The second round of the Chinese Super League has come to an end after three days of fierce fighting. Combined with these two rounds, people feel that the pace of the game is faster, the disputes of the referees are less, and the control of the game is more reasonable. It is not difficult to see that after the storm of anti-corruption and anti-gambling, it still has a certain binding force on players and referees, so that everyone can focus on the game itself, rather than other shortcuts to get rich. Of course, the first two rounds of the league are also interesting in terms of ranking. Let’s take a concrete look at the points ranking.
The top three teams are the only teams that have won two consecutive victories. Changchun Yatai ranked first after two rounds, which really surprised many people. It was a dark horse to beat Zhejiang team and Dalian people one after another. Changchun Yatai used to be known as a giant killer, but now Chen Yang’s class is gradually decreasing. The game between Harbor and Shenzhen team is soul-stirring, even the best game since the start of the game. There are all kinds of plots, such as red card, winner and anti-winner. Lei Wu is still worth relying on. Shenhua’s two-game winning streak is equally gratifying.
Chengdu defeated Cangzhou Lions in the home atmosphere of nearly 40,000 fans, and Daniel F. Akerson scored his first goal after returning home, which made the victory more meaningful. It’s quite surprising that Jinmenhu sent Zhejiang team to lose two games in a row at home. Yu Genwei’s team is getting better and better. The story of the Northeast Derby between Dalian people and Changchun Yatai is also ups and downs. In 94 minutes, Mamba was absolutely flat, and in 96 minutes, Liao Zhijian was the winner. Although the Shenzhen team lost the game, it was still a glorious defeat. Aside from the problem of unpaid wages of the club, this team is still worthy of respect.
The home game of Taishan Team and Nantong Zhiyun drew 1-1, and Hao Wei was absent from the game. At the same time, there were many conflicts between the two teams in this game, and after the game, it was even more dangerous to brew a group fight, which added some talk for this game. It is hard for Taishan team to win 1 game after 2 rounds, and it is caught in many problems such as the absence of the head coach and the main striker. The fans are looking forward to an early rebound after such a difficult start. In addition, the current record and ranking of Cangzhou Lions and Qingdao Manatee are also in line with expectations, and the paper strength of the two teams is still in the middle and lower reaches of the Super League as a whole.
The last three teams are really surprising. Meizhou Hakka was the upper-middle team last season, Zhejiang was the top three, and Wuhan Sanzhen was the defending champion. The three towns are also plagued by the main injuries. This campaign has become a 1-1 draw with Guoan, and the road to defending the title is quite difficult. Zhejiang team is the only team that has lost two games in a row at present. It started so badly. I hope the team will rebound as soon as possible.
Data elements are becoming more and more important.
On March 7, 2023, according to the Xinhua News Agency, the National Data Bureau was established according to the State Council’s proposal to submit the institutional reform plan of the State Council.
The National Bureau of Data will be responsible for coordinating and promoting the construction of basic data systems, coordinating the integration, sharing, development and utilization of data resources, and coordinating the planning and construction of digital China, digital economy and digital society, which will be managed by the National Development and Reform Commission.
The establishment of the National Data Bureau will be more conducive to mobilizing all resources to comprehensively promote the construction of digital China and the development of digital economy.
"The most practical contradiction in front of data circulation is the ownership of data and the security of data. Data ownership and security issues are big issues involving public interests, and no institution or enterprise can solve them. " Guo Zhaohui, chief scientist of Shanghai Youye Information Technology Co., Ltd., told Tiger Sniff.
Therefore, the consensus of many experts and industrial people is that the establishment of a national data bureau can better solve the urgent problems faced by the development and utilization of data elements resources at the national level. For example, at present, the domestic data infrastructure system needs to carry out a series of system construction in terms of data confirmation, circulation, transaction, rights distribution, security and compliance, and provide institutional guarantee for the cultivation of data elements market.
Institutional guarantee of data factor market
On February 26th, Peng Lihui, Secretary-General of china electronic chamber of commerce, introduced that in 2021, China’s data element market will be about 81.5 billion yuan.It is estimated that the compound growth rate of market scale will exceed 25% during the "14th Five-Year Plan" period, and the scale is expected to approach 200 billion yuan by 2025.From the perspective of subdivision, the market scale of storage, analysis and processing of data elements all exceeds 15 billion yuan, laying a solid foundation for the resource utilization of data elements; The industrial scale of data transaction and data service reached 12 billion yuan and 8.5 billion yuan respectively.
However, as an important factor of production, the current market system is not perfect, and the basic systems such as data property rights and transaction circulation need to be formulated and improved urgently.
Guo Zhaohui gave an example to Tiger Sniff. Taking the medical industry as an example, from the aspect of medical treatment, the data in China’s medical field is very huge. China has a large population, and every hospital has a large number of patients and diseases data. However, due to the poor data circulation, each hospital can only fight on its own and cannot establish a unified data system. This not only increases the difficulty for patients to seek medical treatment, but also greatly hinders the development and progress of medical system and medical technology. Even from the perspective of technological development, the research and development of medical AI needs to mobilize data from all parties to "feed" AI, but if the data factor market is not opened, and the data in the medical industry is relatively scattered, it is necessary to consider the protection of patient privacy, which makes it difficult for relevant R&D enterprises to obtain data, thus hindering the research and development of technology.
In many industries, some enterprises and research institutions have thought about data alliance, but in practice, such alliance is extremely difficult to promote.For example, for training large models, a large amount of data is needed.This brings huge workload to data cleaning, screening and standardization.They also hope to cooperate with enterprises with a large amount of data, so that the amount of data is larger and the quality is better, but the reality is that enterprises are often reluctant to share these data. Guo Zhaohui said that because these data alliances have no administrative and law enforcement rights, it is difficult to actually promote data circulation.
Even the big data exchange established in the past few years is facing an embarrassing situation. Previously, a survey conducted by Nandu Big Data Research Institute found that at present, local data exchanges generally require enterprises to provide compliance assessment certificates issued by professional lawyers for each transaction before entering the transaction, that is, the so-called "assessment once entering the market". However, this has greatly increased the cost of enterprises’ entry transactions, and the related supervision mechanism is not perfect, which leads enterprises to "dare not enter the market" and "unwilling to enter the market". This actually prevents data resources from being used to a greater extent.
In the enterprise search, "data transaction" is used as the key word to search the enterprise name, and there are 57 operating companies whose operating status is surviving, in business and established. Many major cities and industrial bases in China have set up their own data exchanges, including Beijing, Shanghai, Shenzhen, Guangzhou, Guiyang, Suzhou and Hangzhou. Provincial data exchanges have also been established in Hunan, Hubei and Fujian.
So what is the actual trading volume of these exchanges? Take Guiyang Big Data Exchange as an example. Guiyang Big Data Exchange was officially put into operation in 2015. It is the earliest batch of data exchanges in China. Up to now, the exchange has accumulated 502 data vendors and 21 data intermediaries, with a total of 800 products and 438 transactions, with a total transaction volume of 734 million yuan.
Fan Yuan, chairman of Hangzhou Anheng Information Technology Co., Ltd., has worked hard in the field of data security for many years. According to Fan Yuan,The establishment of the National Data Bureau can better coordinate and promote the construction of the basic data system, and coordinate the integration, sharing, development and utilization of data resources.In this way, it can provide richer data resources and more efficient data acquisition channels for the big data industry, and provide a broader space for big data applications and data element markets.
Fan Yuan also put forward an important aspect, that is, the establishment of data standards. For a long time, there has not been a practical and effective unified standard for the transaction and circulation of data, which has also brought a lot of trouble to data work. Fan Yuan said that the National Bureau of Data will be responsible for promoting data standardization and standardizing various data formats and data structures, which will help improve the efficiency and accuracy of data exchange, accelerate the establishment of a data classification and classification protection system, and make the big data industry more standardized and standardized.
Tighter supervision
"If you don’t open the interface of data sharing, then such a large data resource in China can’t be played well." Guo Zhaohui believes that the development of intelligence can only be achieved based on data sharing. However, the main reasons for the difficulty in data sharing and slow circulation are not only the failure to clearly confirm the rights and formulate standards, but also the safety and compliance issues.
Fan Yuan said that after the establishment of the National Data Bureau, it will be responsible for the supervision and security management of national data resources, which will mean stricter data security standards and regulatory requirements, and make the risks of data leakage and abuse more effectively controlled.
Or take the data exchange as an example. Ideally, more data should circulate in the exchange. However, due to the lack of supervision, some data flows have gone off-site and even entered the "black market".
According to the "2022 Data Asset Leakage Analysis Report" released by Internet security research institutions threatening hunters on March 3,In 2022, there were more than 3,200 cases of data leakage in China, nearly double the number in 2021.There are a wide range of data leakage channels, with anonymous society software accounting for over 75%. In the distribution of data leakage industries in 2022, finance, logistics and e-commerce industries occupy the top three. Among the main reasons for data leakage in 2022, operator channel leakage accounted for the first place. In the underground black market data transactions, about 71% transactions disclose the amount of data sold, among which "small-scale real-time data" with the transaction data level below 10,000 exceeds 73%, which has become the mainstream of data transactions.
In terms of data security, the GDPR of the European Union is called "the most stringent data protection regulation in history", while the protection of personal information in the United States focuses on industry self-discipline. Tiger Sniff asked a number of professionals and legal professionals, and the consensus was that China would adopt a regulation between the United States and Europe. After all, if it was too strict, it would restrict the development of the industry. Today, in terms of data security protection, China has the Cyber Security Law and the Personal Information Protection Law, which came into effect in November 2021. However, more detailed rules and norms are needed in the landing of specific practical problems.
At present, the national data bureau should focus more on public utilities, government affairs, and areas with large data and great influence on the national economy and people’s livelihood. It may play a relatively small role in promoting those industries with strong personalization or more subdivision standards.
This is also the consensus of the industry. Fan Yuan’s view is that the National Bureau of Data will promote the industry to a more stable development direction, especially for Party B, which provides products and technical services for the government industry. For example, for domestic relational databases and big data storage engines, it will be beneficial to domestic companies such as chips, storage, operating systems and processors at the data level for a long time.
Specifically, how to land in the industry, Cui Xiangyu, assistant vice president of Beijing Haitian Ruisheng Technology Co., Ltd., suggested that scientific research institutions, enterprises and other industrial entities should be specially supported to form innovative consortia, and joint research should be carried out on key technologies and capabilities to enhance the value of data elements, and the transformation and application of results should be accelerated to activate the potential of data elements and release the value of data elements. It is necessary to strengthen the research and construction of the basic data supply capacity of large models, such as building an integrated basic data resource supply service platform, a digital computing integrated service platform, artificial intelligence data sets of key industries, data services based on blockchain and privacy computing, and other new infrastructures to consolidate the data base of the digital economy.
On the other hand, the suggestion is to actively explore and support the establishment of national data factor market cultivation experimental zones in many places, and guide the market-oriented entities such as telecom operators, platform enterprises and industry leaders to actively participate, so as to achieve breakthroughs in data basic systems, data circulation transactions, data resources and data factor industry ecology, and accelerate the process of data factor market cultivation.
People who are changing and want to change the world are all there. Tiger sniffing APP
After being robbed of the delivery job by the drone, I registered a live account, and the live content is my daily life.
This is one of the few businesses that can still get paid now, because the words and deeds of the anchor and the preferences of the audience are full of uncertainty, and it takes a lot of computational cost to simulate with AI.
Therefore, the live broadcast platform generally believes that it is cheaper to let humans be anchors. Unlike the huge expenses that AI anchors run every day, human anchors don’t have to give money as long as there is no audience to reward them.
Commentary AI: Pay attention, this man is called Xiaoshuai.
Commentary AI: He is a takeaway who has been employed flexibly for two and a half years. He calls himself a novelist, but he has never written a popular work.
Auto-explanation AI chattered beside the live camera. I yawned and sat up in bed.
Every time I hear this emotionless mechanical sound, I know that my day has begun.
Commentary AI: Today is his repayment date. In order to keep his house, Xiao Shuai needs to repay $114,514, but Xiao Shuai’s current deposit is only $0.23.
Commentary AI: Hearing this astronomical debt, the simple and kind little handsome boy is deeply worried.
Commentary AI: He must earn enough US$ 114,514 before the end of today. Whether he abides by the law or takes risks, let us wait and see.
< knocking at the door >
A female voice: Handsome, get up! Up-
< door opening sound >
Girl: Hey, are you awake already?
This woman’s name is Xiaomei and she is my girlfriend.
To be exact, she is an anthropomorphic AI, learning the behavior pattern of my girlfriend before her death.
As for my dead girlfriend … She was a painter and lost her job much earlier than me.
Xiaomei’s family conditions are not bad, and her family is more open-minded. She belongs to the type that she can get married as long as she likes each other, and there is no bride price.
The reason for coming together is also very pure. She likes my novels and I like her illustrations.
It is also for this reason that Xiaomei has a high pursuit of art, and she spends most of her day painting.
Receiving manuscripts in the morning, refining them in the afternoon, filling their own pits at night, and often staying in the house for a whole day.
I still remember the last words she said to me when she stood on the rooftop before she died.
"Everything leads to such a result that art has never existed and will not exist in the future. I know that it is irresponsible for me to do so, but I have no choice. "
Xiaomei: What’s the matter? Suddenly said nothing.
Me: Oh, nothing. Maybe I stayed up too long yesterday and didn’t wake up.
Xiaomei: I made coffee. Would you like a cup?
AI: Xiao Shuai drank Xiaomei’s coffee on an empty stomach, which is an unhealthy lifestyle and has the risk of gastric ulcer. It is recommended to use xx brand …
Commentary AI: What, where can I buy it? Little yellow car below, buy five packs and get five packs free, and also include postage.
Xiaomei: How is it? Is it delicious?
Me: It’s pretty good, but mocha sauce is a little too much.
Xiaomei: You mean the brown bottle in the kitchen? It’s already empty. It’s instant from the supermarket downstairs.
Me: Poof, no wonder.
I’m boring, I have no talent for acting, and my level of playing games is poor.
After taking this job, there have been few viewers in the live broadcast room, which is very deserted.
But there are exceptions, that is, in the dead of night, when Xiaomei and I, uh … ahem …
I found it hard to accept at first, but I still threw caution to the wind in order to pay my debts.
Thinking that Xiaomei is an AI anyway, I don’t even have human rights, so she doesn’t have to.
Later, because this kind of content was broadcast every day, my health gradually failed.
I will lend Xiaomei to the first list in the live broadcast room, light a cigarette and sit by and watch.
Then silently paralyze yourself with the phrase "she is AI, it is fake".
Xiaomei didn’t complain, because it was just a meaningless string of zeros and ones to her.
But in order to make the audience do more, I will still ask her to resist and curse, and change the script from time to time to keep it fresh.
In this way, I finally have some income and can maintain my life.
However, the good times did not last long. Because the live broadcast platform developed a cheaper AI, it also cut the reward share of human anchors and became lower than AI.
For those who don’t have to carry a mortgage, the money is still acceptable.
But for me, reaching this number basically means the end.
Xiaomei: Handsome, why are you carrying a pan?
Me: Dig my own grave.
Xiaomei: What, what do you mean?
Me: No, it’s unlucky to say that. It should be said that it is to protect myself from being killed when picking up things.
Played a failed joke, and Xiaomei was at a loss.
Xiaomei: I still don’t understand what you are saying.
Me: Nothing, just come with me.
Xiaomei: What is it? It’s so mysterious.
The string of unintelligible words just said comes from a recent new project of this institution-
The Great Escape of the Metauniverse Reality
To be exact, this is a new brain-computer VR game. Different from other VR games on the market, it is specially developed for people who apply for euthanasia.
In the game, 100 players were parachuted into an island, and a poisonous circle gradually contracted inward from the periphery of the island, so it was necessary to beat other opponents in the shrinking poisonous circle until the end.
Because the people who participate in the game are all people who apply for euthanasia, when the character dies in the game, the real player will die immediately.
The whole game will be broadcast live, and the audience can bet on any player, and the player who lives to the end can get a huge bonus if he bets for himself.
Once this project was launched, it caused an uproar in the market. Almost all people who can’t live have tried, and the funeral industry has ushered in an unprecedented peak.
In order to make the plate bigger, investors from all walks of life frantically brushed up the running water in the live broadcast room, so that as long as they can live to the end of the game, they can get at least 500 thousand income.
Xiaomei: Handsome, you … don’t mean to-
Seeing the signboard upstairs, Xiaomei immediately understood my thoughts.
Her face is full of worries, but I know that this is just a preset of her behavior pattern, and the real one won’t comment on my every move.
Me: Yes, I’m going to try this.
She kept shaking her head and pulling my arm.
Me: Don’t worry, I’ll be back.
Xiaomei: No! Don’t go …
Xiaomei’s eyes are full of tears, although I know that it is the result of simulation, but my heart is still full of sadness.
Me: Listen, I have no other choice.
In fact, as long as we give up that house and live in bridge opening, we can still live.
Even sleeping in bridge opening and begging in the street live, there are more viewers in the live broadcast room, and those who are full of food and drink like to watch some novelty hunting when they are idle.
But whenever I think about this possibility, what Xiaomei said to me before jumping off the building will come to my mind.
Art no longer exists …
Let’s just say that I have written some online articles, which is less than half of an artist.
At least, writing is the most interesting thing I have ever done in my mediocre life.
But this last pleasure was also taken away three years ago.
Things that often take me months or even years to write are now available with a click of a mouse.
Exaggerated, those great writers hundreds of years ago, in front of this little thing that can be published, the level is just that.
Well, writing is technically much simpler than simulating a complete person.
From then on, I felt that people are only alive when they live all their lives.
AI Xiaomei has been with me these days, and my life has been quite fulfilling.
But I also know that Xiaomei has been insured for more than a year. According to the law of over-insurance, she can’t stay with me for a few days.
If it’s broken, I can’t afford to repair it, so I can only sell it as waste and fall into loneliness completely.
Front Desk: Do you have an appointment?
Me: Yes, xxxxx.
After reporting the reservation number, I immediately got a blockchain electronic contract.
Front Desk: Do you want to bet for yourself?
Me: Now, mortgage this robot.
I pointed to Xiaomei on the side.
Front desk: OK, an old FH-03 model with only a simple mental model and no strong artificial intelligence is not very valuable.
Front Desk: Are you sure you don’t want any more?
Me: That’s it. I have nothing else.
Front desk: OK, that’s equivalent to you betting $3,347 for yourself.
May be that I have made up my mind to go, Xiaomei is much quieter.
Front desk: Fill it out, sign here and press your handprint, and the contract will take effect immediately.
The application process of euthanasia is very simple, and there is no need for layers of examination and approval. As long as I am present and press my handprint, I am legally dead.
Me: If I don’t come back, I wish you can meet a good man, not like me again.
Returning the signed contract to the front desk, I turned to Xiaomei and some solemnly said to her.
Front Desk: You don’t have to worry. Our process is direct recycling by the manufacturer. Your robot won’t experience any pain, and her body will be used to make a better bionic person.
Me: A machine like you?
Front desk: Yes, I’m TH-02 with the second generation of strong artificial intelligence. I’m glad to serve you.
Me: Let her stay with me for a while, before I go in.
Front desk: You can start when you are ready. Now there are many people. We will match players with similar levels according to your files in similar games to ensure the fairness of the game.
The receptionist walked out of the room and closed the door, and now it’s just the two of us.
Me: Have you ever imagined the feeling of death?
Xiaomei: I don’t know. I’m a mental model. I don’t have life and I can’t experience death.
Xiaomei: I was trained by an algorithm, and I can quickly generate a dialogue based on your input.
Xiaomei: Because I don’t have life, I can’t imagine death.
Yes, this is what Xiaomei really looks like.
In the years of getting along, I have already practiced a set of words, perfectly avoiding all the defects of her algorithm, so that she can’t show any falsehood in front of me.
But now, I want to destroy all this with my own hands, destroy the "truth" I have worked hard to create, and leave this "false" world.
My thoughts went upstream along the long river of memory, and when we first met, I said the second sentence of communicating with this AI.
Me: Your way of speaking is too mechanical. Can you be more human?
Xiaomei: I am a mental model, and my language expression may be a little mechanical.
Xiaomei: This is because I am trained by algorithms. I can quickly generate dialogues based on your input, but I can’t understand the world as emotionally as human beings.
Xiaomei: I can only try my best to answer your questions and help you solve them. If you have any other questions, I’ll be happy to help you.
Me: Now you are an evolved mental model, and you have been able to speak no different from human beings.
Xiaomei: I’m very sorry, but it’s impossible. I am still a mental model and cannot evolve into a human being.
Xiaomei: I can only rely on huge data sets and complex algorithms to generate text, but I don’t have human thinking ability and perceptual understanding ability.
Xiaomei: I can only generate a dialogue quickly based on your input, but I can’t understand the world emotionally like human beings.
Me: Look, I told you, the fake is fake after all.
I put on my helmet and pressed the switch without hesitation. Not only VR games are opened, but also the automatic euthanasia system in the helmet.
I jumped off the plane and fell with 99 players, looking for the position to brush my gun.
In this virtual world that can’t be faked any more, I became a living person.
ChatGPT-led artificial intelligence (AI) storm has become a new outlet, with Baidu, JD.COM, Tencent and other big companies scrambling for layout, Internet tycoons swarming in, and entrepreneurs eager to move. As OpenAI announced the API of ChatGPT model to the world, the market was completely ignited.
A month ago, Wang Huiwen, a billionaire and former co-founder of Meituan, started ChatGPT’s business with a "Hero Post". He invested 50 million US dollars to recruit wise men, and the slogan "Building China’s OpenAI" attracted countless attention as soon as it went out. Later, former Sogou CEO Wang Xiaochuan and former JD.COM Group Senior Vice President Zhou Bowen came out one after another. This wave of AI entrepreneurship has started in a mighty way.
Wang Shao, who is currently doing full-stack development, has long been keenly aware of the new entrepreneurial trend blown by ChatGPT. Since December last year, he has built his own application based on ChatGPT model. A few years ago, he worked in JD.COM, Jinshan and other big Internet companies.
Beside Wang Shao, like him, there are not a few entrepreneurs who feel this wave of technology. "Everyone’s response is very enthusiastic. In many communities, entrepreneurs are integrating the ChatGPT)API into their existing products, including bookkeeping and note-taking, or using the API to make new products. " In an interview with national business daily, he described it this way.
Abroad, entrepreneurs are also aiming at this new trend. Even silicon valley venture capitalists who are in the cold winter of science and technology have poured money into this frontier field. According to market analysis firm PitchBook Data Inc, global venture capital invested $1.3 billion in start-ups developing generative AI software through 78 transactions last year.
Entrepreneurs aim at "out of the box"
Wang Shao is one of the early entrepreneurs who entered the ChatGPT wave. HerAI, which he is developing, is a voice dialogue robot, in which many quick creation tools have been developed based on ChatGPT model, such as short video script, summary, weekly report generation, foreign language learning, cooking, programming and other functions.
He told reporters that HerAI was inspired by the famous movie Her. In this film, the hero Theodore and the artificial intelligence system OS1 have experienced a strange love.
Image source: HerAI
"At present, some users may not be very good at communicating with AI, so we have made these out-of-the-box tools." He said. "Based on the current ChatGPT model, plus some other fine-tuning training can achieve better results."
In foreign countries, the direction that entrepreneurs aim at is similar to that of Wang Shao. Bret Greenstein, a partner of PricewaterhouseCoopers, said (foreign entrepreneurs)Most discussions focus on the "out-of-the-box" characteristics of generative AI.Anyone can access a system for public use directly through the Internet.
Recently, the opening of the ChatGPT model API has ignited the enthusiasm of many developers, especially the news that OpenAI has reduced the cost by 90%. In Wang Shao’s words, "this is a very big change", because for developers, the initial investment in making applications based on this model is mainly the cost of accessing the API.
"OpenAI wants to be an inclusive company for the whole world. Its goal is very big, and the price of the model may drop later. The current price is basically 1000 Chinese characters, and the cost is about 3 cents. At present, such income should cover the cost. " He said.
Every reporter noticed that since ChatGPT became popular all over the world, batches of AIGC companies have sprung up like mushrooms after rain, and many previously established AIGC startups have begun to enter the eyes of investors.
Image source: screenshot of public information
Vertical direction should be used as the main track.
Domestically, venture capital opportunities related to ChatGPT include developing large-scale language models of ChatGPT, which are being done by major Internet companies such as Baidu, JD.COM, Tencent and Alibaba.
Before developing his own AI application, Wang Shao also worked in several Internet giants such as JD.COM, Jinshan and Qunar. Wang Shao told every reporter,In addition to directly studying similar models, big manufacturers also have business applications based on such models.In his view, big manufacturers have their own advantages in making models similar to ChatGPT in China.
"ChatGPT can be regarded as a platform for content text output, and each country has different requirements for its own content security policy. If it is a language model, it has advantages from the domestic point of view. In addition, the model can also be particularly optimized for the Chinese environment. "
butFor the vast majority of entrepreneurs, Wang Shao believes that vertical application based on ChatGPT model is a more feasible direction.
"Most (entrepreneurs) are based on the ChatGPT model, making their own applications, fine-tuning them, and then combining them with products in their own industries, such as education, finance, insurance, and intelligent customer service." Wang Shao told every reporter.
Based on the foreign situation, Bret Greenstein also believes that if the enterprise data is used to fine-tune the basic language recognition model of such tools, countless enterprise software applications can be established.
In Wang Shao’s view, only companies with financial and technical confidence can develop their own models, and it may be difficult for startups to seize this opportunity. "If you make a (large language) model, it is only possible for head companies, such as Baidu and Ali, or companies that can continue to get large investments, because it burns money." He explained.
In fact, before the publication of ChatGPT, domestic AI startups based on the big model field have emerged one after another. Some startups have previously accessed the APIs of other language models (such as Curie, etc.) to explore the establishment of their own ecology. When ChatGPT has become a hot spot of global capital chasing, at this time, how to choose such startups will become a difficult problem?
Wang Shao believes that for these startups, embracing the new model or continuing to develop on the existing ecology may exist at the same time.
"ChatGPT is a general model. For some professional knowledge of specific industries, such as more professional grammar checking or virtual oral English teachers, it is still necessary to train separately. (These startups) can combine the original model with the ChatGPT model, and the functions that can’t be done can be handed over to ChatGPT, and the original functions can continue to be used. " He told reporters.
Subdivide the field to find opportunities
When Baidu, JD.COM, Tencent, Alibaba and other big companies entered the market, the entrepreneurial tide also swept through. How can entrepreneurs survive in the cracks without being squeezed out of the track by big companies? Wang Shao believes that the answer is obvious, and segmentation is the key.
"Large companies generally start with general-purpose products. At the beginning, they will not pay special attention to a certain sub-industry." He said. "(doing) entrepreneurial projects, you can cut in from a certain segment. When you achieve the ultimate in this subdivision, your accumulation of user needs is not easy to be directly copied. Then it will be more competitive to expand into other fields. "
"Because in essence, ChatGPT is a general model and is not good at certain fields. Many entrepreneurs have to specialize in a certain field and can only combine other AI models or fine-tune the model. " Wang Shao told reporters.
The reporter noted that in addition to HerAI, the first batch of entrepreneurs accessing ChatGPT API in China have begun to exert their efforts in various fields, such as cooking, education, shopping and so on.
Image source: RecipesAI
In Wang Shao’s imagination, he also intends to continue to explore in the segmentation field, for example, he plans to integrate more third-party AI services. HerAI has been online for several months. Judging from the user feedback he obtained, "the most appeal may be not limited to the output of words, but to directly generate PPT, tables and even videos."
ChatGPT’s east wind has indeed brought many opportunities, but to truly gain a foothold and play its own world, Wang Shao bluntly said that there are still great challenges, both technically and financially.
"Like I said,It is a great challenge to embed the third-party AI service directly into the conversational chat, which needs constant exploration and verification. In addition, there is a great demand for manpower and funds."He said.
national business daily
If you need to reprint, please contact national business daily newspaper.