Jining, Shandong Province cracked the case of "12.15" mega-network spreading obscene articles for profit.

   Legal Network Jining (Shandong) August 20th Reporter Xu Peng On the evening of August 16th, seven main suspects who fled Malaysia landed at Yaoqiang Airport in Jinan by plane and surrendered themselves to Jining police in Shandong. At this point, the "12.15" cross-border mega-network dissemination of obscene articles for profit has basically ended.

   In this case, 12 related yellow-related live broadcast platforms were sorted out. The suspects were spread all over the country in 16 provinces and cities and overseas in Malaysia, Myanmar, Thailand and other places. 62 suspects involved in platform management, operation, promotion, family heads and anchors were arrested. The first batch of 16 people were transferred, and 22 people were to be arrested. More than 21 million people involved in the case were frozen and one case was seized. From the root, the whole chain destroyed the criminal gang of the webcast platform and completely eradicated this yellow "cancer" that seriously endangered society.

   Statistics show that these platforms have more than 2 million registered members, with hundreds of millions of viewers clicking, and the amount involved exceeds 100 million yuan. In 2018, the "12.15" case was listed as a supervision case of "eliminating pornography and illegal publications" by the National Anti-vice Office and the Ministry of Public Security.

   It is understood that due to many factors such as low crime cost, high specialization and virtual network, this case is very difficult to crack down, and it is not easy to achieve the whole chain attack from anchor, family leader to management members, shareholders and top management personnel.

   According to the investigators, at present, more than 260 live broadcast platforms related to pornography are online at the peak, and only 12 are related to the case, which highlights the confusion in the management of the current live webcast platform, especially the obscene video images spread, which makes people collapse instantly, greatly endangering the physical and mental health of young people.

   A small clue leads to a major case involving 100 million yuan.

   Sexual intercourse, promiscuity, incest, props performance, including obscene videos of children, mother and child, mother and daughter, human and dog, same sex, exposure of sexual organs, etc., were vividly shown in this case. In fact, this case was led by a small clue.

   At the beginning of December, 2017, the office of the leading group for "eliminating pornography and illegal publications" in Jining City immediately set up a "12.15" task force based on a case clue provided by the Public Security Detachment of the Municipal Public Security Bureau to open a live broadcast room for obscene performances, which was jointly operated by the Comprehensive Law Enforcement Bureau of Jining Cultural Market.

   According to relevant clues, the task force found out that since 2017, criminal suspects Yan and Chu have been engaged in profit-making activities of spreading obscene articles on webcast platforms such as "September Live". Subsequently, the task force tracked down a "small clue" and found out a rare large-scale cross-border network to spread obscene articles for profit.

   The public security department and the cultural law enforcement department of Jining City shall, in accordance with the scope of their duties, define the division of labor and work together. Cultural law enforcement departments are responsible for providing online places for handling cases and obtaining evidence, equipment for obtaining evidence and funds necessary for handling cases and obtaining evidence online. At the same time, they give full play to their professional and technical advantages to conduct online remote inspection and obtaining evidence through the network monitoring platform, and assist the public security departments in interrogating and arresting criminal suspects. The public security department integrates a variety of investigation elements, takes advantage of big data technology to investigate the identity information, fund account number, fund chain and personnel chain of criminal suspects, is responsible for the interrogation and arrest of criminal suspects, and assists cultural law enforcement departments in network remote inspection and network video forensics.

   Subsequently, the task force followed suit, and after more than eight months of hard work, the personnel chain, capital chain and technology chain involved in the "September live broadcast" have been identified. The pornographic webcast platform has more than 2 million registered members, hundreds of millions of viewers’ clicks and hundreds of millions of yuan involved. The personnel composition is: "Member — Anchor — Family leader — Manage members — Shareholders — The top management members are six levels, and the suspects are distributed in 16 provinces and countries such as Malaysia, Myanmar and Thailand. It is a rare case in the country to spread obscene articles through cross-border mega-networks. After the case was found out, the office of the leading group for "eliminating pornography and illegal publications" in Jining City began to close the net. As of August 16, the task force had arrested 62 suspects involved in the case and frozen more than 21 million yuan of funds involved. At the same time, 12 platforms related to this were destroyed, such as "live broadcast of honey juice", "shallow live broadcast", "cat", "allow beauty" and "ram"

   According to the investigators, members recharge the platform through Alipay or WeChat in exchange for platform currency, and the exchange rate is about RMB to 10 platform currency in 1 yuan. There are various virtual gifts in the studio, such as cucumbers, bananas, diamond rings, flowers, sports cars, bombers, passenger planes and rockets. Among them, sports cars, 888 diamonds/car; Bomber, 1314 diamonds/frame; Passenger plane, 8888 diamonds/frame; Rockets, 18888 diamonds/piece, etc.

   "There are three main forms of communication. One is the free and open live broadcast of members, preparing for the next charge to provide obscene videos, and occasionally performing obscene performances to attract audiences, which is equivalent to advertising in physical stores; Second, the obscene performances charged by time or time; The third is to pay for the virtual gift to the anchor. When the gift reaches a certain value (usually a sports car or bomber), the anchor adds members WeChat and QQ, and sends obscene videos to members after establishing WeChat groups and QQ groups. " The case handler said.

   It is extremely difficult to strike with clear hierarchy.

   Through the analysis of "live broadcast in September", the task force found that the division of labor was very clear. Management members, shareholders and top management members are responsible for the management of the platform, the head of the family manages the anchor, and members pay to watch. The income distribution among management members, shareholders, top management members, family heads and anchors is 37-30, that is, about 70% of the total revenue of the platform is paid to the family heads, and the family heads are then distributed to the anchors (the share of family heads and anchors accounts for 20%: 50% of the total revenue of the platform); 30% is platform gross profit, and after deducting some operating costs (cloud service fees, venue rental fees, personnel employment fees, etc.), it is platform net profit.

   According to the investigators, the same anchor can be online on multiple platforms at the same time, and hundreds of live broadcast rooms on the same platform can be broadcast live at the same time. The number of real-time online viewers on the same platform is very large.

   "The income of the anchor is very considerable. Ordinary" anchors "earn as little as 1,000 yuan a day, and well-known" anchors "earn more than 20,000 yuan a day. If they broadcast live on multiple platforms at the same time, the income will be higher." According to the investigators, each anchor watched hundreds of members and thousands of people watched online, and found that more than 240,000 people watched the same anchor performance. "What’s more, there are a large number of teenagers in the audience, and there is a trend of younger development."

   Although Moulido is harmful, the investment of the whole platform is not high, the threshold is low, and it is very difficult to crack down.

   "The cost of building an entire live broadcast platform does not exceed 1000 yuan at most." Fan Cunchang, deputy director of the Comprehensive Law Enforcement Bureau of Jining Cultural Market, often said that the first step is to buy the platform source code only by 10 yuan, the second step is to rent the server, then set up the platform, and use the purchased source code to build the mobile phone application of the live broadcast platform, and the third step is to apply for third-party login interfaces such as WeChat, QQ and Sina Weibo, and third-party payment interfaces such as Alipay. "‘ Live in September ’ For example, the initial investment is only more than 600,000 yuan, and the illegal profit is as high as tens of millions of yuan. "

   Moreover, the live broadcast platform has many ways to avoid supervision, and it is spread by scanning QR codes and downloading, and it also avoids supervision by changing its name and platform version from time to time. Viewers can register as members to watch only through the verification code of mobile phone number or through the binding of third-party software such as WeChat, without real-name authentication.

   "The people involved have strong anti-reconnaissance ability and their true identity is hidden." The case-handling personnel said that the illegal live broadcast platform did not fulfill the principle of "real name in the background and voluntary in the front desk" of the regular live broadcast platform, and the anchor room could be directly certified for the female anchor to perform without real-name certification. Most of the people involved in the case contact each other through the Internet, and even they don’t know each other. They can contact each other only through virtual identities such as WeChat, and even the accounts for collection and payment steal other people’s identity information, which brings great difficulties to the identification of the true identity of the people involved.

   Webcast chaos is in urgent need of supervision.

   "The aggregation platform steals the traffic of the live broadcast platform through hacking technology, which can maximize the collection of common online platforms related to pornography." The police investigating the case said that the analysis found that about 260 live broadcast platforms were online at the same time during the peak of live broadcast. "The chaos in the field of webcasting can be seen."

   "Webcasts are mainly supported by the number of viewers. With a huge audience, there will be no income. Moreover, the technical requirements of the live broadcast platform are low, and it is very simple for live broadcast women to want to broadcast live, as long as they have a computer or smart phone. " The police handling the case said.

   At the same time, the whole criminal process is highly specialized. "For example, in the past, the transfer of stolen money was converted between different bank cards, and now it is directly handed over to a professional money laundering company, and it is very difficult to check the capital flow." According to the police investigating the case, information such as WeChat, QQ number and telephone number are basically false, and identity authentication is not simple.

   "Like this case, the main suspects are abroad, and the servers are also abroad. When arresting, international police law enforcement cooperation is needed, and sometimes it is not smooth." The police handling the case said.

   Fan Cunchang suggested that according to the new situation that the network market has become the main battlefield of the cultural market, we should select high-quality talents such as computer majors, set up a professional network culture law enforcement team, strengthen the supervision and management of network violations, and be fully responsible for investigating cases in the network field, so as to create a clear cyberspace environment and escort the healthy growth of teenagers.

   "The National Anti-vice Office has established a major case filing system. It is suggested that the Provincial Anti-vice Office should also establish a major case supervision system and set up special funds to reward key cases. After research, the cases declared by various cities will be included in the supervision cases for publication. At the end of the year, the cities that have completed the supervision cases will be given special funds for handling cases. At the same time, special regulations for handling special funds for cases were introduced to ensure that grassroots case handlers can effectively and permanently investigate the project. " Fan Cun often said.

   In view of the new situation and new characteristics of frequent illegal cases such as webcasting, Fan Cunchang suggested that the national and provincial anti-vice offices should hold more training courses on how to investigate and deal with online cases in order to adapt to the challenges brought by the increasingly diverse and hidden criminal means to law enforcement.

When you are young, you should rush forward and see which of the seven models of Kaiyi Xuanjie suits you.

Spring is the season when everything recovers, and it is also the spring day when heavy winter clothes fade away. Young people want to live their own spring and summer style, so how can they get a car as a wingman? The premium price of 53,900 yuan is sincere, and the appearance and configuration are also modest. If you are interested in this SUV built for "Houlang", you may wish to take a look at the purchase guide of this car.

Kaiyi has a total of 7 models, with the price range of 53,900-79,900 yuan. The appearance of the new car is fashionable and young, and the interior style is simple and generous. Two 10.25-inch screens in the car bring excellent visual impact. In addition, the new car also has a good performance in configuration, which is very suitable for young consumers to buy.

The appearance, space and control of dazzle world are remarkable.

As a fashionable SUV specially designed for young people, the cool shape is definitely self-evident. Kaiyi not only provides consumers with two front grille styles of "wingspan" and "full of stars", but also has up to nine color schemes to choose from, including five solid color bodies, namely, floating clouds white, proud red, willful blue, slash gray and wind blue, and four two-color bodies, namely, hard core black+floating clouds white, hard core black+proud red, hard core black+willful blue, hard core black+wind blue, which are enough to satisfy young people.

Kaiyi’s body size is 4400/1831/1683mm, and it has a wheelbase of 2632 mm. The interior space of the car is very considerable, and it will not be crowded when there are three people in the back row. The reasonable and exquisite interior will also leave some spare space on the heads of the front and rear passengers, even if the passengers are taller, they don’t have to worry about close contact with the ceiling of the car. The trunk can easily meet the daily household use in the normal state, and the rear seats support 4/6 ratio, which can expand the carrying space and meet the space requirements for loading oversized luggage.

In terms of power, Kaiyi is equipped with 1.5L, its maximum power is 85kW(116PS), and its peak torque is 143 N m.. Matching 5-speed manual or CVT continuously variable gearbox according to different models, CVT models are also equipped with two driving modes, normal mode can bring smoother power output and lower fuel consumption, while sports mode can increase the speed faster, which can bring more surging driving experience.

There are three types of automatic transmission models, which are also divided into low, medium and high. The configuration of low-music type is almost the same as that of manual transmission type, with only one more function. Compared with the music type, the fun type has more functions such as fixed-speed cruise, keyless start/entry, driving mode switching, etc., and the price is also controlled in the early 70 thousand yuan. Finally, let’s take a look at Yao, the model with the highest price. Its configuration level is almost the same as that of the manual transmission, and it adds functions such as remote starting and automatic parking. Combined with its superior price of less than 80,000 yuan, its super high price-performance ratio is attractive. If you want to buy an automatic transmission model, you can bring you an excellent car experience without thinking about Yao.

Walking every day, is it right? Walking skills in several situations

Editor’s note: In life, we will inevitably encounter all kinds of trivial problems. Mastering some tips for a healthy life may solve the life problems. Let’s look at the arrangement of Xiaobian!

Walking every day, is it right? Let’s look at the walking skills in several situations

"Walking" can be described as an activity that accompanies people all their lives. Many people often get tired after two steps, or feel pain in their waist and knees. Apart from the lack of muscle strength, it is also related to the way they walk. Look at the walking skills in these situations, which can prevent fatigue and protect the waist and legs.

1. Walk indoors.

Error: When walking indoors (barefoot or wearing slippers), if you can hear the sound of "knocking", it proves that you are using your hind feet to land. Before stepping out of the foot, the human body will bend the forefoot to support the weight, and the calf muscles will be tense; After stepping out of the foot, the heel hits the ground first, and the toes bend upward, which has poor stability, which will weaken the ability of the arch to mitigate the impact and bring a burden to the knees and waist.

Correct: when stepping, don’t step on the back foot deeply, just gently touch the ground with the toe, and pay attention to relaxing the calf muscles; After stepping out, land on the ground with all your feet. In this way, the calf muscles switch between relaxation and tension, which helps blood circulation. Care should be taken not to stride too much.

2. Go up and down the stairs.

Mistake: When going up stairs, if you put the center of gravity of your front foot and upper body on the same side, it will not only make your body unstable, but also the muscles of your front foot will bear the pressure of moving your body and increase the burden on your knee joint. When going down stairs, if you put your front foot on the center line of your body, it will destroy the balance, especially for people with stiff spine and pelvis, the leg burden is heavier and you need handrail support.

Correct: When going up the stairs, the distance between your feet is equivalent to that of the pelvis. Lift your feet directly above to avoid being biased to the center of your body. Should be able to feel the pelvic force to drive the body upstairs, rather than using leg muscles; The head and feet are in different directions, for example, when taking the right leg, the head leans to the left. When going down the stairs, the distance between the feet is slightly wider than the pelvis, and the whole foot is on the ground; The pelvis tilts slightly with the front foot, flexibly uses the strength of the back, and the head leans to the side opposite to the foot to stabilize the body.

3. Walk long distances.

Mistake: If you wave your arms back and forth, your back will be easily stressed and your body will soon get tired; In addition, the stride will naturally increase, and you will land with your heel unconsciously.

Correct: Keep your upper body flexible and soft, slightly bend your elbows, feel your shoulder blades moving back and forth, and swing your hands left and right. In order to cooperate with the activities of the arms and shoulder blades, the pelvis will naturally move forward, driving the legs rhythmically, and it is not easy to feel tired.

4. Carrying heavy luggage.

Mistake: When carrying heavy luggage, it is easy for the body to lean forward and sway from side to side. If you put your front foot in the center of your body like walking the "model step", in order to maintain balance, your shoulders and neck will consume strength and bring a great burden to your body.

Correct: The key is to narrow the range of bearing gravity, stick the luggage close to the body as far as possible, keep the distance between your feet equal to the pelvis, and step forward, so as to ensure the stability of your body..[Full text]

Low-salt healthy diet has comparable antihypertensive effect to drugs.

A new study in the United States shows that low salt and healthy diet have comparable antihypertensive effects on patients with early hypertension or mild hypertension.

In this study, the standard of low salt is no more than 1 teaspoon of salt per day, which is also recommended by the US Food and Drug Administration. The healthy diet here refers to the DASH diet rich in fruits, vegetables, coarse grains, low-fat or skimmed milk products, fish, poultry, beans, seeds and nuts. DASH is the English abbreviation of the term "diet prevents hypertension".

Researchers say that people have long known that low salt and DASH diet can prevent hypertension or prevent blood pressure from rising, and their research shows that the combination of the two is better. The research report was published in the latest issue of Journal of American Heart Association. One of the authors, Lawrence Appel of Johns Hopkins University School of Medicine, said that they observed that the combination of low salt and DASH diet had "at least as great a effect as prescription drugs" on people at high risk of severe hypertension..[Full text]

How to wear underwear is the healthiest

Underwear is the most intimate and close-fitting clothing. Although it is "small", the health problems involved are very important. In particular, the following points need to be noted:

Underwear should be made of pure cotton, light color and few colors. Underwear is best made of pure cotton, and underwear made of chemical fiber may rub against vulva skin, leading to inflammation, allergy and itching. In the choice of color, dark underwear uses more dyes and contains chemicals like "hodgepodge". It is recommended to buy light-colored underwear.

Underwear should be cleaned separately every day. Many people are used to stacking underwear in a basin for a few days, and then washing them after saving enough, even wearing a pair of underwear for two or three days. This will make underwear a "bacterial nest" and it will be more difficult to clean. Wash it by hand with clean water and special laundry soap and underwear cleaner every day, and separate it from other clothes to prevent cross-infection of bacteria.

It is best to change a batch of underwear every six months. Many people think that underwear, like clothes, can be worn for several years, especially for the elderly, and they are often worn out and deformed before being thrown away..[Full text]

Don’t buy flowers for glass tableware.

Colorful glasses with different patterns are pleasing to the eye and are a must-have item in many people’s homes. However, recently, researchers at Plymouth University in the United Kingdom found that some glass tableware with colored patterns or patterns contain high levels of heavy metals such as cadmium and lead, which will harm human health.

The researchers used X-ray fluorescence analyzer to inspect the new and second-hand glasses, wine bottles and other glass products in the supermarket, and found that most of the glasses contained cadmium and lead seriously exceeding the standard. When people drink water, the paint that falls off may enter the human body through the mouth, damaging the brain and nervous system. Experts suggest that if you have this kind of tableware at home, it is best not to use it to hold food, let alone acidic drinks, so as to avoid heavy metal precipitation..[Full text]

Good hygiene habits can prevent respiratory infectious diseases.

Disease control experts remind the public: wash your hands frequently and don’t wipe your hands with dirty towels; Wash or wipe your hands after touching respiratory secretions; Avoid sharing cups, tableware and other items with others; Pay attention to environmental sanitation and indoor ventilation. If there are patients with symptoms of respiratory infectious diseases around, increase ventilation times, avoid through flow when opening the window, pay attention to keep warm, and dry clothes and bedding in the sun often. Drink more water and eat more fruits and vegetables to increase the body’s immunity; Try to avoid going to crowded public places; Cover your nose and mouth with handkerchiefs or paper towels when sneezing or coughing in crowded places. Don’t spit everywhere, and don’t throw away the toilet paper used for spitting or wiping your nose at will. Avoid contact with cats, dogs, birds, rats and their droppings and excreta. Once in contact, be sure to wash your hands. Do not touch dead or suspected sick animals..[Full text]

5 tips for preventing teeth from yellowing

People’s teeth are light yellow to yellowish brown. As you get older, your teeth will get darker. Some foods and drinks with high pigment, such as coffee, tea and some drugs, smoking and so on, will make teeth yellow. American MSN Health website published an article, which introduced five home-made tricks to prevent teeth from yellowing:

1. Drink drinks through a straw. John C. Moon, a dentist in Half Moon Bay, California, said that drinking coffee, soda water, tea and other drinks through straws can minimize the contact between drinks and the surface of teeth, which can significantly reduce the chance of teeth being colored by drinks and make teeth whiter and more lasting.

2. Eat tooth whitening food after drinking red wine or coffee. Apple, for example, is slightly sour and astringent, and the rough pulp rich in fiber makes it an ideal food for cleaning teeth. Chewing apples, celery, carrots and other foods carefully can clean teeth like a toothbrush and wipe off stubborn stains on teeth.

3. Self-made whitening toothpaste. Jennifer Jablo, a plastic dentist in new york, said: "Particles can remove stains and polish teeth, but they are not corrosive enough to damage enamel." Self-made whitening toothpaste can be made with baking powder and water. It can be used several times a month to remove the coloring on the surface of teeth and make teeth white.

4. Choose the right lipstick. Jessica Libiskey, a makeup artist in new york City, said: "Lipstick with a blue background will make teeth look white." In order to make it easier to distinguish, three or four lipsticks with different background colors can be identified side by side.

5. Reduce heartburn and acid reflux. Edmund Hugh Park Jung Su, a consultant of the American Dental Association and a professor at the College of Stomatology, University of California, said that the acid water in the digestive system will return to the mouth and corrode the enamel like the acid in soda water or sports drinks. Therefore, it is necessary to consult a professional doctor and control the acid reflux..[Full text]

Related reading:

Bulk nuts and minced meat go to the supermarket to buy less of these six foods.

Improper use of humidifier is dangerous. Three kinds of plants are suitable for bedroom.

Scientific morning exercises and fitness tips for grasping six key office gaps

What should I do if I stay up late and hurt myself? Do these things to save a little damage.

When is the best time to sleep? Taking a nap for 25 minutes can "recharge" the body.

Walking 8,800 steps every day, staying in the hospital, practicing late or not, does it hurt the yang?

17 "healthy habits" should be abandoned. What is the healthiest day?

Go to the bathroom and play with your mobile phone? It may hurt you from head to toe

Can drinking tea clear the intestines and detoxify? Making tea in a microwave oven may be the healthiest.

Don’t put hot water in disposable paper cups. "Drink 8 glasses of water every day" is unscientific.

How long do you need sunshine? Research helps you determine the best duration.

Falling asleep is difficult to become an epidemic. Can sheep fall asleep quickly?

Fluorine-containing toothpaste does not cause cancer, and umbrellas do not protect against the sun.

Is the home really infested with mites? Six common misunderstandings of microwave ovens

The bowl is soaked with natural antipruritic prescription in the crazy life without washing bacteria.

Many pots of plants at home buy less medicine. Four cleaning mistakes are easy to cause disease.

Coup: How to keep leeks fresh How to cut onions without tears?

Don’t dry the dishes that have been washed with floor mats in four places at home.

Housework can never be finished? Send you a set of "guide to the rules of housework"

Experts teach you how to choose toothbrushes and throw kitchen waste the same day.

Combs are changed every four years, and pots and pans are placed vertically.

Don’t dip the detergent in water before squeezing toothpaste. Don’t use it if you can.

Traffic Management Bureau of Ministry of Public Security introduced new measures to optimize motor vehicle registration service.

  CCTV News:"Traffic Management Bureau of the Ministry of Public Security" WeChat official news, recently, the Ministry of Public Security launched a number of measures to ensure high-quality development of public security organs’ services. New measures to optimize motor vehicle registration service and further promote automobile consumption are one of the important contents. The Traffic Management Bureau of the Ministry of Public Security deployed local public security traffic control departments to further refine measures, optimize processes, simplify and optimize vehicle registration procedures, better facilitate mass enterprises to handle affairs, promote the circulation of automobile consumption, and better serve and ensure high-quality development.

  First, simplify the registration procedures of new cars, and facilitate one-stop registration of new cars.On the basis of the previous pilot, we will further expand the pilot of pre-inspection of domestic passenger car registered production enterprises, and implement the pre-inspection of new cars in 33 production enterprises in 36 cities, so that the people are exempted from submitting their cars, reducing the waiting time for people to do things in line, adapting to new car sales modes such as online car sales on the Internet, and better promoting automobile consumption. Actively support the establishment of motor vehicle registration service stations in banks, auto financing companies and other financial institutions, so as to facilitate the people to handle loan and mortgage registration procedures at one stop.

  Second, facilitate the registration of second-hand car transfer and promote the circulation of second-hand car transactions.In conjunction with the commerce, taxation and other departments, we will implement the convenience measures such as canceling the restrictions on the movement of used cars and the registration of second-hand passenger cars. The implementation of motorcycle registration in the province, the province (district) for motorcycle registration, transfer registration, residence move in, the applicant can rely on the resident ID card "one card through office", no need to submit proof of temporary residence, to facilitate people to buy cars in different places. If local laws and regulations implement a separate management policy for the purchase of motorcycles, local regulations shall prevail.

  Three, the implementation of second-hand car export registration in different places, to promote the development of second-hand car export.For second-hand car export enterprises to purchase vehicles in different places, they are allowed to apply for cancellation of registration at the place where the enterprise is located, and continue to promote cross-provincial traffic control services. For second-hand car export enterprises that have the conditions to issue invoices for second-hand car transactions, actively support the establishment of motor vehicle registration service stations, facilitate enterprises to register in one stop, reduce costs and burdens, and create a good environment for second-hand car exports.

The quality requirements of SF aging parts are continuously upgraded.

Source: Financial Network

  The performance of SF Express in the first quarterly report was like a deep-water bomb, which caused an uproar in the market. Although the company explained that this was due to the short-term impact caused by the superposition of various factors, the company’s long-term development trend did not change. SF Express, which was too "excellent" in 2020, was constantly questioned. What’s more, it was even said that SF Express lost money because its business pieces were not profitable and it did not ask SF Express to send paper documents in the electronic era. But is SF Express really just a courier company that sends a "contract"? Is it really only a contract that needs SF Express today?

  SF Express, which started from business parts, has now become a comprehensive business entity with business clusters such as time-limited parts, economic parts, express business, city, cold transportation, international business and supply chain business. The proportion of time-limited parts only accounts for 43.09% in the 2020 annual report, but in 2020, the business income of time-limited parts excluding tax was 66.36 billion yuan, a year-on-year increase of 17.41%. As the core product of SF, the profitability of SF’s aging parts continues to be stable.

  Today, SF has upgraded the aging products to achieve a new definition. First, it is the aging commitment, making accurate commitments to customers and taking the lead in aging. Through the combination of air/rail/steam transportation modes, priority is given to shipment, transit and delivery, providing competitive timeliness. The second is the operation guarantee, which realizes the traceable door-to-door standard express service through the exclusive process guarantee of each link of receiving and forwarding. The third is based on different scenarios, rich product matrix to meet the diverse customer needs of different customers. The fourth is to maintain high efficiency. If it is simply understood, SF time-limited products are trying to deliver them today and arrive tomorrow.

  Some insiders said that in different times, customers who need prescription products are different. The core of SF aging network is not aimed at the business parts market, but its core is high-aging and reliable service. Today, high-end consumer goods and fresh food are its biggest use varieties. Founder Securities analysts believe that the aging parts have entered a mature stage in the pattern, but the high growth of e-commerce demand has given the aging parts a second vitality, which is the core driving force for SF Express’s performance growth at present.

  Among SF Express’s over 60 billion yuan time-consuming parts revenue, the proportion of consumer business has increased rapidly, accounting for almost half of the country, and the proportion of document business is about 30%, while the contract invoice business is only a small part of document products. Simply calculate, if the revenue of SF Express is 154 billion yuan in 2020, then the revenue of document products only accounts for about 13% of the total revenue.

  The reason why SF is a growth stock rather than a cyclical stock is to get rid of the dependence on business parts and equalize the so-called cyclical influence. In the era of big consumption, SF’s aging parts have already been upgraded, and they are active in the demand side of all consumers and B-end users, keeping up with the high-end demand of customers, and the customer’s stickiness is constantly increasing.

  Figure: Category structure calculation of SF aging parts (2020H1, from orient securities research report).

  High-end demand for aging parts in the era of great consumption

  Since 2020, affected by epidemic factors, the online shopping penetration rate of residents has been improved, and the online promotion of high-end e-commerce has increased, and consumers’ online shopping habits have continued.

  After the epidemic situation has been basically controlled, online shopping has maintained a rapid growth trend, and high-quality consumer goods, such as high-value and high-margin department stores, high-time survival fresh goods, etc., are also accelerating the online trend. These high value-added goods also have a relatively rigid demand for high-quality express delivery services. Driven by the demand of domestic high-end e-commerce, traditional commerce is accelerating to penetrate online, and the demand space of aging network is opening. The demand is shifting from traditional business parts to undertaking consumption upgrading and the rise of high-tech industries, and the space of SF aging parts is further opened.

  Orient securities analysts believe that the demand for aging parts is shifting from transporting traditional business parts to undertaking online demand for high-quality consumption. From the perspective of sub-categories, the growth rate of documents has slowed down due to electronicization, and high-end e-commerce devices have benefited from the accelerated growth of high-quality online consumption. In the medium and long term, with the rise of high-tech manufacturing industry and the commissioning of Ezhou Airport, aging network is also expected to undertake the delivery demand of precision instruments, spare parts, biomedical products and other goods brought about by the rise of high-tech industry.

  The analyst believes that the increase in the proportion of consumption in aging parts is not only due to the explosive traffic caused by the epidemic in a short period of time, but also inevitable, that is, the advantages of e-commerce channels in resource allocation, circulation mode, efficiency and cost, the continuous investment of luxury e-commerce sellers in digital construction, the continuous optimization of fresh e-commerce sellers on fresh supply chain, and the continuous enhancement of consumers’ trust and stickiness in online channels.

  SF’s online penetration rate in 3C electronics, high-value clothing, medical health, skin care and other categories has been continuously improved. The strict requirements for express quality of Cordyceps sinensis, gold jewelry, high-end clothing, high-end cosmetics such as lamer, and high-end small household appliances such as Dyson have promoted the accelerated development of SF’s prescription parts. At the same time, the development of the company’s supply chain business will also bring in new prescription parts, and the development of SF’s prescription parts will be strengthened by the joint efforts of the two.

  Innovating Value-added Space of Business Components and Enhancing Customer Stickiness

  In the field of business parts, SF also provides extended value-added services on the basis of express delivery service, and provides one-stop information scheme for the whole process management of documents and bills, including internal operation management optimization, logistics receipt and delivery management and personalized value-added services at the end, helping enterprises realize intelligent document and bill management, helping them reduce operating costs and enhance customer stickiness.

  In the 2020 annual performance briefing, the company also stated that SF B2B business parts (industrial and commercial express) are the main source of income and profit. With the progress of science and technology, the popularity of electronicization, including the disappearance of the demand for mailing invoices, the industry and commerce are challenged to some extent. SF creates some time-sensitive segmented products in the supply chain, turns passive into active, and maintains the advantages of the company’s time-sensitive parts, as follows:

  First, continue to deepen the B-side and dig deep into some industrial and commercial needs. The aging products created by the company can improve the aging of products, and then from the perspective of management, they can create greater value for customers.

  Second, use data to reflect the specific scenes in the supply chain. Instead of the traditional supply chain, the company uses technology and express delivery network to build a brand-new supply chain. After these links are opened by technology, the products can be reproduced and fissile quickly, and many SF resources can be used, such as Ezhou Airport, airplanes and transportation resources with obvious advantages. By using technology and data to penetrate into all walks of life, many common scenes can be extracted and seamlessly connected with SF Express’s network. There is an internal project called "Lego", which is to let customers use the Lego model to combine on the billboard.

  Broaden the extension of aging parts for potential customers

  According to the research of CICC analysts, the persistence, stability and reliability of SF Express’s service during the epidemic period have improved the recognition of its brand in the public mind, and SF Express has a good mass base.

  Zhao Xiaomin, an expert in the express delivery industry, is concerned about SF Express’s comprehensive ability in the fresh season. He believes that in recent years, the competition in the field of "fresh fruit" in the express delivery industry is extremely fierce, and SF Express has adopted multi-arms joint operations in the field of fresh fruit, from providing ordinary transportation to providing industry solutions and continuing to customized solutions, which has solved the blank of standardization in the field of fresh fruit in the past.

  In the consumption upgrade of online shopping, there are a large number of high-end e-commerce brands and express delivery needs that require high timeliness services, such as fruits and fresh food, and these are the potential customer groups of SF Express Service.

  SF’s powerful "network management and control" ability and aggregated distribution network have solved the product sales for the majority of fruit farmers and made great contributions to increasing production and income. The support of national policies for the logistics industry has also made SF Express’s aging parts have a new growth dividend, and the company’s market share in high-end aging parts may be further enhanced.

Reporting/feedback

Huatai computer: measuring the global AI computing space from the evolution of large model

We believe that from the evolution path of the large model, the volume of the model will be further expanded, which will bring about the continuous growth of computing power demand. In the long run, the operation of the mature big model is expected to bring an incremental server market of $316.9 billion, which still has a large room for growth compared with the global AI server market of $21.1 billion in 2023. Based on this, we believe that the continuous iteration of the large model is expected to bring a large number of computing infrastructure needs, and it is recommended to pay attention to the investment opportunities in the computing industry.

Core view

Global demand for AI computing power continues to rise

With the continuous iteration of the large model, the capability of the model is constantly enhanced, which is the result of the increasing number of model parameters and data sets under "Scaling Law". We believe that from the evolution path of the large model, the volume of the model will be further expanded, which will bring about the continuous growth of computing power demand. Specifically, the demand for computing power of large models is reflected in three links: pre-training, reasoning and optimization. According to our calculation, taking the 100 billion parameter model as an example, the total computing power of the three links is about 180,000 PFlop/s-day, which corresponds to 28,000 pieces of A100 equivalent GPU computing power. In the long run, the operation of the mature big model is expected to bring an incremental server market of $316.9 billion, which still has a large room for growth compared with the global AI server market of $21.1 billion in 2023. Based on this, we believe that the continuous iteration of the large model is expected to bring a large number of computing infrastructure needs, and it is recommended to pay attention to the investment opportunities in the computing industry.

The volume of the model is getting bigger and bigger, which drives the demand of computing power construction.

Large Language Model (LLM) is a model pre-trained on a large number of data sets, which shows great potential in dealing with various NLP tasks. The appearance of Transformer architecture opens the way for the evolution of large-scale models. With the increasing number of decoding modules, the parameters of the model continue to increase, and gradually evolve into different versions of models such as GPT-1, GPT-2, GPT-3, PaLM, Gemini, etc., and the parameters also increase from billions to billions and trillions. We can see that the evolution of each generation of models has brought about the enhancement of capabilities, and a very important reason behind it lies in the growth of parameters and data sets, which has brought about the continuous improvement of model perception, reasoning and memory. Based on the scaling law of the model, we believe that the future model iteration will continue the path of larger parameters and evolve more intelligent multi-modal capabilities.

The computing power requirements of large models are embodied in: pre-training, reasoning and optimization.

From the perspective of disassembly, the computing power demand scenarios of the large model mainly include pre-training, Finetune and daily operation. For the three parts of computing power demand, our calculation ideas are as follows: 1) pre-training: based on the assumption of "Chinchilla scaling law", the amount of calculation can be described by formula C≈6NBS; 2) Reasoning: Based on ChatGPT traffic, the amount of calculation can be described by formula C≈2NBS; 3) Tuning: Backstepping by tuning the required GPU core hours. Taking the pre-training/reasoning/optimization of the 100 billion parameter model as an example, the computational power requirements of the three links are 13889, 5555.6 and 216 PFlop/s-day respectively. We believe that under the blessing of Scaling Law, with the growth of model volume, the demand for computing power is expected to be released continuously.

Infrastructure demand is expected to continue to be released, focusing on investment opportunities in computing industry.

Combined with the calculation of computing power demand for pre-training/reasoning/optimization of large models, we predict that the demand for A100 equivalent GPU will be 28,000 from development to mature operation of a 100 billion model. According to our calculation, the operation of the mature big model is expected to bring the global AI server incremental market of $316.9 billion. In contrast, according to IDC, the global AI server market will be 21.1 billion US dollars in 2023, and it is estimated that CAGR will reach 22.7% in 2024-2025, and there is still much room for growth in the future. In addition, considering the limited access to high-performance chips in China, the localization of AI GPU is also expected to accelerate further.

Risk warning: macroeconomic fluctuation, lower-reaches demand less than expected, and the calculation results may be biased.

main body

"Scaling Law" drives the demand for large-scale model computing power to grow continuously.

The appearance of Transformer opens the way for the evolution of large models. Large Language Model (LLM) is a model that is pre-trained on a large number of data sets, and the data is not adjusted for specific tasks. It shows great potential in dealing with various NLP (Natural Language Processing) tasks, such as natural language understanding (NLU) and natural language generation tasks. According to the development of LLM in recent years, its routes are mainly divided into three types: 1) encoder routes; 2) codec route; 3) decoder route. From the development characteristics: 1) the decoder route is dominant, which is attributed to the excellent performance of GPT-3 model in 2020; 2)GPT series models keep ahead, or it is attributed to OpenAI’s insistence on its decoder technology; 3) The model closed source has gradually become the development trend of head players, which also originated from the GPT-3 model, and companies such as Google have begun to follow suit; 4) The codec route is still developing continuously, but the number of models is less than that of the decoder route, or due to its complex structure, it has no obvious advantages in engineering implementation.

The large model may evolve towards larger parameters. We see that from GPT-1 to GPT-4 model, and from PaLM to Gemini model, the capabilities of each generation of models are constantly strengthening, and the achievements in various tests are getting better and better. As for the source of capability behind the model, we think that parameters and data sets are the two most important variables. From the scale of one billion to tens, hundreds and trillions, the increase of model parameters is similar to the increase of the number of human synapses, which brings about the continuous improvement of model perception, reasoning and memory. The increase of data sets is similar to the process of human learning knowledge, which constantly strengthens the model’s ability to understand the real world. Therefore, we believe that the next generation model will continue the route of larger parameters and evolve more intelligent multi-modal capabilities.

From the perspective of disassembly, the computing power demand scenarios of the large model mainly include pre-training, Finetune and daily operation. From the practical application of ChatGPT, starting from the framework of training+reasoning, we can further divide the computing power requirements of the large model into three parts according to the scene: 1) Pre-training: training the basic language ability of the model mainly through a large number of unmarked plain text data, and obtaining a basic large model like GPT-1/2/3; 2)Finetune: On the basis of completing the pre-trained large model, conduct two or more trainings, such as supervised learning, reinforcement learning and transfer learning, to optimize and adjust the model parameters; 3) Daily operation: based on the information input by users, the model parameters are loaded for reasoning and calculation, and the feedback output of the final result is realized.

Pre-training: the demand for computing power is expected to continue to grow under the scaling law

The pre-training effect of large model is mainly determined by parameter quantity, Token quantity and calculation quantity, and it satisfies the "scaling law". According to the paper Scaling Laws for Neural Language Models published by OpenAI in 2020, in the process of large language model training, the parameters, the number of Token and the amount of calculation have a significant impact on the performance of large models. For the best performance, these three factors must be amplified at the same time. When it is not restricted by the other two factors, the model performance has a power law relationship with each individual factor, that is, it satisfies the "scaling law".

OpenAI thinks that the calculation amount of model pre-training can be described by formula C≈6NBS. According to the paper Scaling Laws for Neural Language Models published by OpenAI in 2020, the computational power (c) required for pre-training a Transformer architecture model is mainly reflected in the process of forward feedback () and backward feedback (), and is mainly determined by three variables: model parameter quantity (n), Token batch consumed in each training step (b) and iteration times required for pre-training (s). Among them, the product of b and s is the total number of Token consumed by pre-training. Based on this, we can use C≈6NBS to describe the computational power required for pre-training of large models.

Among them, OpenAI believes that the model parameter is the most important variable, and the larger the parameter, the better the model effect. OpenAI believes that as more computing becomes available, model developers can choose how much to allocate to train larger models, use larger batches, and train more steps. Assuming that the amount of computation increases by one billion times, most of the increase should be used to increase the model size in order to obtain the optimal training of computational efficiency. In order to avoid reuse, only a relatively small data increment is needed. Among the increased data, most of them can increase the parallelism by larger batch size, and the required serial training time only increases a little.

Google put forward "Chinchilla scaling law", which holds that model parameters and training data sets need to be scaled up in equal proportion to achieve the best results. According to "Training Compute-Optimal Large Language Models" published by Google DeepMind in 2022, the relationship between the model performance and the number of tokens and the amplification of parameters required for model pre-training is not linear, but the best model effect can be achieved when the number of model parameters and the number of training tokens reach a certain ratio. In order to verify this rule, Google trained a 70-billion-parameter model ("Chinchilla") with 1.4 trillion Tokens, and it was found that its effect was better than that of Gopher, a 280-billion-parameter model trained with 300 billion tokens. Further research by DeepMind found that the approximate relationship between the parameter quantity of the optimal language model and the data set size satisfies: D=20P, where d represents the number of Token and p represents the parameter quantity of the model, that is, it satisfies the "Chinchilla scaling law" at this ratio.

We estimate that the computational power required to train the 100 billion parameter model is above 10,000 PFlop/s-day. We assume that the models with different parameters and volumes all satisfy the "Chinchilla scaling law", so as to calculate the optimal data set size required by different models and the computational power required for pre-training. Taking a large language model with 100 billion parameters as an example, the number of training Token required under "Chinchilla scaling law" is 2 trillion. According to the computational formula C=6NBS proposed by OpenAI, it can be calculated that the computational power required to train the 100 billion parameter model is about 1.39X10 4 pflop/s-day. Similarly, the computational power required for training a 500 billion parameter model is about 3.47× 10 5 p flop/s-day, and that for training a 1 trillion parameter model is about 1.39× 10 6 p flop/s-day.

Reasoning: High concurrency is the main driving force of reasoning computing requirements.

The underlying architecture of GPT model is composed of decoder modules. In a large language model such as GPT, the decoding module is equivalent to the basic architecture unit, and the underlying architecture of GPT model is pieced together by stacking each other. The number of decoding modules determines the scale of the model. Generally, GPT-1 has 12 modules, GPT-2 has 48 modules and GPT-3 has 96 modules. The more modules there are, the greater the model parameters and the larger the model volume.

The decoding module realizes large model reasoning by calculating tokenized text data. According to the paper "Scaling Laws for Neural Language Models" published by OpenAI in 2020, after the large model is trained, the model itself has been fixed, and the reasoning application can be carried out after the parameter configuration is completed. In essence, the reasoning process is to traverse the parameters of the large model again. By inputting the vector encoded by the text, the result is output and converted into words through the calculation of the attention mechanism. In this process, the parameters of the model depend on the number of model layers, the number of feedforward layers, and the head of attention mechanism layer.

The computational power required in the reasoning process can be described by the formula C≈2NBS. Because the decoding module mainly performs forward propagation in the process of reasoning, the main calculation amount is embodied in text coding, attention mechanism calculation, text decoding and so on. According to the calculation formula given by OpenAI, every time a Token is input and goes through such a calculation process, the required amount of calculation is =2N+2, in which the second half of the formula mainly reflects the size of the context window. Because this part accounts for a small proportion of the total amount of calculation, the required bytes are often expressed in K level, so it is often ignored in the calculation. Finally, we get that the calculation requirement of large model reasoning is the product of a single calculation and the number of Token, that is, C≈2NBS.

With the same number of visits to ChatGPT, we expect that the computing power required for the reasoning of the 100 billion model will be more than 5000 pflop/s. According to Similarweb data, ChatGPT visited official website 1.8 billion times in March 2024. We assume that there will be 10 questions and answers for each user visit, and the number of tokens consumed in each question and answer is 800, so it is calculated that the number of tokens consumed by ChatGPT official website in April is 06 million. Considering that the computing infrastructure construction is determined according to the peak demand rather than the average demand, we further assume that the peak Token demand is five times the average. Finally, assuming that different parametric models have the same access to ChatGPT, according to the formula of C≈2NBS, the reasoning power requirements per second of 1000, 5000 and 1000 billion parametric models are 5555.6, 27777.8 and 55555.6 PFlop/s, respectively.

Tuning: The computing power demand mainly depends on the tuning times.

After the pre-training, the parameters of the large model need to be optimized to meet human needs. Generally speaking, after the pre-training, large language models need to be continuously Finetune to achieve better running results. Taking OpenAI as an example, the process of model tuning adopts human feedback mechanism (RLHF). Reinforcement learning guides model training through Reward mechanism, which can be regarded as the loss function of traditional model training mechanism. The calculation of reward is more flexible and diverse than the loss function (for example, the reward of AlphaGO is the outcome of the game), but the price is that the calculation of reward is not derivative and cannot be directly used for back propagation. The idea of reinforcement learning is to fit the loss function through a large number of samples of rewards, so as to realize the training of the model. Similarly, human feedback is not derivable, and it can also be used as a reward for reinforcement learning, resulting in reinforcement learning based on artificial feedback.

Taking ChatGPT as an example, the tuning process mainly goes through three steps. Based on the reinforcement learning technology of human feedback, the tuning process of ChatGPT is mainly divided into three steps: 1) training supervision model; 2) Training reward model; 3) Strengthen the learning of PPO parameters. After optimization, the parameters of the model will be updated, and the generated answers will be closer to the expected results of human beings. Therefore, the demand for computing power in the tuning process is actually similar to that in the pre-training, and the model parameters need to be traversed, but the data set used will be much smaller than that in the pre-training.

The computational power requirement of large model tuning can be reversed by the number of GPU kernel hours needed for tuning. For the computing power requirement of large model tuning, we use the method of actually consuming GPU core hours to push back. According to Deepspeed Chat (Microsoft’s service provider focusing on model tuning), it takes 9 hours to tune a 13 billion model, using 8 A800 accelerator cards. According to NVIDIA official website, the peak computing power of A800 accelerator card is about 312 TFLOPS(TF32, using sparse technology). According to this calculation, it takes about 0.9 PFlop/s-day to tune a 13 billion parameter model. By analogy, the computational power required for once tuning the 30 billion, 66 billion and 175 billion parametric models is 1.9, 5.2 and 8.3 PFlop/s-day respectively.

We estimate that the computing power required to tune the trillion parameter model every month is above 2000PFlop/s-day. For the convenience of comparison, we further assume that the models with different parameters are all tuned by a single A800 server instance (that is, eight A800 accelerator cards), and the training duration is proportional to the model parameters. In addition, considering the problem of tuning times, we assume that large model manufacturers need to tune the model 30 times a month. Based on this, we calculated that the computational power required for the monthly optimization of 100 billion parameter models is 216 PFlop/s-day, and that for the monthly optimization of 1 trillion parameter models is 2160 PFlop/s-day.

The demand for computing infrastructure is expected to continue to be released, paying attention to the opportunities of computing industry.

Large-scale model training/reasoning/tuning brings computing hardware requirements. At present, the mainstream method is to use AI server to carry the computing requirements of large models, and the core devices are AI GPU, such as NVIDIA A100, H100, B100, etc. According to NVIDIA, the peak computing power of a single A100 accelerator card TF32 is 312 TFLOPS (using sparse technology) and that of FP16 is 624 TFLOPS (using sparse technology). Considering that in the actual workload, multi-card interconnection is often used for model training and reasoning, it is necessary to consider the problem of effective calculation. According to the GPT-Neox-20b: An Open-Source Autoregressive Language Model published by Sid Black et al. in 2022, the effective computing power of a single card is about 117 TFLOPS(TF32, using sparse technology), that is, the effective computing power ratio is 37.5%. We assume that the effective computing power ratio of the reasoning process is equivalent to that of the training process, and the reasoning computing power of a single card is 234 TFLOPS(FP16, using sparse technology).

We estimate that the demand for A100 equivalent GPUs for training/reasoning/tuning of the 100 billion model is 28,000. For the number of computing infrastructure needed for large models, we measure it by the number of GPU/ servers. According to our calculation framework, the total demand for computing power of large model is the sum of computing power demand of pre-training, reasoning and optimization. Considering that after the model pre-training, the infrastructure such as servers will usually be used for the development of the next generation model, we assume that the computing power requirements of pre-training, reasoning and tuning will occur concurrently. In addition, we assume that the training, reasoning and tuning are all completed within one month. Based on this, it is estimated that the demand for A100 GPU for the 100 billion parameter model is 28,000, that for the 500 billion model is 218,000, and that for the 1 trillion model is 634,000. We further assume that all servers are integrated with eight A100 acceleration cards, so the demand for AI servers in the 1000, 5000 and 1000 billion parameter models is 0.3, 27,000 and 79,000 respectively.

The operation of the mature big model is expected to bring the AI server market space of $316.9 billion. According to the Research Report of China Artificial Intelligence Large Model Map released by China Institute of Science and Technology Information, as of May 2023, 202 large models have been released in the world, and the number of large models in China and the United States accounts for nearly 90% of the global total. We predict that the number of large models in the world is still increasing, but with the iteration of large models, the competition among model manufacturers will gradually become balanced. Based on this, we conservatively assume that in the future, 30 manufacturers will realize the mature operation of 100 billion parameter models, 20 manufacturers will realize the mature operation of 500 billion parameter models, and 10 manufacturers will realize the mature operation of 1 trillion parameter models. According to JD.COM, a single Inspur NF5688M6 server is equipped with eight A800 accelerator cards, and the price is 1.59 million yuan/set, which is about 220,000 US dollars/set according to 1:7.23 USD. Based on the server demand of the above different models, we estimate that the server demand of global large model manufacturers is $316.9 billion.

In contrast, the current global AI server market is only $21.1 billion, and there is still much room for growth. According to Gartner, the global AI chip market will reach 53.4 billion US dollars in 2023, and the year-on-year growth rate is expected to reach 25.7% in 2024. According to IDC, the global AI server market will reach $21.1 billion in 2023, and it is estimated that the market will reach $31.8 billion in 2025, and the CAGR will reach 22.7% in 2024-2025. In contrast, the continuous competition and mature operation of global large model manufacturers are expected to bring about a space of 316.9 billion US dollars, while the current market size is only 21.1 billion US dollars, and there is still much room for growth. We believe that with the emergence of global large-scale models and AI applications, the demand for training/reasoning/optimization is expected to drive the rapid growth of computing infrastructure construction.

Under the background of localization, domestic AI GPU is expected to accelerate the catch-up. On October 17th, 2023, the Bureau of Industry and Security (BIS) of the U.S. Department of Commerce issued export restrictions on advanced computing and semiconductor manufacturing items in China, and domestic imports of high-performance AI chips were restricted. On the other hand, we also see that there is still a gap between domestic AI GPU and overseas advanced level. Among domestic AI GPU, Atlas 300T based on Huawei Ascent 910 has strong computing performance, and FP16′ s computing performance is about 90% of NVIDIA A800 SXM without considering sparse technology. However, compared with the most advanced products such as B100 in NVIDIA, there is still a gap of at least two generations. We believe that under the background of limited import of AI chips, the localization of AI GPU is expected to accelerate, and under the technical iteration, the gap between home and abroad is expected to gradually narrow.

To sort out the industrial chain companies, please see.The original research report.

Risk warning

Macroeconomic fluctuation. If the macro-economy fluctuates, the pace of industrial transformation and the landing of new technologies may be affected, and the macro-economic fluctuation may also have a negative impact on IT investment, resulting in the overall industry growth being less than expected.

Downstream demand is less than expected. If the downstream demand for computing power is less than expected, the related computing power input will increase or be slower than expected, resulting in the industry growth being less than expected.

There may be deviations in the calculation results. In this paper, assumptions such as "scaling law" and "Chinchilla scaling law" are used in the calculation process, which is subjective to some extent. If it is inconsistent with the actual model training process, it may lead to deviation in the calculation force demand.

Related research report

Research report: "Global AI computing power demand continues to rise" April 12, 2024

This article comes from: Selected research reports of securities firms.

Reporting/feedback

Du Xiaohua, another prototype of the film Dear, painted his son and his wife said that she had lived in expectation for ten years.

03:26
Public Network Poster Journalist Wu Junlin Jie Qiangmin Jinan Report
On December 30th, Du Xiaohua, one of the prototypes of the film Dear, his wife and little daughter came to Jinan with the Inner Mongolia police, and asked Lin Yuhui, a simulated portrait expert, to paint his lost son for 10 years. Referring to the appearance of Du Xiaohua and his wife, Lin Yuhui drew Du Houqi’s 17-year-old "appearance" based on the photos of Du Houqi’s childhood and the appearance of Du Houqi’s younger brother.
During the interview, Du Xiaohua specially arranged the T-shirt, so that the photo of his son, missing information and his phone number on the T-shirt appeared completely in front of the camera. Du Xiaohua is a native of Shangrao, Jiangxi. His eldest son is Du Houqi, born in 2005. In 2011, Du Xiaohua took his wife and children to work in Nancun Woodware Cabinet Factory, Dingdu Longgui Village, Qingshan District, Baotou City, Inner Mongolia. On the evening of March 6, Du Houqi disappeared.
Du Xiaohua said that his son had a tiny scar on the upper part of his nose.
"During the day, we agreed between adults that we wanted to take a bath in the street or buy some daily necessities. After dinner, I went from the east factory to the west factory to find my colleagues and see if anyone went. My son followed me from behind and asked me if I wanted to go to the street. I said that it was so cold and windy today, and the boss’s car was not at home, so I couldn’t go. " Du Xiaohua said that before his son came to see him, his wife told him that if he didn’t take a bath in the street, he would go home quickly and boil some water to wipe his body.
"My son learned that I stopped going to the street and went back alone. At this time, I took a phone call and didn’t notice my son going back." Du Xiaohua said that at about 7: 40 that night, his wife asked where his son had gone, and he found that the child was gone. In just 10 minutes, my son disappeared from the aisle seven or eight meters between the east and west factories!
In the panic, workers and villagers helped to find it, and the police also came to investigate, but the son was so "into thin air".
Du Xiaohua said that his son has a great feature. If he cries very sadly, he will have a nosebleed. It was very cold the night he disappeared, and his son had a slight cold. He was worried that if his son was forcibly abducted by traffickers and struggled to cry, he would bleed profusely.
"After the loss of our son, especially on holidays, his birthday and the day of his disappearance, we are extremely sad. We dare not imagine where he is. I don’t know how he is living. I don’t know if he has food or clothes. Will he wander in the street?" In the past ten years, Du Xiaohua has traveled all over the country to find his son, and has been to all provinces except Xizang, Hainan, Hong Kong, Macao and Taiwan. His wife, Jin Nana, does business at home and makes a living.
The most depressed time was from 2015 to 2017, and there was no news from my son, and the clues were pitiful.
Du Xiaohua and his wife turned to Lin Yuhui, a simulated portrait expert, for help.
After the son disappeared, the family advised Du Xiaohua and his wife to have another one. "Otherwise, it will be ruined."
"When I registered my second son, our local department wanted to cancel my eldest son’s account. I have been begging not to cancel it. I said that if my eldest son’s account is cancelled, what will he think of us on the day he comes back? " Du Xiaohua said, "I have been telling my second and third children that if dad really can’t walk, you can help mom and dad get his brother back."
Portrait of Du Houqi
Because Du Xiaohua was on the phone when his son disappeared, his wife always complained about him. "Every time he comes home, I am not happy to see that he came back by himself. Even if he is injured, I feel nothing, as long as I can get my son back. " Jin Nana said that for many years, he has been living in expectation, expecting his son to call suddenly. "My son has no reason not to remember his mother."
Du Xiaohua said that in the old mobile phone, there was a jigsaw puzzle of Mona Lisa. Every time my son put it together, he said that he had put his mother together.
After Sun Haiyang, another prototype of "Difficult Friend" and "Dear", found his son Sun Zhuo, Du Xiaohua followed Sun Haiyang and his son to Hubei and Shandong to seek exposure, hoping to gain clues about his son.
"Recently, there are so many clues on his mobile phone that I can’t see them. I also help him see them. Every clue feels like a hope." Jin Nana said, "I feel very lucky to find children. I hope we can be so lucky."
Du Xiaohua said that he was more and more confident in the "reunion" action carried out by the Ministry of Public Security after Guo Gangtang and Sun Haiyang, the prototype of Lost Orphan, and he was more eager to have his son back.
Lin Yuhui revised Du Houqi’s portrait
Reporting/feedback

The relationship has deteriorated! Japan will restrict the export of high-tech materials to South Korea, and South Korea says it retaliates.

       CCTV News:According to Yonhap News Agency, the Minister of Industry, Trade and Resources of the Republic of Korea, Chang Yunmo No.1, said in Seoul that the South Korean government would file a lawsuit against Japan for restricting the export of semiconductor raw materials to the Republic of Korea, and would take necessary measures to counter it in accordance with the principles of international law and domestic law.

       Chang Yunmo, Minister of Industry, Trade and Resources of Korea, characterized Japan’s export control to Korea as economic retaliation. Cheng Yunmo also said that in order to guard against Japan’s unilateral measures, the South Korean government has been working with relevant industries to diversify the import market and key technologies, and expand domestic production facilities. In the future, we will maintain close communication with the industry, strive to minimize the losses of Korean enterprises, and take this opportunity to enhance the competitiveness of Korean parts, materials and equipment industries.

one

       It is understood that the high-tech materials that Japan intends to restrict Japanese enterprises from exporting to South Korea include fluorine-containing polyimides, resists and high-purity hydrogen fluoride. These materials are important raw materials in smart phones, chips and other industries. From July 4th, Japan will require enterprises to apply for permission from the government before exporting the above materials to the ROK, and the approval process will take about 90 days.

       It is reported that the fluorine-containing polyimides and resists manufactured by Japanese companies account for about 90% of the global output and hydrogen fluoride accounts for about 70%, respectively. Once export restrictions are imposed on South Korea, it will impact major Korean electronic products companies such as Samsung Electronics and LG Electronics.

one

       According to the Kyodo News Agency quoted the Korea Trade Association, in 2012, South Korea’s semiconductor exports were US$ 50.4 billion, accounting for only about 9% of the total exports, while in 2018 it increased to US$ 126.7 billion, accounting for about 21%. In this regard, some experts said that South Korea’s economic growth is "over-dependent" on the export of Korean semiconductor industry, and Japan’s export restrictions will hit the Korean semiconductor industry and have a huge impact.

       On the 1st, Japan’s Ministry of Economy, Trade and Industry announced that the Japanese government would restrict the export of important high-tech materials to South Korea from July 4th, because it was dissatisfied with South Korea’s demand for Japanese enterprises to compensate South Korean workers forced by Japan during the Second World War. Since last year, Korean courts have successively ruled that three Japanese companies should make compensation for the forced recruitment of Korean workers during World War II. The Japanese government insists that such claims have been "settled" according to the Japan-Korea Claims Agreement signed by the two countries in 1965. The above-mentioned compensation dispute has become one of the main factors for the recent deterioration of relations between the two countries.

Tsunami in Indonesia: The restaurant was washed away by the 3-meter wave and the cross-sea bridge was washed away.

  "The situation in the affected area is chaotic, people are running in the streets, buildings collapse and ships are washed ashore." Dwikorita Karnawati, head of the Indonesian Meteorological and Geophysical Bureau, described the scene after the September 28 earthquake and tsunami.

  On September 28th, an earthquake of magnitude 7.7 occurred in Central Sulawesi, Indonesia, which subsequently triggered a tsunami.

  According to the statistics of Indonesian Disaster Reduction Agency on the 29th, the earthquake and tsunami have caused 384 deaths and hundreds of injuries. With the advance of rescue work, the death toll of tsunami caused by earthquake in Indonesia is rising further. The Indonesian Disaster Reduction Agency also said that many "bodies" were found on the beaches in Indonesia.

  The above-mentioned institutions also said that due to the paralysis of local communications, they have not received a more comprehensive report of casualties for the time being. Nugroho, spokesman of Indonesian Disaster Relief Agency, said the impact of the disaster was "extensive", with thousands of houses washed away and some people missing.

  "This is the strongest earthquake I have ever encountered, and we all ran out of the building." Yanti, a 40-year-old housewife from East Gala, said.

  Mohammed Fikree, another resident from East Gala, said that he fled from his home after the earthquake, but he did not encounter large-scale panic.

  "Everything in my house is shaking and cracks appear on the wall." Fikree said, "This is not the first time (earthquake). The last earthquake caused us a stronger sense of shock, so this time we didn’t panic very much. We fled the building and now everything is back to normal."

  According to Singapore’s Lianhe Zaobao, the first earthquake occurred at about 2 pm local time yesterday (1 pm Beijing time), with a magnitude of 6.1, and another earthquake with a magnitude of 7.5 occurred about 4 hours later. The epicentres of the two earthquakes were located 78 kilometers north of Palu, the capital of Central Sulawesi Province. The second earthquake triggered a tsunami.

  After the tsunami, Indonesia’s local TV station broadcast a video taken with a mobile phone on the 29th. In the video, when the distant waves suddenly hit the coast of Palu, people screamed and ran away in panic. The photographer of this video is located in a tall building not far from the coast. The coastal town of Palu has a population of 350,000.

  Where the lens swept, the small restaurant set up along the beach was washed away by the waves, and the city streets were full of building debris wrapped in seawater, and the telephone poles were skewed. In the picture, a mosque has partially collapsed, but it is not clear whether it is due to the impact of the earthquake or tsunami.

  The city of Palu, Sulawesi province was devastated by the earthquake and tsunami. According to the BBC, this video was shot about 80 kilometers from the epicenter of the earthquake.

  According to previous media reports, the 3-meter-high tsunami triggered by the earthquake in Central Sulawesi province hit two cities, Palu and Donggala, and many residential areas were invaded by seawater.

  Photos spread in the media and social networking sites showed that some victims’ bodies were wrapped in blue body bags, parked on the roadside where building debris remained, and survivors walked by. Some affected hospitals transferred the wounded at the first time.

  The body of the victims temporarily parked on the roadside, Palu City, was built according to makassar strait, and the narrowness of the strait increased the destructive power of the tsunami impact. On the streets of the city, fragments of buildings are scattered everywhere. In the nearby city of Donggara, a sea-crossing bridge with a yellow arch was washed away by huge waves.

  Initially, the Indonesian Meteorological and Geophysical Bureau declared that earthquakes would not cause tsunamis. Subsequently, Dwikorita Karnawati, the head of the agency, confirmed to Reuters that the tsunami hit the city of Palu in makassar strait, which connects Celebes and the Java Sea.

  The disaster situation in nine villages is still unknown.

  To make matters worse, the lack of electricity and communication has made the search and rescue work more difficult.

  After a staff member of the Central Sulawesi Museum in Palu confirmed the impact of the tsunami on the museum to Jakarta Post, the media reported that the communication connection of the mobile phone "collapsed".

  Mirza ali Sam, a resident from Kendari, the capital of southeast Sulawesi province, told the media that his uncle’s family of five was on vacation in Palu, and he had been unable to get in touch with them since the tsunami hit.

  Nugroho, spokesman of Indonesian Disaster Relief Agency, said that the Indonesian military is deploying troops to Palu and Donggara, and the police have mobilized to help with emergency response. Search and rescue personnel and disaster relief agency personnel are working at the same time.

  At present, the evacuation work is in progress, and people are warned to be vigilant and stay outside.

  "People are encouraged to gather in safe places, away from hillsides." Nugroho said. "It is best not to enter houses or buildings, because there is a great possibility of aftershocks." After the magnitude 7.7 earthquake on the evening of 28th, there have been many aftershocks, including a magnitude 6.7 aftershock.

  "Many houses collapsed," Akris, a staff member of the Indonesian Disaster Relief Agency, told the Associated Press that the disaster situation in nine villages is still unknown.

  The main airport in Palu was also closed after the tsunami and is expected to be closed for at least 24 hours. Nugroho said that helicopters can still land at the airport if necessary, but AirNav, which is used to monitor the flight system, shows that the runway of the airport has cracked and the control tower has been damaged.

  Over the past month or so, there have been many earthquakes in Indonesia, including an earthquake measuring 5.9 on the Richter scale in northern Sulawesi province of Indonesia on September 8, an earthquake measuring 5.5 on Sombawa island of Indonesia on August 26, and several earthquakes in Lombok, Indonesia during August. This series of earthquakes has caused 515 deaths and 7,145 injuries.

  On the evening of 28th, Indonesian Prime Minister Joko said that he had instructed the Minister of Security to coordinate various government departments to deal with the earthquake and tsunami disaster in Central Sulawesi province.

  Stephane Dujarric, a UN spokesman, said that UN officials are in contact with Indonesian authorities and are "ready to provide support as needed".

  Indonesia is located in the Pacific volcanic belt, where the continental plates meet, resulting in frequent earthquakes and volcanic activities. The worst is the December 2004 earthquake, when a super strong earthquake of magnitude 9 or above occurred on the Indian Ocean seabed west of Sumatra, which triggered a tsunami and killed 226,000 people in 13 countries in Southeast Asia and South Asia. In Indonesia, 168,000 people were killed.

The second generation of Haval H9 shock hit: Highlander and Prado combination? Price only 199,900!

[ITBEAR] September 26th news, after ten years of expectation, the second generation H9 finally made its debut. This new model officially met the public on September 25th, bringing three different versions of Exploration, Extreme, and Extreme. The price range is set to 199,900 to 229,900 yuan, which is 6,000 yuan lower than the pre-sale price. What is more noteworthy is that compared with the 2022 models that have been discontinued, the entry price of the new car has been reduced by 14,900, and the top version has been greatly reduced by 49,900, which undoubtedly increases its market competitiveness.

According to ITBEAR, the price range of the second-generation Haval H9 is similar to that of the 2.T model of the Tank 300, but the positioning of the two is different. The Tank 300 prefers a pure off-road experience. However, the second-generation Haval H9 is not only satisfied with providing strong off-road performance, but also has made significant improvements in comfort, striving to satisfy the driver’s off-road fun while also bringing a comfortable ride experience to the rest of the family.

In terms of spatial layout, the second-generation Haval H9 shows its unique charm. It has a 2850mm wheelbase, and its body size exceeds that of similar family cars such as Highlander and Wenjie M7. It is particularly worth mentioning that its two rows of seats can be completely flat to form a comfortable queen bed up to 1.8 meters long, providing great convenience for family travel.

In the seat design, the second-generation Haval H9 also spares no effort. The ergonomically designed seat is more in line with people’s physiological curves. The backrest of the seat adopts a 9-layer structure design, which adds multiple functional layers compared to the traditional 5-layer structure, and is wrapped with skin-friendly and environmentally friendly materials. In addition, the seat cushions and flanks have been customized to lengthen and widen, providing better wrapping and support.

In order to further enhance the tranquility of the ride, the second-generation Haval H9 has also made many efforts in sound insulation and noise reduction of the vehicle. For example, the use of double-layer laminated sound insulation glass, the design of multiple laminated panels on the body, and the optimized exhaust noise control system. These meticulous designs are all designed to provide passengers with a more peaceful and comfortable riding environment.

As an off-road vehicle that focuses on family use, the second-generation Haval H9 is no less intelligent in terms of configuration and safety performance. It is equipped with 8155 chip, which makes intelligent voice interaction more sensitive and user-friendly. At the same time, it also supports seamless interconnection with mainstream domestic smartphones, adding more convenience and fun to the journey. In terms of safety, the second-generation Haval H9 adopts a non-carrier body design and is equipped with an L2-level driver assistance system, providing a full range of safety for long-distance off-road travel.

Since the first generation Haval H9 came out in 2014, its cumulative sales have exceeded 140,000 units, which not only accumulated a wealth of hard-core SUV experience and user reputation for cars, but also laid a solid foundation for the prosperity of the Chinese brand off-road vehicle market. Now, the second generation Haval H9 has debuted with a new attitude, not only meeting the needs of off-road enthusiasts, but also taking into account the actual needs of family users, opening up a new market segment of "family off-road". However, under the impact of the new energy vehicle wave, whether the second generation Haval H9 can stand out in the market remains to be further tested by the market.

#2nd generation Haval H9 ##family off-road ##comfortable experience ##strong off-road ##smart configuration #

Source: http://www.itbear.com.cn/html/2024-09/519515.html