Technical and resource background of ChatGPT
ChatGPT is a language model based on big data, which calculates a large amount of data, such as the probability of language occurrence and applicable environment. There are approximately 200 billion language model parameters used by ChatGPT. Since the calculation of ChatGPT requires a large amount of computer resources, it is very difficult to implement it in engineering, which can be said to be a kind of brute force calculation. Mr. Wu Jun once mentioned: "How much electricity does a ChatGPT training consume? It is roughly equivalent to the electricity consumption of 3,000 Tesla electric cars each running 200,000 miles. This training process is very expensive." In recent years, the development of deep learning technology The development has brought great help to language models, making them achieve great success in various fields.
Application areas of ChatGPT
As an AI language model, ChatGPT is best at using language processing technology to communicate and interact with humans, and can answer a variety of questions related to knowledge, information, processes, guidelines, suggestions, etc. In addition, it can also provide technical support in natural language generation, natural language understanding, text classification, entity recognition, etc., and provide corresponding help based on user questions.
In fact, what computers are best at processing are fact-based questions with clear answers. As long as they have the ability to store and search, they can complete a large amount of knowledge reserves. For those complex problems, a large number of calculation rules and logical information need to be integrated. Therefore, for those tasks that have a fixed format, we can use ChatGPT to help us complete it. For example, there is a large amount of imitation content on the major short video platforms, which does not have much nutrition for us. If we use ChatGPT to handle such tasks, I believe it will be better than us humans. Therefore, we still need to engage in content creation and technical in-depth cultivation, and become those who truly explore the mysteries of human knowledge.
Us in the ChatGPT era
Since ChatGPT requires a lot of computing resources behind it, it is likely that only the resource owner will benefit from it. As Ren Zhengfei said, ChatGPT will promote the expansion of data pipelines. I can feel this around me. Almost everyone is watching videos and surfing the Internet. Especially when the rural infrastructure is not perfect enough, I think this will be the next growth point. In the context of national rural revitalization, this field is bound to usher in rapid development.
In the current very hot environment of ChatGPT, we need to calm down, study and use it seriously with an empty cup mentality, and do the things in front of us well.
First of all, when a new thing appears in front of us, we need to understand and learn it, and then think about what it can do for us and how to make our life, work and study more ideal. Secondly, don't be too anxious. Fear often arises from not understanding something. Believe in yourself and with the enthusiasm for learning and mastering new things, you will be able to do it with ease. Because we have what others have, we have nothing to envy and be anxious about.
AI can help us deal with fixed and regular tasks. The future society will need talents who are innovative, diligent in research, able to solve personalized problems and provide spiritual services.
Therefore, we need to improve our ability to learn and explore and provide emotional value. Cultivate children's learning ability, enable them to experience life, protect their innovation and curiosity, and avoid falling into fixed patterns of learning and rote memorization.
Below I will share some of our current experiences in ChatGPT practice for like-minded friends to learn from and refer to. Discussions on technology and prospects are welcome. Our system has a source code demonstration station, if you want to experience it, please feel free to contact us.
technical practice
At work, we have begun to actively deploy our own AI and released the Java version of the ChatGPT system machine "feeding AI" version. We provide WeChat mini programs, official accounts, H5, IOS, Android and Web multi-terminal versions.
The technology stack we use is as follows:
Java background technology uses the renren framework: springboot+mybatis-plus+redis
management front-end vue +elementUI
mobile client: uniapp (vue3)
pc client: vue3
The functions we support include:
1. Home page
AI chat, streaming output, new chat window, AI drawing, model switching, sensitive word filtering
2. Role models:
There are various AI chat models such as investment consultants, accountants, legal consultants, medical consultants, learning consultants, translators, marketing planners, etc.
3. AI drawing:
Wenshengtu, Tushengtu, Meitu Square, secondary creation, prompt word translation expansion, poster generation
4. The toolbox has
more than 200 chat dialogue templates. You can collect templates and create your own templates, which is convenient and fast.
5. My
Q&A records, card secret redemption, my team, check-in, inviting friends, cash withdrawal, membership activation, purchasing Q&A coins, upgrading partners, commission withdrawals, Q&A coin turnover, about us, withdrawal records, team member list
H5 interface
PC interface
