Recently, Zhongke Jiahe, a domestic artificial intelligence software infrastructure, announced that it has completed tens of millions of yuan in seed round financing, led by BV Baidu Ventures. Finally, the east wind of the era of computing power has blown to the field of computing power software~
If you still don't understand what computing power software is?
You can see a little chestnut.
Through computing power software optimization, each domestic chip/Nvidia limited version chip (represented by the A800 and H800 with limited functions) can be integrated into the ecology, and it can also overtake in curves in the era of large models.
At present, the artificial intelligence software and hardware ecology is built around NVIDIA in an all-round way, with strong external dependence.
The era of computing power is coming faster and faster, and the international situation is becoming more and more uncertain. The challenges of localization of my country's computing power ecology and large-scale model implementation are not so big.
Among them, at the level of computing power hardware, domestic computing power is fragmented and the software ecosystem is not perfect, which leads to high user costs and low availability of chips;
At the AI application level, the surge in demand for computing power and the differentiation of vertical applications have put forward more urgent and stringent requirements for hardware-side performance iterations and overall software optimization; at the same time, operator fragmentation and long tails have also exposed Chip manufacturers have shortcomings in terms of corresponding support capabilities.
These problems all point to a common solution—they all need computing power software to optimize the performance of the chip.
And we are gathering experts and mentors from National University of Defense Technology, Chinese Academy of Sciences, Central South University, Hunan University, etc., to cultivate computing power software talents. For a long time, some netizens who don’t know much about this technology often wondered about our motives. Today, we can systematically explain what our team is doing.
The emergence of ChatGPT has accelerated the artificial intelligence industry from the earliest "make it work" era to the "make it perform" era. It is foreseeable in the future that there will be countless large models and exclusive models, so how to reduce the comprehensive landing cost and enter the era of "make it cheap" is the next structural opportunity for the AI industry.
What are structural opportunities?
To put it simply, it is to consider what drives this industry? Who contributed the most? This is a structural opportunity. Computing power software is undoubtedly an important structural opportunity for our country.
Those who don't want to lose need a structural opportunity to take off. The same goes for countries.
In recent years, the state has attached great importance to the construction of a unified ecosystem of independent controllable domestic basic software and operating systems. The 14th Five-Year Plan clearly listed "independent controllability" as a necessary requirement for the strategy of manufacturing a powerful country. Then the generalized, low-cost, high-performance basic software tool chain is an important part of the domestic chip industry ecosystem.
The tool chain is not only in the potential track of rapid development, but also has a technical threshold. Compared with the relatively fierce competition in models, computing power, and upper-level AI applications, it is relatively in a non-red ocean state.
This is one of the fields that our high-performance computing and artificial intelligence R&D talents will be engaged in.
With the continuous development and iteration of large-scale models and domestic computing hardware in the future, the relevant computing software and computing software talents will surely show extremely high scarcity value in reducing the cost of model implementation and the difficulty of hardware application, and become the AI industry chain The key card holders.
Lei Jun's recent annual speech said that "many people's dreams are ignited when they are young because of a book, a movie, or a person." For us, we are very lucky in this uncertain era. What to do, where to go.
The Whampoa Military Academy is committed to becoming a high-performance computing talent, helping to solve the bottleneck problem. This is our vision and mission. We took it seriously and tried to break it down into achievable goals one after another. And you?