Recently, I got the internal test permission of Keling, a big model of fast AI video, and made many attempts. Especially after opening the graphics video, Keling has very high availability value.
At the same time, Sora, which was released in February, is still drawing PPT. In the field of open source big models, Ali's Tongyi Qianwen 2.0 has become the top in the world.
From the practical application in recent months, AI technology really has no moat. Better than OpenAI, the first mover advantage is also getting weaker.
At the press conference, Apple announced that it would integrate several AI tools such as chatgpt, and it may cooperate with Wenxin in China. Some netizens worry that the experience will be different if they change to domestic AI.
However, Xingkong Jun carefully studied Apple's AI function, and found that it was nothing more than automatic extraction of ID image information (Shunfeng has long been implemented), automatic summary of email summary (nail conference summary has been used), automatic audio transcription of text and semantic summary (nail, Tencent conference, iFLYTEK said: this is it), AI generation of expression packs (bad street), etc.
Players like Star Kong Jun, who use AI in a high intensity, can responsibly tell everyone that most domestic AI can do even better.
Compared with the frequent use of ladders to visit Chatgpt and ClaudAI last year, this year it is basically enough to use domestic Kimi, Tongyi Qianwen and ChtGLM. In terms of research report analysis, semantic summary, etc., the domestic AI is basically the same as Chatgpt4 or even partially surpasses it. Occasionally, the use of ladders is mainly to visit the C station to download AI painting models.
Recently, two news has attracted the attention of the Star King:
First, OneThing founded by Li Kaifu announced that the company has successfully developed a new vector database "Descartes" based on the full navigation map, and has won the first place in the evaluation of six data sets in the authoritative list ANN Benchmarks.
Second, OpenAI announced the acquisition of Rockset, a database analysis company. Coincidentally, Rockset is also developing vector databases.
The so-called vector database is the vectorization of text, voice, image, video, etc. Compared with traditional databases, vector databases can handle more unstructured data (such as images and audio). In machine learning and deep learning, data is usually expressed in vector form.
Compared with traditional database applications, the data stored in AI is mainly unstructured data, so it is more suitable to use vector databases.
In terms of domestic databases, Tencent has just released cloud native vector databases, and GBase has also released vector databases.
With the development of AI technology, the database also ushered in a new era of letting a hundred flowers blossom. Traditional databases such as Oracle have gradually withdrawn from the historical stage.
01
Damon Data Technology Innovation Board was listed
Many young developers now think that domestic databases such as Damon and Jincang are simply modified and encapsulated on the basis of open source databases, so they can be directly adapted with components of Oracle and other databases.
In essence, it is still the same sentence: as soon as the source is opened abroad, it will be self researched at home.
What they didn't understand, however, was that the earliest version of Dream Database CRDS came out in 1988, even before MySql.
During the period of Sino Japanese friendship, Japan supported China's construction of WISCO.
In 1978, Japanese technicians installed equipment in the hot rolling workshop of WISCO. Before they left, they destroyed all technical data, including database software, which amounted to three trucks. At that time, a young teaching assistant in the computer department of Huazhong Institute of Technology happened to study at WISCO. Seeing this scene, she was shocked.
Since then, the young man has determined to develop a database that he can master.
He is Feng Yucai, the founder of Damon Data. When Damon Data was listed, he was 80 years old.
According to the prospectus of Damon Data, the company serves China Construction Bank, Bank of Communications, Everbright Bank, Industrial Bank, China Guangfa Bank, China Development Bank, China Life Insurance, Postal Savings Bank, PICC, State Grid, China TravelSky, China Mobile, China Tobacco, the State Market Supervision Administration, people's procuratorates at all levels, people's courts at all levels, the National Development and Reform Commission Well known users, including the National Immigration Administration, China Securities Regulatory Commission, Shanghai Stock Exchange and Shenzhen Stock Exchange, have been successfully applied in dozens of fields, such as finance, energy, aviation, communications, party and government organs. According to the report released by CCID Consulting and IDC, the market share of the company's products from 2019 to 2023 will be among the top domestic database manufacturers in China's database management system market.
It can be seen that China's core lifeline industries and databases are firmly in their own hands.
However, according to Xingkong Jun, the market share of excellent domestic databases such as Oceanbase and TiDB is also high, and the current domestic databases are in the bloom stage.
Combined with the current wave of domestic substitution, we can almost infer that the market space of Dream will further explode.
02
How rough is the self controlled track
Xingkong Jun often told his friends that life is a wilderness, so don't stick to plans, plans, assumptions, etc.
In the field of investment, the track of self control is a boundless wilderness.
Five years ago, Star King advocated the new energy vehicle track; Five years later, the Star King began to advocate independent and controllable track.
Autonomy and controllability is not just a matter of state-owned enterprises buying a high priced domestic computer. With the promotion of big data exchanges, data asset entry and other platforms and application policies, China has opened an unprecedented path of digital economy.
The foundation of this road is autonomous and controllable. The database of the railway system has been changed from Sybase to self-developed database, and the database of two barrels of oil has been changed from Oracle and Myssql to Oceanbase, Jincang and Dameng
In order to make safety assessment more convenient, private enterprises docking with these platforms also began to gradually migrate to domestic systems, thus gradually driving the whole industry to use domestic autonomous and controllable systems.
According to the prospectus, Dameng's performance has experienced ups and downs. After reaching the "peak" in 2021, it will decline in 2022.
Data source: Flush iFind Drawing: Poetry and the Starry Sky The main reason is that in 2021, a wave of localization substitution process was carried out in central enterprises and institutions. Due to the large base, there will be a decline in 2022. However, based on the company's own advantages, the performance in 2023 will enter the growth channel again.
03
AI will restructure all traditional businesses
Although Starsky doesn't think AI has any threshold, the application of AI will bring revolutionary changes, that is, AI will restructure all traditional businesses. The process of restructuring is a huge opportunity for enterprises and investors.
For example, Xingkong Jun participated in the AI transformation of a traditional information system, replacing the manual copying of the document entry system with AI+RPA (the domestic ChatGLM used for the base is ready to switch to Qwen2.0 later).
The database industry will also encounter such changes. AI can change the manual operation of traditional database into automatic operation.
In the face of large-scale data and different application scenarios, traditional database components are not sensitive to business types and have weak query optimization capabilities. At present, there are researches on replacing traditional database components with machine learning algorithms to achieve higher query and storage efficiency and automate various tasks, such as automatically managing computing and storage resources, automatically preventing malicious access and attacks, and actively implementing intelligent database tuning. Machine learning algorithms can analyze a large number of data records, mark abnormal values and patterns, help enterprises improve security, prevent intruder damage, and automatically, continuously, without manual intervention, perform repair, tuning, backup and upgrade operations when the system is running, so as to reduce human errors or malicious acts as far as possible, and ensure the efficient and safe operation of the database.
