Rebirth of the big era of online games

Chapter 464 Boiling Cloud Computer

In the field of artificial intelligence research and development, whether it is in this life or in the past life, whether it is hardware or software conditions, China is currently far behind foreign countries.

At present, the domestic artificial intelligence level, especially the intelligence level of game AI, is still in the stage of "supervised learning".

The so-called "supervised learning" refers to a learning method in which humans create questions, tell AI the correct answer, and give corresponding countermeasures, similar to cramming education in primary and secondary schools.

Reflected in the game application scene, it is the kind of recognition that after seeing a human player, the human will take the initiative to attack, and at most some settings will be added. For example, when the HP is less than 30%, it will find a cover to avoid it, and when the enemy hides somewhere for a while If it is too long, a grenade will be thrown at it.

But these are very rigid procedural judgments. Strictly speaking, they do not have the characteristics of artificial intelligence, and they are not artificial intelligence in the true sense.

Internationally, as early as 97, IBM used a specially designed computer to win the chess championship.

Of course, the so-called computer at that time could not be called AI. The reason why it was able to win the chess championship was to use the brute force method, that is, the exhaustive method. All possible moves were calculated and then compared. Human chess score, choose the optimal solution.

This is the so-called "powerful miracle".

But when it comes to Go, there is no way to be so exhaustive.

No matter how great the power is, there will eventually be a limit. The possible moves of Go far exceed the sum of all atoms in the universe. Even with the most powerful supercomputing at present, it will take tens of thousands of years. Before the quantum computer matures, how many years will the electronic computer be used? It is impossible to perform such a task.

Therefore, the programmer added an extra layer of algorithm to AlphaGo: first calculate where to calculate and where to ignore, and then calculate in a targeted manner.

In layman's terms, the computer has the ability to learn, can perceive, judge and complete decision-making, which has the basic characteristics of AI.

However, the learning method of machines is qualitatively different from that of humans. By observing a few features, people can infer most unknowns, that is, they have the ability to "infer others".

But the machine can't, it has to constantly try to eliminate, and constantly identify and calculate, in order to get the result.

And in this learning process, three basic conditions are needed: algorithm, computing power, and data.

Algorithms are computing methods, which can also be called computing models. For example, recognizing a dog and recognizing a piece of human speech have different algorithms and different application scenarios. This requires programmers to design and optimize accordingly.

But with algorithms, AI cannot learn without computing power and data.

Just like "a smart woman can't cook without rice", the algorithm is the recipe, the computing power is the cooking tool, and the data is the ingredients. Just knowing the cooking method, without the cooking tools and ingredients, it is impossible to cook delicious food.

Therefore, in order to make achievements in artificial intelligence, the most fundamental thing is to solve the three basic conditions of algorithm, computing power and data.

At present, the research and development of artificial intelligence technology in the world is slow, mainly due to the constraints of computing power and data, but this is exactly the advantage of boiling.

After years of layout, the current three basic conditions and their extension conditions, although they cannot be compared with foreign countries, are basically available in China. Therefore, Fang Jie believes that it is time to make efforts in the research and development of artificial intelligence technology.

First of all, in terms of algorithms, to put it bluntly, there are two branch conditions: program development talents and algorithm training platforms.

There is currently no shortage of talents in China. The Boiling Research Institute, the Institute of Automation of the Chinese Academy of Sciences, Huawei, and Tencent already have a large number of scientific research talents in related fields.

And this time, Fang Jie directly used money to smash many Chinese experts working overseas back to China.

As for the algorithm platform, that is the old business, isn't it just games, in this regard, the Boiling Group has a natural advantage.

AI learning is learning through games.

Of course, AI games are slightly different from human games, that is, they are more targeted. The purpose is not to give AI a sense of accomplishment from the game, but to make it learn and improve, and help humans complete various tasks.

For example, there is a very simple game in the red and white machine: playing squares.

The gameplay is that the player slides left and right to control an elastic short board, catches the ball and bounces the ball, eliminates the rows of small squares on the top of the game interface, and clears all the squares to clear the game and gain a sense of achievement.

If this game is used as an AI training platform, through specific algorithms and reinforcement learning, the AI ​​will not be able to catch the ball at all from the beginning, and then not only can catch the ball, but also accurately calculate the rebound angle of the small ball. Remove all blocks above.

Even after a period of training and learning, AI can still find the best way to clear the customs.

For example, first accurately eliminate a few squares in the upper left corner, open up the paths above all the squares, and then hit the small ball in, so that the small ball is above the square array and bounces back and forth between the wall and the square array. The combo eliminates dozens of blocks, which greatly improves the speed of customs clearance and obtains more combo score bonuses.

Such game ideas and technologies may only be thought of and achieved by high-level players. At this stage, the technical level of AI in the game of playing cubes has obviously surpassed that of ordinary players, and its learning method is not "supervised learning". , but "reinforcement learning", AI has the ability to learn independently.

In the same way, if AI is put into other games, or if it is continuously learned and summarized in different algorithm models, then it can also train and cultivate highly capable AI.

In essence, training and teaching AI is not much different from teaching children.

Then on the other hand, learning through games is the best way of education and learning when human beings are young. It is not advisable to regard games as beasts. In terms of educational methods, self-reinforcing learning is also better than cramming.” Supervised learning" works better.

The algorithm problem is solved, and the next step is to meet the conditions of computing power and data.

Needless to say about the data, the game training platform itself is constantly collecting data. With the continuous improvement of AI intelligence, the speed and quality of data collection are also constantly improving. In actual application scenarios, what about the big data spread all over the network?
For example, in the future of artificial intelligence driverless technology, Boiling Cars and Boiling Travel are already collecting relevant data.

Why data is important is important here. Whoever owns big data can do AI. Otherwise, even the threshold of artificial intelligence will not be reached.

Obviously, Boiling Group and Fangzhou International are definitely the world's leaders in data collection, and their software and hardware foundations are better than other international Internet giants.

The collection, statistics, and processing of these data require massive computing power support, which involves the supercomputing center project.

After two years of layout and development, the Boiling Supercomputing Center has established more than 100 supercomputer rooms, with an investment of more than 300 billion yuan, but the computing power is still in short supply.

Among them, scientific research cloud computing and cloud Internet cafes are the two largest business segments of the supercomputing center at present, and the demand for scientific research cloud computing is the most.

The reason is very simple. Only the Boiling Supercomputing Center is the only supercomputing center that provides such powerful computing power and low service costs in China.

As we all know, doing scientific research is very expensive, and computer research is burning money mainly in manpower, hardware and electricity.

In terms of manpower, it is good to say that, after all, they are high-tech talents, and they are worth the price. To do related scientific research, this is a necessary investment.

In terms of hardware, according to Moore's Law, computer hardware is updated very quickly. If scientific research institutions are allowed to upgrade hardware equipment from time to time, the cost is too high and cannot be consumed at all, but if they do not update and upgrade hardware equipment, the energy consumption will not be high , the work efficiency is also low, and it is not a long-term solution.

Moreover, the issue of hardware depreciation must also be considered. Many hardware and electronic devices operate at full capacity for a long time, and their service life is much lower than that of home computers. The Boiling Supercomputing Center is building a new computer room at the front, and replacing old and damaged equipment at the back, so the operating cost is high. No less.

As for electricity, the entire photovoltaic industry is now at a huge loss. If it does not rely on state subsidies, it will not survive at all, let alone a supercomputing center that uses high-cost solar power.

And these costs are now borne by the Boiling Supercomputing Center, and scientific research institutions are of course rushing to snatch the wool of the Boiling Supercomputing Center.

It is precisely because of this that even though the business is booming, the Boiling Supercomputing Center that has been operating for two years is still losing money, let alone making money. .

This also prevented the bank from approving further loans to Boiling Supercomputing, and several shareholders who participated in the investment were not willing to continue financing investment. After the Boiling Supercomputer Center established more than 100 computer rooms, it began to lack success.

It's so difficult to do business. You can't fool investors by telling some stories and concepts. Only by showing solid performance can it be convincing.

If the supercomputing center is not strong, the artificial intelligence research and development work of each company can only be intermittent. Generally, during the late-night computing power valley period, relatively cheap computing power is used to train AI, but the computing power cost is still high, burning money Burns quickly.

Therefore, to make efforts in artificial intelligence research and development, Fang Jie, the real problem to be solved in the end is actually the problem of computing power, or the operating cost of the supercomputing center.

To reduce the operating cost of a supercomputing center, the most direct way is to increase the basic price of computing power, but doing so runs counter to the purpose of running a supercomputing center.

If the price of computing power is increased, then the cost of scientific research will also rise, which will naturally severely dampen everyone's enthusiasm for scientific research and scientific research progress. Some scientific research directions with unclear prospects, especially basic science research with little possibility of profit in the short and medium term The project is likely to be dismantled here.

The purpose of Fang Jie's scientific research center is to accelerate scientific and technological progress and innovation, so he cannot hinder the speed of scientific research for the sustainable and healthy development of the supercomputing center.

Therefore, in the middle of 2012, Fang Jie convened a large-scale seminar with the Feiteng Supercomputing Center, the Feiteng Research Institute, the Feiteng Technology, and the Feiteng Energy, as well as the senior shareholders and technical researchers of the affiliated companies.

The purpose of the meeting is to brainstorm and discuss how to generate income for the supercomputing center without raising the price of computing power, so that it can turn losses into profits, at least to ensure that it can continue to expand and develop while maintaining capital.

The seminar lasted for a full week, with more than 1000 participants. Senior management and technical personnel from various parties expressed their ideas and put forward many good ideas. Some of them were approved by Fang Jie on the spot. The future has far-reaching implications.

The first project to be approved is the project proposed by Feiteng Technology, which is called: Feiteng Cloud Computer.

What is "Boiling Cloud Computer"?
To put it bluntly, this computer, after being connected to the Internet, has the function of cloud computing.

When users use Boiling Cloud Computer normally, it is no different from ordinary computers, but when the computer is in an idle or standby state, the computer will join the supercomputing center network, become a part of it, and undertake a certain amount of supercomputing center distribution. computing tasks.

By analogy, these Boiling Cloud Computers are actually no different from the meat machines in the hacker world on the surface. It is nothing more than that hackers use them for illegal purposes, while Boiling Cloud Computers are used in the right way.

That is to say, the idea of ​​the Boiling Cloud Computer project is that since the operating cost of the supercomputing center is too high to make ends meet, and the construction speed cannot meet the huge demand of users for computing power, then a sea of ​​​​tactics will be launched to convert each personal user computer The computing power provided during idle time is used collectively.

Perhaps the computing power provided by a boiling cloud computer is negligible, and it can only perform some calculations that are not too complicated and last for a long time. What if the computing power of 100 million, 1000 million or even hundreds of millions of computers is combined?
To put it bluntly, Boiling Cloud Computer is actually a computing power server in the supercomputing center. When thousands of Boiling Cloud Computers based on individual users are assembled, the computing power they provide far exceeds hundreds of supercomputers. The computer room of the center can solve the dilemma faced by the current supercomputing center to a certain extent.

And this function does not affect the normal use of Boiling cloud computer by individual users. When the cloud computer is used by users, the cloud supercomputing system will not occupy memory and system resources, and users can also set the resource allocation ratio by themselves.

For example, when users are reading online novels, most of the computer system resources are actually wasted. At this time, users can allocate resources to the cloud supercomputing center to provide computing power support according to a certain proportion.

And this support is not free. According to the computing power and time it provides, the supercomputing center will pay users a certain computing power fee through the transaction treasure.

That is to say, users who use Boiling cloud computer, after the computer is turned on, may make some money by the way while eating, working, reading novels, listening to songs, etc. without taking up much system resources.

The money may not be much, but it is no problem to subsidize some electricity bills and reduce some living and working costs.

Anyway, the computer is idle, as long as it does not affect the normal use, and can make some extra money by the way, the users of Boiling Cloud Computer will naturally not mind.

In addition, Boiling Cloud Computer has three advantages.

One is that the technology of boiling technology has learned the trick of boiling cars.

That is, Boiling Cloud Computer is actually renting, not selling.

The lease term for each period is two years. After two years, Feiteng Technology will update and upgrade the computer hardware for free, and will also provide a full range of after-sales guarantee services during the period.

The rental fee for each period is different according to the computer configuration. The highest grade is 2400 yuan, the average annual rental fee is 1200 yuan, and the lowest grade is 600 yuan per year. The cost is only 50-100 yuan per month.

You must know that the monthly Internet fee starts at 60 yuan, and the 60 yuan/month computer rental fee is nothing at all.

This money does not need to be paid immediately. As long as the user uses the Boiling cloud computer and allows the computer to participate in cloud computing work in his spare time, the rental fee will be automatically deducted from the cloud computing revenue.

In other words, it is possible that after using the Boiling Cloud Computer for two years, the user does not have to pay a penny of rent, which is equivalent to using the computer for free.

Of course, it can't actually be free, because the electricity bill is borne by the user.

But then again, don't you also use your own computer to consume power?

What's more, many computer users still use the unit's electricity, and the electricity bill is reimbursed by the unit.

Of course, some users may also say that it seems a bit uneconomical to rent a two-year boiling cloud computer for 2400 yuan. You can buy a computer for 3000 yuan, so why rent it.

This refers to the second biggest advantage of Boiling Cloud Computer: high configuration.

For the needs of cloud computing, the configuration of the boiling cloud computer is obviously not comparable to the 3000-yuan low-end computer.

According to the configuration officially announced by Feiteng Technology, the performance of the 2400-yuan rental cloud computer is equivalent to the current computer configuration of 1 yuan on the market.

So the 2400 yuan is rented back, but it is a server-level computer. Even if the user watches videos on the computer and allocates computing power to the supercomputing center, there will be no obvious bad feeling.

If you use such a computer to simply play games, all the 3A masterpieces on the market will run without problems, and they are very smooth. If they are used to render graphics images, they are enough to meet the needs of professional designers.

If it is your own computer, after 2 or 3 years, the performance will be outdated or even eliminated. In the end, it can only be sold as scrap iron, and even scrap iron is not necessarily required by anyone.

However, there is no such problem in leasing Boiling Cloud Computer, because after the 2-year lease expires, you can choose to renew or stop the lease, and there is no depreciation loss.

If you choose to continue leasing, Feiteng Technology will update the computer and provide new high-end computers of the same period, which can be replaced with new ones just like the boiling car.

Therefore, with the three advantages of low cost, high configuration, and replacement, I don’t have to worry that no one will buy the Boiling Cloud Computer. It is even foreseeable that the Boiling Cloud Computer will become a rare commodity like the Boiling Car.

Because the core components of Boiling Cloud Computer, such as CPU, motherboard, memory, etc., are actually second-hand goods that have been replaced by supercomputing center upgrades and depreciation, and the quantity is limited.

For supercomputing centers, these second-hand accessories may have poor performance and high energy consumption, and are no longer suitable for continued use, but for ordinary home and business users, they are still high-end accessories when assembled.

Therefore, the Boiling Cloud Computer launched by Boiling Supercomputing and Boiling Technology is actually waste utilization, and it does not cost much. Otherwise, the rental fee will not be so cheap. The price is 2400 yuan / 2 years. Actuaries and market analysts The result of precise calculation, it is impossible to lose money.

As for two years later, when it is time to upgrade Boiling Cloud Computers, the eliminated computers are not real waste products. All the recycled ones are recycled, then refurbished and assembled. Through Fangzhou Technology, these third-hand computers will be sold at low prices. Sell ​​to Southeast Asia, Africa, South America and other underdeveloped regions.

For the people in these underdeveloped areas, even if it is a third-hand computer, it is still a high-end configuration, but the price can be relatively low, and there is no reason not to pay for it.

What's more, with the characterization of Ark International, it is fully implemented by the capital. Fang Jie never pays attention to the ethics of earning money from foreigners. He said that the computer is new, and everyone can only believe that, there is nothing wrong with it. .

All in all, the Boiling Cloud Computer project was established during the seminar. The first batch of 200 million Boiling Cloud Computers was officially launched to the market less than a month after the meeting, and then it was sold out in less than a week, and the supply exceeded demand.

After all, there is a precedent like Boiling Car, and now Boiling Computer has been launched with the same sales model. Experienced people will not wait and see any more.

If ordinary people have a strong desire to buy, then corporate users can be described as crazy, and they all order hundreds of computers in batches.

Because the bosses of the company are too aware of the operating costs of their companies, especially the Internet companies, the cost of electricity has always been high. During the work period, the employees generally turn on the computer all day, but often fish, and do not use the computer 100%. Resources, this ratio can reach 40% even if it is not bad.

Since the company's office computer idle rate is very high, is it that Mao Bu is making some extra money to subsidize the company's operating costs?

Moreover, the company's computer always needs to be replaced every 2 or 3 years. During this period, it is often damaged and needs to be maintained, which is also a big cost. If it is not maintained, it is always complained by employees that the computer is too slow and unusable. Now the emergence of Boiling Cloud Computer, These problems have been completely solved, and after a little calculation, you will know that renting a Boiling Cloud computer is definitely more cost-effective than buying a computer yourself.

From the perspective of the Boiling Group, anyway, these accessories, except for peripheral accessories such as monitors, keyboards and mice, and the core accessories are all obsolete things. It is not a pity to rent them cheaply to the public for use. of.

From Fang Jie's point of view, another advantage of doing this is that it greatly improves people's life and production efficiency. In the past, using low-end computers, whether for work or games, was very troublesome, or even impossible. Now the 200 million boiling cloud Computer users don't have to worry about this.

If the whole society is using Boiling Cloud Computer, not to mention how much additional computing power it can bring, it will greatly promote and accelerate the development of the entire society and the economy.

Therefore, the project of Boiling Cloud Computer is also a good thing to benefit the country and the people with one stroke, and achieve a win-win situation for all parties.

The only problem is that the output of Boiling Cloud Computer is too low, and its production capacity depends on the speed of hardware elimination of Boiling Supercomputing Center. According to Moore's Law, theoretically, the weekly shipment of Boiling Cloud Computer in the short term is at most 60 units, which cannot fully cover the whole country. market.

In fact, this shipment volume is already very strong, accounting for almost two-thirds of the country's annual PC sales market, and monopolizing the low-end computer sales market in one fell swoop. Even if you look at the world, the market share exceeds 25%.

Of course, the purpose of engaging in Feiteng Cloud Computer is not to make money, and anyone with a discerning eye in the industry can see that the project of Feiteng Cloud Computer is almost unprofitable when the rental fee is so low.

Ultimately, this project aims to provide massive computing power to the supercomputing center and accelerate scientific research, especially in the direction of artificial intelligence.

However, when the Feiteng Group made such a move and launched the Feiteng cloud computer with such a high cost performance, it directly offended the entire PC industry. The low-end and mid-end computer manufacturers were scolded, and in order to survive, they took the rhythm on the Internet.

For example, Boiling Cloud Computer is just a second-hand computer. Who would be willing to rent a second-hand computer for 1200 yuan per year?
Moreover, second-hand computers are very prone to various failures, and the maintenance costs are high and troublesome.

As for making money by computing power, that's just a gimmick. Even if a computer provides full computing power around the clock, it will not earn 10 yuan. After deducting the cost of electricity, the annual rental fee can't actually be earned back, right?

Fang Jie ignored these voices and continued to go his own way.

If you want to do something in society, it is impossible not to be scolded by others. In the past, he tried his best to value peace and make fewer enemies, but if the PC industry was to be offended, he would be offended.

Anyway, with the advent of the era of mobile smart phones, the PC industry will decline sooner or later. Longer pain is better than short-term pain. At least it is replaced by boiling cloud computer, which can provide huge computing power. Generally speaking, the advantages outweigh the disadvantages.

As for the unemployment problem of a large number of people in the industry brought about by this move, Fang Jie also gave a solution to let the businessmen of the computer city transform their business from computer sales to computer after-sales service.

That is, in the past, the merchants in the computer city mainly made money by selling assembled computers, but now they are responsible for the recycling, delivery, and software and hardware maintenance services of Boiling Cloud Computer to maintain their livelihoods.

This is actually a kind of industrial progress and upgrading.

More than 10 years ago, when the domestic personal computer market started to take off, computers were still a tall and new thing, and users who could use computers, even those who could only type, were considered high-tech talents. At that time, there were very few personal computer users, and the quality of users was It is also relatively high. Generally, they understand the basic maintenance of computers. There is almost no market for the computer after-sales business.

But it is different now. Personal computers have been popularized to the public. The user base is large, and not everyone can maintain computers. For example, reinstalling the system is impossible for many people. In addition, the pace of life Hurry up, we don't have time to deal with these little problems.

Especially for enterprise users, people may be a big business of tens of thousands of millions in minutes. How can they spend their energy on hardware maintenance? If you specially hire technical staff for computer technical maintenance, the cost is too high, and you may not be able to handle it. All problems, so professional computer home repair and maintenance services have broad market prospects.

Although Boiling Cloud Computer adopts a rental and sales model and also provides after-sales service, this after-sales service only provides basic services. If there are software and hardware problems caused by the customer during use, additional after-sales maintenance fees will be charged.

Feideng Technology's approach is to establish a Feideng cloud computer after-sales platform. Customers place orders on the platform to invite maintenance personnel, while self-employed technicians take orders from the platform based on the principle of proximity, and get one order. Earn as little as 50, more Then there are several hundred, not to mention making a lot of money, there is no problem in maintaining a normal basic life.

Moreover, the business of Boiling Cloud Computer after-sales platform is not only for cloud computer users. Users of other brands or users who have problems with assembling computers can place orders on it and enjoy on-site professional services.

Whether it is the Boiling cloud computer project or these related businesses, it is only one of the projects discussed in this seminar. In addition, there are many planned projects that need to be talked about.

For example, Boiling Games has proposed a concept of "generative adversarial network learning" for artificial intelligence learning.

The proposal of this concept provides an idea for saving a lot of computing power for artificial intelligence training and research and development.

The concepts of "supervised learning" and "reinforcement learning" have been mentioned before, but these require a lot of computing power, while "generative adversarial network learning" requires relatively low computing power, because in this way, artificial The object of intelligent learning to complement is human, and it is carried out in the form of network confrontation.

In layman's terms, it is possible to make a series of online games specially used to educate and cultivate artificial intelligence, but these online games can also be played by ordinary players, and are highly playable and confrontational.

Then, the artificial intelligence continuously learns the game experience from the players by constantly confronting the players in the game, so as to achieve self-growth and intelligence improvement.

To put it simply, the player is used as a sparring partner, so that the player's logical judgment and behavioral decision-making become the learning nutrients of the artificial intelligence, rather than relying on another program code to train the artificial intelligence, which saves part of the computing power.

The disadvantage of this idea is that the player's behavioral decisions are often uncertain and random, which will generate a lot of invalid data, making the AI's learning efficiency very low, and even biasing the AI, just like adults teach children badly.

But then again, Boiling Games is meant to be a game, and players also want to play games. During the operation of the game, Boiling Games will have collected a lot of player data, so arrange to train an AI system in the game. Continuing to learn against players is a matter of course, it will not delay anything, and it will not increase the cost.

As for what kind of game to develop, there are no restrictions on the theme, type, and style. Of course, the best multiplayer online FPS and MOBA competitive online games with strong confrontation are.

Now that he has ideas and wants to do things, Fang Jie is of course happy to see it happen. He turned around and stuffed a game plan of "PUBG Mobile" for Boiling Games, asking them to develop this game within one year, and join the AI intelligent system.

(End of this chapter)