The ultimate evolution of search engine to artificial intelligence

　　

K.K in the documentary "Google and the World Brain," he asked Larry Page in the early days of Google start-up, now has a good performance search engine, why do one? ' Instead of developing a new search engine, we're going to do artificial intelligence, ' Larry page explains. Search engines are the most important connection between our initiative and the world of the Internet, and have a unique advantage in long-term data accumulation, storage technology, cloud services, supercomputing and machine learning, and perhaps they will become the closest Internet applications to artificial intelligence.

For now, Google's founder's desire has come true – Google has become the world's most successful internet company in artificial intelligence. Domestic search giant Baidu, also in the last two weeks of technology open Day detailed introduction of the "Big Data Engine" program, which refers to data mining and machine learning technology, the specific application of large data, as well as the grasp of future trends is equally exciting.

　　First, the prototype of brain-computer interface--Search engine

The Brain-machine interface (Brain-computer Interface, also known as Brain-computer interface) focuses on how we can interact directly with external machines by thinking, by building a unified code that links the brain to the machine, and realizing the ultimate dream of "controlling machines with ideas" is an important technical support for artificial intelligence. The movie "The Matrix" and "Avatar" use consciousness to control the "Avatar" of the virtual world; Russian "Global Future 2045" predicts that scientists will use the brain-computer interface in 2020 to implement the robot with consciousness control; Ray Kurzweil that 30 years later, humans will be able to upload their own consciousness to the computer to achieve "digital immortality", these ultimate artificial intelligence are through the brain-computer interface to achieve.

Those dream version of the goal is too far away from us, for the vast number of Internet users, from our recent brain-computer interface class application is a search engine, brain-computer interface is essentially for us to provide a kind of active awareness and external environment to interact with the way, from the source of consciousness-the brain, With the outside world to achieve the most direct contact and interaction. The similarities between search engine and brain-computer interface are mainly embodied in two aspects.

1, the active connection of consciousness to the outside world

The brain-machine interface transmits our initial conscious motivation-the nerve impulses-to the outside world and gets feedback. And when we use the search engine, the same is the first generation of awareness, and then converted into keyword input to the search engine, the latter through the search to our awareness and external information matching, after repeated debugging, for us to find an accurate answer. Therefore, the search engine carries our initiative to contact the outside world's will, through the valuable feedback solves our question, has fully manifested our consciousness initiative.

This may also be the essential difference between search tools and other Internet services, and when we browse the web, watch videos or view social tools, we are in the position of a passive recipient of information, and what we see depends entirely on what the other person is showing. Using search tools is an initiative, we need the brain to think and manually input, but also based on the information to be adjusted until the final search results are consistent with our original motivation. We can use other Internet services without thinking, but can not do the network search without brains, when we enter the keyword into the search box This behavior occurs, it has manifested our initiative awareness. Because of this, search engines need more "brains" and "hands" than other Internet tools, and these relatively cumbersome operations are actually the transformation of our consciousness into behavior.

2. The process of information transmission and interaction

The brain-machine interface, which is based on mastering the principles of the brain, establishes a unified interactive rule between the brain and the external object, making the brain perfectly docking with the outside world, with the ultimate goal of providing an idealized conduit for information transmission. Our nervous system can also be seen as a "computer" structure that communicates information through bio-electricity. Interconnected neurons communicate through electrical signals and transmit information through neurotransmitters, forming multilayered, functional neural circuits and electronic circuits. Therefore, electrical signals become a communication medium that is accepted and communicated by the brain and external intelligent machines.

For the search, the key words as we and search tools between the media, with the help of the keyword this "compiler", we have realized the communication between the self-consciousness and the search engine, forming a consciousness generation, keyword conversion, search, information acquisition, motivation to meet the information closed loop.

In short, the search tool embodies our awareness of the initiative, but also for us to create an effective way to interact with the outside world, in this respect and brain-machine interface is not the case. Although it is in the technology and efficiency and the ideal brain computer interface There is a huge gap, but already in our role as a smart assistant. Therefore, we can use the current widely used search engine as the future advanced brain-computer interface of the embryonic or initial state, in other words, the current search engine is currently the most promising to evolve into a brain-computer interface, the realization of artificial intelligence of the Internet service form.

Ii. the necessary factors for the evolution of the search engine to artificial intelligence

Search engines with similar human-machine interfaces have natural advantages, but the following conditions are required if they are to go farther on the road of machine learning and artificial intelligence.

1. Focus on strategic data rather than "big data"

If the artificial intelligence is to be compared to the brain, the Internet is the human head of the blood vessels, a network to reach all parts of the system, and data and information is the most important, can provide energy for the brain activities of the blood, they are necessary for machine learning and the input and research base. Baidu's "Big Data Engine" structure of the bottom of the "open cloud" service is to achieve this function.

In the data acquisition, can not aimless pursuit of all-inclusive data, because large data pursuit of the "n= all" full sample is not possible, while large data samples can not solve the problem of sample deviation, but caused a large number of small data problems. Search engines should pay more attention to the filtered and valuable strategic data. For example, the recent MIT Big Data Challenge was the theme of Boston traffic; Google's big data function was to start by predicting the flu, and Baidu's big data engine first chose the medical, transportation and financial sectors as the specific directions for big data strategies. Search engines have accumulated enough on the PC side, and with the outbreak of mobile phones and other wearable smart devices, search engines will get more strategic, finely differentiated and situational data. "Baidu Migration", "Attractions Comfort Forecast" and "urban tourism forecast" is Baidu to fine differentiation of strategic data as the breakthrough point of the success of the case.

2, the search engine data mining advantage

When the age of information explosion comes, it's no longer important to have information, it's important to find the information quickly, and search engines have a natural advantage in this area, and the only way to solve the contradiction between the quantity of information and the efficiency of information (valuable), and search engine data mining will have more obvious effect. For example, Watson, who defeated humans in intelligence programs, was able to learn because of "massive data plus search matching".

The astonishing efficiency of Internet search contrasts with the traditional way of information search, therefore, each rational person in the acquisition of information will give priority to the use of Internet search tools, in the enjoyment of the efficiency of the search to contribute personal information, so that the search tools to better understand us, so that next time to provide us with more efficient, more accurate, More personalized search services, as information acquisition speed and accuracy of the increase, we will be more frequent use of search tools to contribute more personal data ... This forms the "flywheel model" that Jeff Bezos put forward, a closed, benign cycle.

This virtuous circle brings the continuous growth of user habits and data volumes to the search tool, and more importantly, when the data accumulates to a certain extent, the search engine will master enough user's personal label and industry label to form a precise definition of the user and industry after the multi-dimensional information integration. So that through their own data accumulation and mining for users and industries to provide more intelligent information services. And this process should be the "Big Data Engine" meeting mentioned "the critical point of technological change", but also vice President Wang Jin mouth that "Baidu years of deep technical accumulation of a qualitative process." The definition of individual tags is a standard data element, the process of labeling all users with standardized labels for statistical analysis is referred to the "quantification of self." And Baidu's "Big Data Engine" related to the owner also mentioned in addition to the search tool on the PC side of the accumulation, mobile phones, wearable smart device growth also accelerated the volume of data growth, which will contribute to "quantify the self" of more digital labels.

Therefore, the advantage of search engine in data mining is not only embodied in its own algorithm and computing ability, but also in the quantitative analysis of user information and data mining, and on the basis of the more intelligent information services provided. It is worth mentioning that these valuable tags can only be obtained after a long period of accumulation and long-term user precipitation. Just as people and people only in the long-term communication can be in-depth understanding of each other, the machine to the user's understanding is also based on a long-term, countless times the basis of interaction, and the user's understanding of the degree of the decision of the core algorithm machine learning. As a result, the search engine, which already has a leading position in the market, will have a more obvious advantage in this respect.

3, the emergence of technical singularity

After the search engine collects strategic data and makes data mining, the next step is to realize the last phase of AI-machine depth learning. Search engine to realize the transition from traditional search to artificial intelligence, not only because of its accumulated data quantity reached the level of qualitative change, but also because of hardware storage, cloud computing, supercomputing, Analog neural network and other related technologies mature.

Search engines can be the leader in this technological change by virtue of their unique advantages, and they have already started the relevant preparations. Google has long been able to assemble the necessary technologies for artificial intelligence through its own research and acquisitions, including a learning-enabled Google Brain that uses 16,000 of processors to simulate the human nervous system; The acquisition of DeepMind machine learning technology applied to sound and text search to better understand the human natural language, the machine to learn the leader Geoffrey Hinton incur under the command. And Baidu is the first domestic machine learning (Baidu brain) and the formation of relevant research institutions (IDL, Baidu in-depth study institute) of the Internet company, proposed the "less handsome plan" to recruit young scientists, has a super computing capacity beyond the Tianhe No. second; The world's largest 20 billion-parameter deep neural network has been formed. These technologies, together with the data advantages accumulated by the search engines themselves, have become an important driving force for them to enter the field of artificial intelligence.

4, the correct way to learn the machine

Before this article "latent strong undercurrent: The Ultimate Intelligence Group Evolution Path" has mentioned that leads to the ultimate artificial intelligence to have two paths, first, embarks from the theory research, after completely grasps the human intelligence principle to reproduce through the technology; Secondly, from basic technology and concrete application, The corresponding level of artificial intelligence is obtained by combining the research results of each stage of human brain, and then the intelligent level is improved through the long technological evolution.

If you take the first approach, it does not make any sense to simulate the brain structure until you fully understand the brain's principles. Therefore, this approach is not suitable for users to continue to provide services to the search engine companies. Andrew Ng of Stanford University in California has completed the world's largest artificial brain with 11 billion neural connections, but two researchers at Google's brain engineering point out that the neural network computer did not even reach the mouse's intelligence level.

The European Union's human-brain program, which lasted 10 years and cost € billions of, did not aim to fully grasp the principles of the human brain and simulate them, instead, it has been emphasized that the results of each phase should be put into practice, by imitating the principles of some human brain to create intelligent systems or applications of some kind of human brain function (human-like). And for the user-oriented search engine is even more so, they should choose the second bottom-up, distributed intelligence implementation, each technology is the intelligent innovation from the actual needs of users, each technical progress represents the further development of machine intelligence. Google, after acquiring DeepMind, has made it clear that it will not first apply it to the Robotics department, but start with the basic semantic recognition. And Baidu is the depth of learning technology used in the specific user services, such as reducing the relative error rate of voice technology, improve the Chinese speech recognition rate, improve the image recognition capabilities, the world's first full web face search engine.

Above mentioned, the search engine constructs the human and the information consciousness active connection, moreover the user and the search engine each interaction is one kind of mutual understanding, grows together the process. Unlike other Internet services, users are also involved as a "developer" when using search engines, which is what the president of Beihang University said on the opening day of the "Big Data Engine" in Baidu, "a new way of software development that transcends crowdsourcing and integrates computers and people".

5. Machine learning based on human cranial nerve network

The biggest difference between the human brain and the computer is that it is a two-way connection, can keep the new information feedback back, strengthen already have something, this is our learning function. The levels of our brains are not born, they are inherently unrelated modules, there is no pattern between each other, and all the connections and enhancements between modules are realized through learning.

Our brain's learning and reasoning ability is in line with Bayesian mathematical principles, reasoning is based on the probability of experience, when we encounter new situations, we always have to go back to their own memory and experience, and then adjust according to the new situation, and finally come up with a relatively high probability of coping strategies, That's why we've learned how to deal with this probable world. In the implementation of the "Bayesian principles" in the study, we will first encounter the new situation in the previous memory and experience of the search and match, to find a higher probability of success of the decision to put into practice, and then adjust according to the actual feedback, and then repeat the process, until the best solution to find. After such a repeated action of Bayes, we constitute a specific self, but also formed the kind of important learning ability.

At present, advanced computer technology can make Bayesian computing method can be more effective implementation, "The most human" author Christian believes that the computer will be a change of 1-0, not true or false scientific logic, can understand and use Bayesian probabilistic reasoning. After recognizing the learning principles of the human brain, the use of digital "neural network" to simulate the brain and independent collection and feedback of information has become a breakthrough in the field of in-depth learning, for example, robot expert Hod Lipson robot after observing and touching the pendulum all day, through learning to calculate the F=ma , and it took humans thousands of years to come up with this formula.

And the search engine that "keyword input, database search, information matching, multiple adjustments, return results" is exactly the same as the Bayesian learning principle? Therefore, we have reason to believe that the search engine through the construction of analog neural network has been able to obtain a certain degree of human learning capabilities, such as the current "Baidu Brain" has reached the level of intelligence of children aged 2. In the American Academy of Pediatrics, children are called the "scary 2 years of age" in the "Parenting encyclopedia", his vocabulary began to abound; he was able to visualize the image of things, deeds and concepts in his mind; he began to understand the relationship between objects; He began to connect different activities in tandem to form a logical relationship; when he realized that society wanted him to follow certain rules, , he will begin to develop a certain degree of self-control. Although the "Baidu Brain" may not be able to match these intellectual manifestations, but we still feel the great progress of machine intelligence.

6, a more open search engine platform

As for the research topics such as machine learning and artificial intelligence, such as long period of study and large span of discipline, it is necessary to maintain an open mind and a willingness to cooperate actively. The European Union's human brain program has been emphasizing the establishment of open platforms, the opening of interfaces and the integration of multidisciplinary and so on. For search engines, openness has always been its unique advantage, and it can cover almost all users and industries. But the long-term goal of machine learning and artificial intelligence also needs to be kept open by search engines, we can also from Baidu's "Big Data Engine" strategy to see its importance to open and cooperation, it will be large data storage, analysis and intelligent processing, such as a set of core competencies through the platform, interface to open the way, In order to attract more partners, with the Beijing Aerospace University for Scientific research, and transport, finance, medical industry, such as in-depth cooperation. The open platform can not only make the existing machine learning function more extensive value, but also can verify and perfect the existing machine learning model through more application and the introduction of external resources.

Third, the search engine transition to artificial intelligence worries

1. Privacy of personal data

We can be indifferent to artificial intelligence in the laboratory, because no matter how far the research progresses, it will not be directly related to us. But we have a completely different attitude towards the search engine that we use every day, the reason for the search engine's artificial intelligence is that we have countless interactions with each other, and our intelligent definition is that it can help us find the information we need faster and more accurately, because it is so intelligent because we know more about our needs, and its understanding of us is based on the increasing amount of personal information we provide to it, which, of course, means less privacy.

Search engines know our travel routes, geographic location, job information, day-to-day behavior patterns, and social circles, and they know our risk situation better than any insurance company or bank, and as wearable smart devices rise, it may be more aware of our own physical condition than doctors. Or, search engines will become more aware of themselves than we are.

This is the unique background of the information age, the pursuit of efficiency so that we will inevitably enjoy the Internet search engines and other services brought to our information services, but also the inevitable enjoyment of personal information leakage of distress. We have observed that the changes in the world brought about by the increasingly intelligent and humane information services have brought about an increase in our efficiency, but we have neglected that as our personal information becomes more public, our actions are being perceived. Search Engine machine learning is bound to need more and more user information, which has a fundamental conflict with our privacy. Perhaps we have realized this, but there is nothing we can do about it in the face of efficiency.

2, we give the memory to the search engine

"Shallow" mentioned that the internet is shaping our brain, all the while changing us, and search engines are the first to seize our memories. For thousands of of years, humans have relied on each other to memorize the minutiae of everyday life. Now, we are relying on electronic devices and "clouds" that are changing the way we perceive and remember the world around us. The internet may not only replace the "other" kind of external memory resources, but also replace our own cognitive faculties. The Internet not only eliminates the need for us to share information with our peers, but also disrupts the urge to store important information that is immediately acquired into a biological memory system. This is called the "search engine effect".

If we think of our memory as a database, each memory store is to get the information we need at some point in the future, and our every memory is a search in this database. Just imagine if a more intelligent search engine knows us better, when we search for information over the Internet faster and more accurately than we can in the memory of our brain, we certainly do not hesitate to choose the former, whether it is a part of the "memory" that exists in the external Internet or in its own brain.

There is still debate about the pros and cons of this outcome, one idea is that after we share some of the memory with the search engines, they start to think that they know more than anyone before, but in fact, the reliance on search engines just shows how little they know about the world around us, and we completely take the product of Google's algorithm as its own. Know "of something." Another opposite view is that as we become a member of the Inter-mind, we develop a new kind of intelligence that no longer relies on local memory in our brains. When we are liberated from the need to memorize everyday facts, we can use this part of our free mental resources to achieve our individual ambitions. This evolving "interconnected brain" may combine human ingenuity with the abundant knowledge of the Internet, enabling us to break through some of our own manufacturing difficulties.

3. Transfer of Ultimate Intelligence

Although the mystery of consciousness remains unresolved, at least we already know that our consciousness and our thinking are rooted in our neural connections, and the information carried by the neurotransmitters passing between the neurons constitutes ourselves, as Sebastian Seung said "I am my connecting body." The emergence of neuroscience is erasing the "soul", there is no one in the so-called true self, we have only our own brain, which means that we are only the brain in a specific time to show the state, and this state can be reflected by the information. Of course, we can also think of the neural network of the brain as a software that runs on a Hinton. Or, in essence, our neural activity is a transmission and computation of information. There is a line in the matrix that makes the most incisive generalization: "What is real? What is the definition? If it is the sense of touch, smell, taste and vision, these are all signals and messages that the brain accepts. ”

When the various Internet behaviors that embody our consciousness are quantified, counted and regroup by search engines, does this mean that they can grasp the intelligence that is inherently ours? They simulate our intelligence by machine learning, and combine them with the powerful function of being born superior to humans, It can even solve a problem that human beings will never be able to attain. Besides mentioning the application of large data and machine learning in many key industries, Baidu's "Big Data Engine" also mentions the use of machine learning to get a new understanding of the outside world and discover the laws of nature that we have not found. If this happens, our greatest professional philosopher, scientist and artist may have to be in search engines.

Search engines Learn to master human intelligence through machine learning, though not a predatory one, it may be a transfer of control. We have evolved from natural evolution to technological evolution, and technology has brought about the hyper-human doctrine. Internet services, such as search engines, have turned us into cyborg, and the evolutionary direction of artificial intelligence blurs the boundaries between the human brain and the machine, and we can take it beyond the limits of our own cognitive, but this change does not mean that we are in danger of losing ourselves. We are building a powerful, unprecedented connection with the search engine that moves towards intelligence--we just integrate ourselves into a greater thing.