into personalized compositions.

We released a technology called Synthetic Anchor for AI. It is to completely link the relationship between voice, image and text. Before, we turned the mouth shape into a voice and a text. Now we are reversed, and it is the same technology that can turn the text into a sound expression. Our AI synthesizers can synthesize human voices, mouth shapes, expressions, and are the first in the world to replace people with AI.

AI’s synthetic anchor technology has been launched, and many media in the world will purchase it. The online customer service of the largest insurance company in China will also adopt our technology and replace some people with machines. In perceptual vision, machine and human communication machines can do a great job of replacing people and work independently.

Man-machine “new fit”

In the future, as AI technology develops and machines become more accurate, their assistance can gradually give the results of standard answers. It can be seen that from the language processing, our efforts are from search to question and answer, and later using voice questions can give direct results. Now, we can see that 40% of users can give a direct answer. From search to question and answer, the final form is to go to the personal assistant, let the machine and people generate the following communication, can help complete some tasks.

Whether humans really have to be replaced by machines is something we need to answer. There is a paradox, academically famous. Our perception says that the human brain is too smart. We can actually create a machine that is smarter than ourselves, but another person said that the human brain is too smart, we can’t make it. A machine that is smarter than the human brain.

These two arguments are true. When people are smart, whether the machine can surpass people themselves is a difficult question to answer. My opinion is that the artificial intelligence we make in the future is not a direct comparison with the human brain, nor is it to make the anthropomorphic machine a future for the development of artificial intelligence, but to take a different path.

In the field where people are best at, the design principle of the machine can’t be reached from the technology that is currently known; but in turn, these things that the machine is good at can’t be done. Just like a calculator, we don’t compare the speed of calculation with a calculator. At the same time, the use of artificial intelligence in the future will certainly far exceed that of people.

However, our universal intelligence, our vitality, and the ability to adapt to the environment, in the visible technology, the machine can not do. The basic point I want to mention here, first of all, don’t be arrogant. We think that the machine will replace people, because we can’t find a machine that is more powerful than human beings and can completely replace people.

Before, we had a narrow understanding. We were especially afraid that the machine would surpass people in a certain field, just like playing Go. In fact, we have to accept the function of the machine in many cases to replace the human part, or to enhance the function of the human part. For example, many people are wearing glasses, thinking about it, we have been invaded by technology, because glasses make your vision better, because each of us becomes a clairvoyant, easy-going ear. As in the future, technology and people will create new fusions.

Say to a monkey, would you be willing to change genes to turn you into people? The monkey is definitely not willing, because it does not understand or understand. Similarly, if we use technology to change people and turn people into superhumans, would we be willing? This is happening today. Most of us still adopt this kind of refusal attitude because we feel that people are even smaller. However, in the future, people and machines will have a new fit. With technology, we will become More powerful, forming a new ecology, this is the future of co-evolution of people and machines.

Our mission is to do simple work in AI, make expressions and information simple, and find your place in the big times.