ChatGPT is an artificial intelligence language model based on natural language processing and deep learning technology, which is a very large neural network model with billions of parameters. Therefore, the training and deployment of ChatGPT requires a lot of computing power. From the perspective of the training process of ChatGPT, it is necessary to use large-scale training datasets and high-performance computing devices for training, which requires a lot of computing power and storage resources. At the same time, a large number of calculations and parameter optimization are required during the training process, and efficient algorithms and computing platforms are required.
The training of ChatGPT models requires a large amount of computing power and storage resources. Specifically, training a large ChatGPT model can require thousands to tens of thousands of GPUs and hundreds to thousands of petabytes of storage capacity. For example, the latest GPT-3 model is a giant pre-trained language model with 175 billion parameters, and its training requires thousands of TPUv3 model accelerators for training, as well as large-scale distributed storage and computing cluster support. According to reports, the training cost of GPT-3 models can be tens of millions of dollars or more.
From the perspective of the deployment and application process of ChatGPT, the trained model needs to be deployed to the server and needs to support high-concurrency requests. For large-scale ChatGPT applications, it may be necessary to deploy multiple ChatGPT models to cope with highly concurrent requests, which requires efficient computing power and server clusters. The development of ChatGPT has a great impact on the future demand for computing power. In the future, with the development of artificial intelligence technology and the expansion of application scenarios, the demand for computing power of ChatGPT will continue to increase. This requires us to evolve more efficient algorithms and computing platforms to meet the training and deployment needs of ChatGPT.
First of all, let me elaborate on my views on the current development of artificial intelligence in the mainland:
Artificial intelligence big model only big factory play, because manpower and material resources are huge investment, domestic large models to Baidu's "Wenxin big model" and Huawei's Pangu big model, Tencent, Ali, 360 are involved, but these companies have little to do with A shares, only one 360 can be. But for other application companies are clearly bearish, but does not affect their hype, these application companies without cooperation with chatgpt, can be regarded as hype, good technology wind high how to participate, really does not matter, make money on the line, I do not have that ability to participate.
This is also artificial intelligence concept stock, in the past 5 years, they have been engaged, but there is no result, before chatGPT came out artificial intelligence has not seen any closed-loop cases of commercial success, but after chatGPT came out, they these companies shouted that I can also do it, I am very technical and very good, do you believe it? Bullish on their future, then optimistic about which companies developing applications, only optimistic about many application companies in the United States that cooperate with chatgpt, that is, companies with chatgpt this platform have potential, and those without chatgpt are not optimistic, can be seen as hype.
Represents personal views only and is for informational purposes only
Here are the core stocks:
1, computing power chip chiplet stacking scheme
Tongfu Micro Battery, VeriSilicon Co., Ltd., Changjian Technology, SMIC
Here is optimistic about Tongfu Micro Power + Huawei Ascend AI chip chiplet is also the only solution for the return of HiSilicon stacked chips this year
2. Computing power server
China Army Inspur Information Ascend AI Computing Power Topway Information Plus Zhongke Dawn
In total, these three computing cores, you can configure two evenly, so that you can deal with it more calmly
3, CPO low power consumption, low cost and high energy efficiency solution
Cambridge Technology, Tongyu Communication, Guangxun Technology, Tianfu Communication, Zhongji Xuchuang, Xinyisheng
4. The combination of chatGPT and bots
Mingzhi Electric Double Ring Drive Zhongda De Sanhua Intelligent Control Green harmonics
5. Computing power construction provider
China Telecom includes Tianyi Cloud, China Unicom, China Mobile, Alibaba Cloud, HUAWEI CLOUD, Tencent Cloud, and Baidu Cloud
In summary, describe the core of artificial intelligence in a paragraph:
With the chip stacking solution of Tongfu Microelectronics, to help Huawei's Ascend AI chips, Kunpeng chips, HiSilicon chips to build computing power servers need AI chip supply, in the computing power server provider Topway Information Inspur Information With the support of Sugon, China Telecom, Alibaba Cloud, Huawei Cloud, etc. provide computing power for large model companies, build China's chatGPT artificial intelligence platform, and cooperate with robot companies on artificial intelligence platforms such as chatGPT. Robots with high intelligence (including humanoid robots) are manufactured, and then these artificial intelligence robots are combined into smart factories made by artificial intelligence automatic manufacturing, resulting in an exponential increase in productivity and greatly reducing the cost of products.
ChatGPT is a very powerful natural language processing model, which can perform natural language generation, understanding, dialogue and other tasks, and at the same time the improvement of programming efficiency is also huge, such as:
Code auto-completion and error correction: ChatGPT can autocomplete code and correct code errors by analyzing the context and syntax of code. This reduces tedious and repetitive work in programming work and improves programming efficiency and code quality. Generation of documents and comments: ChatGPT can automatically generate code documentation and comments by analyzing the structure and semantics of the code. This improves the readability and maintainability of the code, and reduces human errors and vulnerabilities in programming.
Intelligent code recommendation and optimization: ChatGPT can automatically recommend best programming practices and optimizations by learning and analyzing a large number of code samples and developer behavior. This helps developers better understand programming languages and frameworks, and improve programming and code quality. In contrast, AGI needs the ability to be able to have intelligence in multiple domains, not just language processing. Realizing AGI requires solving many complex problems, including knowledge representation, reasoning, planning, learning, perception, control, etc., and due to the excellent experience of chatGPT, more powerful ChatGPT 4.0 will be launched soon, and we may achieve AGI in the next few years.
I will update the latest digital economy, big data AI, 6G satellite, data elements, new medical infrastructure, integrated die-casting, automotive intelligence, lidar, HUD, automotive chip, air suspension, L3 intelligent driving, PET copper foil, nano battery, 800V high voltage, photovoltaic HJT, TOPCON, perovskite, photovoltaic XBC, BIPV, IGBT chip, silicon carbide SIC, CTP/CTC/CTB battery, 4680 battery, Industrial mother machine, sea wind flexible straight high voltage, new energy vehicle high voltage fast charging, high nickel ternary, carbon fiber, PET aluminum foil, PET copper foil, air source heat pump, new materials, traditional Chinese medicine innovative drugs, traditional Chinese medicine formula granules, rural revitalization, lithium mine, vanadium flow battery, sodium-ion battery, distributed energy storage, centralized energy storage, antigen detection and other latest topics hot mining, the future belongs to the structural market with high expectations, grasping the core track and the inherent value logic of individual stocks is fundamental.
— END —
Like it before you look at it, and make it a habit
Disclaimer: Pictures, data from the Internet, reprinted only for communication and learning, if there is a copyright problem, please contact the author to delete